J. Al-Rafidain Engineering Vol.18, No.4 (2010)

2

Al-Rafidain Engineering Vol. 18 No. 4 August 2010

ENGLISH SECTION

CONTENTS

No. Title PageNo.

1. Performance Measurements of a Wireless Internet Service Providing (WISP)System.Qutaiba I. Ali, Ahmed Z. Saeed Al-wattar.

1

2. Real Time Implementation Of Fir Filter Based On Time Delay NeuralNetworks.Dr. Shefa Abdulrahman Dawwd

17

3. Design And Realization Of Circular Contourlet Transform.Dr. Jassim M. Abdul-Jabbar, Hala N. Fathee.

28

4. Effect Of High Temperature On Mechanical Properties Of ConcreteContaining Admixtures.AHMAD,A.H., Abdulkareem,O.M.

43

5. Finite Element Analysis of Unreinforced Masonry WallsMohammed S. Mohammed.

55

6. Investigation of Handoff Algorithms for GSM Mobile Cellular Networks.S. A. MAWJOUD, H. A. Al-TAYYAR.

69

7. The effect of drawing ratio in deep drawing process on thickness distributionalong the cup.Dr.A.D.Younis.

81

Ali: Performance Measurements of a Wireless Internet Service Providing (WISP) System

1

AbstractIn this paper, a comprehensive practical test to a Wireless Internet Service Provider

(WISP) system is presented. The structure of the system is investigated and its performanceis measured for different time units ranging from a Day to a Year periods. The effect of thedifferent parameters on the system behavior is determined and their contribution isdetected. The goal of this study is to give a realistic picture to the behavior of such a systemas a result of its users demands, network infrastructure and service managementtechniques.

Key word: Wireless Internet service Provider, Latency, Throughput, WLAN Delay, andUser Behavior

/ .

/

. . ,

. : ,.

Performance Measurements of a Wireless Internet ServiceProviding (WISP) System

Qutaiba I. Ali and Ahmed Z. Saeed Al-wattarComputer Eng. Dept. / Mosul University/Iraq

Received 3/5/2009 Accepted 16/9/2009

Al-Rafidain Engineering Vol.18 No.4 August 2010

2

1. INTRODUCTION

Recently, Wi-Fi has become one of the most popular standards for wireless Internet accesstechnology. Using radio frequency connections between a base station and devices with add-onor built-in 802.11 wireless cards, Wi-Fi gives access to the Internet and remote corporate andpersonal data without using the wires and cables of conventional wired networks in public places,homes, and offices. The global push to adopt 802.11 is based largely on its high bandwidth (up to54 Mbps) and rich user experience that is comparable to being on a wired company LAN. Thisstandard is open, unlicensed, internationally adopted, interoperable, and supported by everymajor player in the wireless LAN industry. WLAN options are available today for mostconsumer devices and the current technology of laptops, handheld PCs and PDAs are WLANenabled [1]-[4].There is a growing literature on wireless traffic measurement and Internet protocol performanceover wireless networks. For example, Tang and Baker [5]-[6] discuss wireless networkmeasurements from two different environments: a metropolitan area network, and a local areanetwork. Balachandran et al. [7] report on network performance and user behavior for generalInternet access by several hundred wireless LAN users during the ACM SIGCOMM conferencein San Diego in 2001. They find that for this set of technology-literate users a wide range ofInternet applications are used, user behaviors are diverse, and overall bandwidth demands aremoderate. Kotz and Essein [8] characterize campus-wide wireless network usage at DartmouthCollege, focusing on infrastructure mode using access points. G. Bai et al. [9] focuses on theperformance of standalone wireless Web servers in short-lived wireless ad hoc networks. M.Narbutt et al. [10] experimentally investigates the relationship between resource utilization inWLANs and the quality of VoIP calls transmitted over wireless medium. Specifically theyevaluate how the amount of free bandwidth Influences transmission impairments (i.e. delay, loss,and jitter) and thus call quality.Our paper extends the work given in the above references to present comprehensivemeasurements covering all aspects and parts of a wireless internet service providing system. Themain contribution of this work relies on the long term measurements made during a year period(2006-2007). The data given here could be used by the planners, developers and researchers ofsuch a sophisticated network field.

2. SYSTEM DESCRIPTION

Internet services entered Iraq primarily at 1999. In the beginning, it was subjected to thegovernment control and has limited services and speed. The main Internet distribution technologywas the telephone wire modems with very low bit rate (under 1kbps with high noise &attenuation) and hence downloads speed. After 2003, many private companies began to establishtheir own Internet service providers based on connecting the Internet through Satellite modemsand wireless LAN (IEEE802.11) distribution network. The WISP system under study isconsidered the major internet service provider in the city of Mosul/Iraq. Its establishment beganin the second half of 2003 and witnesses a wide expansion, in terms of coverage area and numberof users, since then. The system gave various services to its clients, such as, Internet browsing, E-mail services, web hosting, HTML pages design, distributed gaming and numerous commercialactivities.


3

From the technical point of view, home Internet services providing is considered to be the majoractivity of the system. It used multiple wireless/wired network techniques to cover the differentareas of the city. Fig. (1) Gives a clear picture of the network

a)

(b) (c)

Fig. (1): Structure of WISP System: (a) ISP (Management) Center (b) Distribution Network (c)The Whole System

The network spans about (25×25 Km) of the city area. The major elements of the system are thedistribution network and the ISP center.

The distribution network: It is the various links between the clients and the ISP center. Thetraffic to/from the center is forwarded through three pairs of point to point bridges, we calledMajor Access points (MAP); each one of them is responsible of covering certain areas of the city.The wireless links between MAPs are subjected to IEEE802.11g WLAN standard, running at aspeed of (22 Mbps). It is worthwhile to mention that two of the MAPs are connected through a

To the NetworkInfrastructure


4

Fast Ethernet LAN to the third MAP, which is the gateway between the WLAN communities andthe ISP center. The rest of the network consists of local access points (working at 11 Mbps datarate of IEEE 802.11b WLAN standard) installed in different sections of the city to serve certainnumber of users. The local access points were segmented into different VLAN groups to preventinter-traffic between them. Fig. (2) Shows the installed access points with their relative users.

Fig. (2): Installed access points with their relative users.

It is clear that local access points have different number of users with an average of (63client/AP). When the number of clients exceeds certain limit, additional access points areinstalled to serve the growing number of users. The coverage area of each access point ismaintained through the use of proper antennas and if necessary, suitable signal boosters.

The ISP Center: It consists of various network devices and servers to manage and provideinternet services. Their functions could be summarized as follows:

Satellite Modem: It is the link between the ISP and the rest of the internet. It has a data rates of(15 Mbps) for download and (3 Mbps) for upload operations.

1-1 NAT Router: The system has an (1500) public IP addresses pool available to the clients onthe one to one Network Address Translation basis (i.e., each public IP is given to a private IPaddress during user log in time).

A Firewall supported with Intrusion Detection System (IDS) capabilities provides certain levelof security to the system. The company’s web server is connected to the demilitarized zone(DMZ) portion of the firewall device. The firewall allows limited access to the DMZ, but becausethe DMZ only includes the public servers, an attack there only affects the servers and does notaffect the other inside networks [3].

AAA Server: AAA is the acronym for Authentication, Authorization, and Accounting.Authentication controls access by requiring valid user credentials, which are typically a usernameand password. Authorization controls access per user after users authenticate. Accounting trackstraffic that passes through the security appliance, gives the ability to have a record of user activity[3].

Web filtering server together with the cache server frees more bandwidth for the benefit of themost important needs. As known [4], these servers have a great influence on the systemperformance and our measurements shows that a Hit Ratio of 0.35 is achieved (i.e., 35% of thetraffic is acquired from the cache server rather than from the internet). On the other hand,filtering server frees 40% of the available bandwidth from the less important applications. Itsdecision is based on Content plus Database filtering algorithms according to pornography,spyware and viruses categories.

Game server provides distributed game network between network clients.


5

Bandwidth Management Server: It is responsible for giving a certain level of service to thedifferent clients according to their individual subscriptions. Each client is given a certain amountof bandwidth using several bandwidth management techniques. Also, many subscriptions have apre defined amount of transferred data per month. Table (1) lists the various types ofsubscriptions.

TABLE (1): SUBSCRIPTION TYPESSubscription Type Number of Clients

1.2 GB 64 Kb/s Home 1-PC 2051.2 GB 64 Kb/s Home 2-PC 821.2 GB 64 Kb/s Home 3-PC 1602.4 GB 64 Kb/s Home 1-PC 3992.4 GB 64 Kb/s Home 2-PC 1462.4 GB 64 Kb/s Home 3-PC 2923 GB 160 Kb/s Home 1-PC 2196 GB 160 Kb/s Home 1-PC 99 GB 160 Kb/s Home 1-PC 16 GB 256 Kb/s Home 1-PC 21.5 GB 160 Kb/s Home-Hours 1-PC 243 GB 160 Kb/s Home-Hours 1-PC 21364 Kb/s Café-IP-Unlimited 5128 Kb/s Cafe-Band 23160 Kb/s Cafe-Band 48192 Kb/s Cafe-Band 8224 Kb/s Cafe-Band 5256 Kb/s Cafe-Band 1600 MB 64 Kb/s Home 2-PC 1600 MB 64 Kb/s commercial 1-PC 51600 MB 64 Kb/s commercial 2-PC 11600 MB 64 Kb/s commercial 3-PC 181.5 GB 64 Kb/s commercial 1-PC 181.5 GB 64 Kb/s commercial 2-PC 11.5 GB 64 Kb/s commercial 3-PC 11.5 GB 160 Kb/s commercial 1-PC 423 GB 160 Kb/s commercial 1-PC 54.5 GB 160 Kb/s commercial 1-PC 43 GB 160 Kb/s commercial 1-PC Full-time 316 GB 160 Kb/s commercial 1-PC Full-time 19 GB 160 Kb/s commercial 1-PC Full-time 13 GB 160 Kb/s commercial-Hours 1-PC 393GB 160 Kb/s Cafe-IP 5 to 9-PC 93GB 160 Kb/s Cafe-IP 10 to 14-PC 13GB 160 Kb/s Cafe-IP More than 14-PC 19600 MB 64 Kb/s Home 1-PC 5Total Number of Clients 2104

The WEB Server contains the web site of the company and hosted web sites for someindividuals.

At last, Domain Name Server(DNS) and Dynamic Host Configuration Protocol(DHCP) serversprovide their usual tasks , such as translating IP addresses to a domain names (DNS function) andsupplying the connected users with their corresponding IP addresses(DHCP function)[4].The last issue to discuss in this section is the Service Level Agreement (SLA) of the system. Itcan be defined as the level of the services given to the user as given in the contract [6]-[7]. Forthe current system, SLA has the following criterion:


6

Average web page response time (Time needed to completely download a web page [8]) lessthan or equal to (14 sec.) for 80 % of the time.The bandwidth given to a user is full for 80% of the time.Average WLAN Latency 100 msec.System (or network) failure rate (Availability) 12 day/yea

3. LOAD MEASUREMENTSIn order to discover the real load applied to the system, a comprehensive load tests were made.The tests were implemented at different points on the network to give the maximum level ofknowledge about the system. These points are: the satellite modem, content (http traffic) filter,game server, MAP1, MAP2, MAP3, a highly loaded local access point (AP4) and a lightlyloaded access point (AP28). The data were collected for a year, month and day periods, as shownin Figs (3 to 5).

(a)

Fig.(3) :Yearly load of the system: (a)Satellite Modem (b)http filter (c)Game server (d)MAP3(e)MAP2 (f)MAP1 (g)AP4 (h) AP28

(a) (b)

(c) (d)

(e)(f)

(g) (h)


7

Fig.(4) :Monthly load of the system: (a)Satellite Modem (b)http filter (c)Game server (d)MAP3(e)MAP2 (f)MAP1 (g)AP4 (h) AP28

(a) (b)

( c )(d)

(e) (f)

(g) (h)

(a) (b)


8

Fig.(5) :Daily load of the system(11/08/2007): (a)Satellite Modem (b)http filter (c)Game server(d)MAP3 (e)MAP2 (f)MAP1 (g)AP4 (h) AP28

The following remarks could be extracted from the above Figures:1. The yearly average value of the downloaded traffic occupy (58.5%) of the available (15 Mbps)bandwidth, while it consumes (60%) of the available (3 Mbps) upload bandwidth. This resultindicates the possibility of successful future expansion which is expected due to the loadincrement throughout the year.2. Http traffic occupies (85%) of the whole download bandwidth. Other applications are: filetransfer applications (FTP), E-mail applications (SMTP), Domain Name Server Protocols (DNS)and Chatting protocols which share the remaining bandwidth.3. The resultant traffic load applied to the system changes periodically between a low load period(16% of the available download bandwidth, from 4 to 10 am) and high load periods (78% of theavailable bandwidth) for the rest of the day.4. The game server traffic witnesses a notable increase during the summer holiday as comparedto other seasons. Also, a similar daily load distribution to that mentioned earlier is noticed here.

( c ) (d)

(e) (f)

(g) (h)


9

5. The traffic directed from the infrastructure network is gathered via the three major accesspoints and forwarded to the ISP center. In addition to the traffic received from the major accesspoint (MAP3), ISP center receives an extra load from some local access points connected to itdirectly.6. The average throughput of the local access points indicates an average user throughput of (4kbps).Not far from the above results is the statistical data gathered about the number of live (active)users of the system. This factor has a direct impact on the load applied to the network. Fig. (6)Illustrates the number of active users during different time units.

(a) (b)

(c)Fig. (6): Number of Live Users: (a) Yearly (b) Monthly (c) Daily

4. LATENCY MEASUREMENTS

In order to investigate the different contributors of the total network response time, systemlatency is measured. This is done using a PING command originating from different sources inthe network to a certain server in the ISP centre during 3 days period, see Fig. (7).The purpose of these experiments is to determine the real reason behind performance change indifferent situations. We chose two local access points, a high load access point (AP4) and a lightload access point (AP28). These access points serve different number of clients and have


10

comparable path lengths to the central ISP. Four latency measurements are done: from localaccess points to ISP centre (PING1 & PING2), from a HOST to ISP center (PING3) and fromISP center to the internet (PING4). The latency values result from the different measurements areseen in Fig. (8).

Fig. (7): Network Setup of Latency Measurements

(a)


11

(b)

(c)Fig. (8): Latency Measurements (a) PING1 & PING2 (b) PING1&PING3 (c) PING4

In these experiments, latency was measured as the summation of three parts:1. from a client to the ISP center: One of the hosts is configured to send continues PINGcommands to the ISP center. Fig. (8b) shows that this latency has an average value of (137 msec.)and subjects to the daily load distribution mentioned earlier. A comparison of this latency valuesto these of Ping1 (from access point to ISP centre) shows a higher contribution of this part of thenetwork in the total latency budget. This is caused mainly by the contention nature of theCSMA/CA protocol (working in Distributed Coordination Point (DCF) mode) and itsperformance dependence on the number of served clients.2. from local access point to the ISP center: The average values of this latency are (5 msec.) forthe lightly loaded access point (AP28) and (17 msec.) for the highly loaded access point (AP4),see Fig. (8a). It is obvious that this latency value is considered as a minor contributor anddepends mainly on the number of contestant clients related to each access point. The relativelylow latency in this section of the network indicates the successful planning of the installeddistribution network.


12

3. from ISP center to the Internet: We chose Yahoo server as the destination in this experimentbecause of its popularity (it was the Internet home page for most of the users). The averagelatency value in this case is (885 msec.) and caused mainly by the effect of the propagation timeto the geostationary satellite. The noted fluctuation in the latency results from various loadconditions during test time, see Fig. (8c).Total two way network latency could be calculated as the summation of both (ping3 and ping4)latencies and have the average value of (1022 msec.).To complete the picture, another test is done, page response time. Yahoo web site is chosen againto be the tested sample because of its popularity and the objects richness of its web pages. Thepurpose of this test is to assure system fulfillness to its service level agreement criteria mentionedearlier. The test procedure lasts for 24 hours and includes configuring one of AP4 (high load)clients to repeat the download operation of the page with and without the use of the cache server.The properties of the web page under test are listed in table (2) and the page response time isshown in Fig. (9).

TABLE (2): YAHOO WEB PAGE PROPERTIES

Fig. (9): Page Response Time Measurements

The average value of the page response time is (12.6 Sec.) when using cache server and (25.6Sec.) without it. This is a clear indication to the importance of web cache technique and its greatinfluence on the system performance. Also, Fig.(9) assures the ability of the WISP system torespond successfully to the SLA criteria.

5. INVESTIGATING USERS BEHAVIOR

As shown in earlier, http was the major internet application and its traffic occupies most on theavailable bandwidth. This result gives a motive to discover the user behavior relating to thisapplication. The first measure to be considered is the most visited web sites. In this manner, weused two metrics to evaluate a web site: the amount of the downloaded data from a site in atypical day and the number of TCP connections made to that web site. Tables (3 and 4) list thetop 50 sites according to the above criteria.

No. of Objects 23No. of Image Objects 17`No. of Java Script Objects 4No. of Style Sheet Documents 1No. of Flash Objects 1HTML Code Size(kBytes) 133Total Page Size(kBytes) 566


13

TABLE( 3): TOP 50 SITES WHICH REPRESENT 82% OF THE DOWNLOADED DATAIN A ONE DAY PERIOD (11/08/2007)

Accessed Site Download Size(MByte)au.download.windowsupdate.com 2000msgr.dlservice.microsoft.com 1000us.js2.yimg.com 754l.yimg.com 480us.i1.yimg.com 444mi.adinterax.com 423www.6rbtop.com 401ads.yimg.com 386us.dl1.yimg.com 373mail.yimg.com 367insider.msg.yahoo.com 334www.google.com 294www.6rb.com 289pagead2.googlesyndication.com 284www.yahoo.com 234tbn0.google.com 231kh.google.com 222m1.2mdn.net 208mosul4all.com 205www.mosul4all.com 199download.windowsupdate.com 198www.microsoft.com 153f3.yahoofs.com 15114.mihd.net 148d.yimg.com 145www.almawsil.com 139alrafidenland.com 137.6server2.mp3quran.net 130content.yieldmanager.edgesuite.net 117www.7shasha.com 114ad.z5x.net 110images.google.com 100pms.panet.co.il 92www.youtube.com 90www.aljazeera.net 89.5ads.adbrite.com 82.6www.bokra.net 82content.sweetim.com 79messenger.yahoo.com 74ad.yieldmanager.com 74video.google.com 71mail.google.com 70www.6rbwow.com 69www.mosulalhob.net 63www.aljazeerasport.net 63img.youtube.com 62www.graaam.com 59.5forum.amrkhaled.net 56.6www.google-analytics.com 54www.hawar11.com 50

http://:@www.6rbtop.com/

http://:@www.google.com/

http://:@www.6rb.com/

http://:@www.yahoo.com/

http://:@www.mosul4all.com/

http://:@www.microsoft.com/

http://:@www.almawsil.com/

http://:@www.7shasha.com/

http://:@www.youtube.com/

http://:@www.aljazeera.net/

http://:@www.bokra.net/

http://:@www.6rbwow.com/

http://:@www.mosulalhob.net/

http://:@www.aljazeerasport.net/

http://:@www.graaam.com/

http://:@www.google-analytics.com/

http://:@www.hawar11.com/


14

TABLE (4): TOP 50 SITES WHICH REPRESENT 80% OF THE TCP CONNECTIONSIN A ONE DAY PERIOD(11/08/2007)

Accessed Site No. of TCP connections (Thousands)us.i1.yimg.com 173.5au.download.windowsupdate.com 164mosul4all.com 133.6us.bc.yahoo.com 94.7www.mosul4all.com 78.6msgr.dlservice.microsoft.com 77.8tbn0.google.com 66pagead2.googlesyndication.com 60ads.yimg.com 50www.google-analytics.com 42.16www.almawsil.com 42.14l.yimg.com 4114.mihd.net 40www.google.com 39www.7shasha.com 37insider.msg.yahoo.com 36us.js2.yimg.com 31messenger.yahoo.com 30kh.google.com 28ad.yieldmanager.com 2772.52.165.228 25content.sweetim.com 24www.mosulalhob.net 23.3mail.opi.yahoo.com 23www.6rb.com 22.7www.yahoo.com 21.42ad.z5x.net 21.39chat.alfnaan.com 21.16www.youtube.com 19.2ad.doubleclick.net 18img.youtube.com 17.3shttp.msg.yahoo.com 17www.ratteb.com 16.9stc.msn.com 15.8www.6rbtop.com 15.5m1.2mdn.net 15.4content.yieldmanager.edgesuite.net 15.3mail.google.com 14chat1.alfnaan.com 13.89www.4arab.com 13.53address.yahoo.com 13.53www.aljazeera.net 13.34qlbe.doook.com 13.1www.clocklink.com 13host1.digichet.com 12.8m1.webstats.motigo.com 12.5674.52.117.218 12.55host7.digichet.com 12.51us.music1.yimg.com 12.38us.i1.yimg.com 173.5

The most visited web site is (au.download.windowsupdate.com). This web site accessedautomatically (and occasionally) by the WINDOWS operating system for the software updateprocess, other web sites are accessed according to users’ demands.

http://:@www.mosul4all.com/

http://:@www.google-analytics.com/

http://:@www.almawsil.com/

http://:@www.google.com/

http://:@www.7shasha.com/

http://:@www.mosulalhob.net/

http://:@www.6rb.com/

http://:@www.yahoo.com/

http://:@www.youtube.com/

http://:@www.ratteb.com/

http://:@www.6rbtop.com/

http://:@www.4arab.com/

http://:@www.aljazeera.net/

http://:@www.clocklink.com/


15

From the above statistics, a better look to the users’ activities in a typical work day could becreated, as listed in table (5) below.

TABLE (5): TYPICAL DAILY USERS ACTIVITIESAverage number of live users 300 out of 2104Average downloaded data/day 69.8 GByteAverage downloaded data/user/day 232.6 MByte ~ (7 Gbyte/Month)Average number of TCP Connections/day 4.77 MAverage number of TCP connections/user/day 15.9 KAverage log time/user 10 hours

6. CONCLUSIONS & SUGGESTIONS FOR FUTURE WORK

This paper abstracts one year work on the test and measurements of a popular WISP system. Thefollowing notes could be extracted from the above statistics:1. The accurate plan prior to build the infrastructure network enhances system performance andallows a smooth future expansion.2. The main contributor in the WLAN latency budget is the delay between the clients and theirassociated access points. In order to get an acceptable performance, the number of clients shouldnot exceed a certain limit (60 node).3. Web browsing was the dominant internet application and occupies most of the downloadedbandwidth ( exactly 85%) .4. The performance of the system could be greatly enhanced using a proper cache server. Thistechnique has influence on both optimizing the used bandwidth and minimizing page responsetime( reduction by 50%). Also, Filtering server frees more bandwidth (exactly 40%)to the profitof the most important applications. 5. Daily http traffic has a periodic nature, i.e., it varies between low (16% of the full bandwidth)and high load (78% of the full bandwidth) values according to the users demands. This behavioris a direct result to the number of active users in these times and it is reflected on the networklatency variation.In spite of the fact that the current WISP system responds successfully to its Service LevelAgreement (SLA) criteria, it is still far away from the level expected as compared to the Internetservices provided to users worldwide. It was found that Internet users in Iraq still suffer fromperformance degradation due to multiple factors such as, contention nature of the WLANprotocols governs Internet sharing task (CSMA/CA protocol working under DCF mode) andlimited bandwidth-high propagation delay features of the satellite communication systems. Thisstudy suggests the following enhancement to the Internet system in Iraq:1. Installing a high speed network (such as SONET or 10Gigabit Ethernet) covers all major Iraqi

cities.2. Connecting the suggested Iraqi WAN to the Internet Backbone through high speed optical

fibers.3. Internet sharing & distribution inside each city could be accomplished using modern high

speed wireless networking techniques such as WLAN(IEEE802.11g) with QoS & PCF mode,4’th Generation mobile system or WiMAX (IEEE802.16).


16

ACKNOWLEDGMENT

We thank AL MEHRAB for Internet and communications for the facilities provided during thepreparation of this study.

REFRENCES[1] S. Dixit and R. Prasad,” Wireless IP and Building the Mobile Internet”, Artech House, 2002.[2] R. Kikta, A.Michael and P. Courtney, “Wireless Internet Crash Course”, McGraw-Hill

Inc., 2002.[3] J. Gutérrez, “Business Data Communications and Networking: A Research Perspective”,

IDeA Group Publishing, 2007.[4] G. Camponovo, and D. Cerutti, “WLAN communities and Internet access sharing: a

regulatory overview”, International Conference on Mobile Business, 11-13 July 2005Page(s):281 – 287.

[5] D. Tang and M. Baker, “Analysis of a Metropolitan Area Wireless Network”, Proceedings ofACM MOBICOM, Seattle, WA, pp. 13-23, August 1999.

[6] D. Tang and M. Baker, “Analysis of a Local-Area Wireless Network”, Proceedings of ACMMOBICOM, Boston, MA, pp. 1-10, August 2000.

[7] A. Balachandran, G. Voelker, P. Bahl, and P. Rangan, “Characterizing User Behavior andNetwork Performance in a Public Wireless LAN”, Proceedings of ACM SIGMETRICS,Marina Del Rey, CA, pp. 195-205, June 2002.

[8] D. Kotz and K. Essein, “Analysis of a Campus-Wide Wireless Network”, Proceedings ofACM MOBICOM, Atlanta, GA, September 2002.

[9] G. Bai ,K. Oladosu, Performance and Robustness Testing of Wireless Web Servers, M.Sc.Thesis, University of Calgary, September 2003.

[10] M. Narbutt and M. Davis, “Experimental investigation on VoIP performance and theresource utilization in 802.11b WLANs”, Proceedings 2006 31st IEEE Conference on LocalComputer Networks,Nov. 2006 Page(s):397 – 403.

The work was carried out at the University of Mosul

Dawwd: Real Time Implementation Of Fir Filter Based On Time Delay Neural Network

17

REAL TIME IMPLEMENTATION OF FIR FILTER BASED ON TIME DELAY NEURAL

NETWORK

Dr. Shefa Abdulrahman DawwdComputer Engineering Department, College of Engineering, University of Mosul

Email: [email protected]

Abstract

An FPGA real time Implementation of Time Delay Neural Network (TDNN) is presented inthis paper. The design and all of the work are geared towards the implementation of the TDNNin a scalable fashion. The TDNN is an adaptive FIR filter with 18-bit input and 18-bit outputresolution. In this paper, the filter adapts its tap weights using the Least-Mean-Square (LMS)algorithm and then stores them in FPGA memory cell. The LMS algorithm that is used forweight adaptation is off chip implemented. The input is processed through a digital tappeddelay line. The FIR neural network is used for real time adaptive noise cancellation. When thefilter order is 10, the filter consumes 1168 Spartan 3E FPGA logic elements.

Keywords: TDNN, FIR neural network, FPGA neural implementation.

. --

.

.

1818 .

.

. .

.

101168 .


mailto:shefadawwd:@yahoo.com


18

1- INTRODUCTION

Many modern-day electronic systems must deal with unknown or changing environmentalvariables such as noise levels, interference, and varying input statistics. These systems frequentlyuse adaptive signal-processing techniques to optimize their performance[1].

The problem of the lack of a priori knowledge of the noise is being treated by suchadaptive filtering. The idea is based on the application of time delay neural networks as anadaptive TDNNs inherently posses adaptability properties as well as they are of nonlinearstructure, hence they represent a natural choice for the construction of adaptive filters. A TDNNis trained by an identity observer with assumptions that the model of the system is precise and themeasurement and the process noise are unknown, regardless whether Gaussian or colored. It isshown by simulations that this concept gives better results when we miss information about thenoise statistics, or when the noise is not Gaussian[2].

However, in application domains such as mobile communications or ubiquitouscomputing, these systems also face severe constraints in power dissipation and circuit die area. Insuch cases, using programmable digital signal-processing (DSP) chips becomes infeasible.Although analog circuits can implement moderate resolution arithmetic at low power and area,these circuits are limited by other problems such as charge leakage, signal offsets, circuitmismatch, error accumulation, and noise sensitivity[3]. Therefore, using Application SpecificIntegrated Circuit (ASIC) seems as a compromised solution. Both approaches (analog anddigital-ASIC) have fixed topology, resulting in a design suited only for one type of targetapplication. Digital implementation, using FPGAs, allows the redefinition of the topology usingthe same hardware. FPGAs have traditionally been configured by hardware engineers using aHardware Description Languages (HDL). Very High Speed Description Language (VHDL) isone of the most efficient description languages used. The disadvantage of an implementationusing FPGA over ASIC is the performance. FPGA normally runs slower than ASICs[4].

In this paper an FPGA implementation of TDNN that is used as an adaptive filter isproposed. The rest of paper is organized as follows: in section 2, the FIR filter is introduced.Section 3 presents the structure of TDNN while the algorithm used to train the network ispresented in section 4. In section 5, the hardware implementation of the TDNN and its proposedhardware architecture is presented. Finally, in section 6, the experimental results and acomparison with the system presented in [5] where an FPGA chip is used to emulate an adaptiveFIR filter of various taps length is performed. Also some conclusions are discussed in thissection.

2- FINITE IMPULSE RESPONSE FILTER

Digital filters represent an essential part of digital signal processing. Filters are employedin preprocessing for reducing noise and enhancing characteristic qualities. An example of this isedge enhancement in pictures. Filters are categorized as linear, non-linear or adaptive[6].Whenever possible, linear filters are preferred since their treatment can be completely put intotheory. According to their impulse response, filters are divided into those with limited impulse


19

response (FIR=finite impulse response)) and those with an impulse response of infinite length(IIR=Infinite impulse response).

A FIR filter computes a convolution between an input data stream and a stored weightvector. Fig. 1 shows the architecture of a FIR filter. It comprises a digital delay line, weight cells,digital multipliers and adders[6].

Real world signal processing problems often require adaptive solutions. Simulations ofthese solutions, sometimes via trial-and-error, will hopefully identify an algorithm to correctlysolve the given problem. The problem such as the lack of a priori knowledge of the noise is beingtreated by adaptive filtering, simultaneously investigating the statistics of the noise and updatingthe filter gain upon them [2].

3- TIME DELAY NEURAL NETWORK

The most common feedforward networks are static networks, which have no internal timedelays. They respond to a particular input by immediately generating a specific output. However,static networks can respond to temporal patterns if the network inputs are delayed samples of theinput signals, i.e., time is treated as another dimension in the problem [7]. Incorporating time asanother dimension in neural networks is often referred to as Time Delay Neural Network(TDNN). These networks, trained with the standard back propagation algorithm, have been usedas adaptive filters for noise reduction and echo canceling and for chaotic time series prediction[8]. Consider a tapped delay line(Fig. 2), that is, a shift register. Consider also a multilayerperceptron where the tapped outputs of the delay are applied at its input, the output has a finitetemporal dependence on the input: y(k)=F[x(k),x(k-1),x(k-2),……..,x(k- )] , where F is a typicalnonlinearity function. When this function is a weighted sum, then the TDNN is equivalent to afinite impulse response (FIR) filter.

For the filter, the output y(k) corresponds to a weighted sum of past delayed values of theinput:

The Tapped Delay Line (or TDL), presented at Fig. 2, is a bank of unit delays used at theinput, sometimes also at the output, of an ANN. This is a standard element used in adaptivefiltering applications of ANNs. In this way the network structure is capable of processing number of time samples at the same time instance.

+ + + +

)0(h )1(h )2(Nh )1(Nh

)(kx

)(ky 0

1z 1z 1z

Fig. 1: The architecture of a FIR filter

(1))()()(0n

nkxnhky


20

The FIR network is a feedforward neural network architecture with internal time delaylines and can be learned by the temporal back-propagation algorithm. It is a modification of thebasic multi-layer network in which each weight is replaced by an FIR linear filter. The FIR filteris modeled with a tapped delay line is shown in Fig. 3.

If the TDNN is formed with two layers(input and output layer, there is no hidden layer),then a 1D FIR network like that shown in Fig. 1 is formulated. The layer weights are adapted fora given problem by using a supervised learning method.

Fig.2: The Tapped Delay Line (or TDL)

ANN

y(k)signaloutput

x(k)signalinput

1z

1z

1z

1z

1z

1z

1z

1z

+

+

+

1

2

3

input layer hidden layer output layer

)( kx

)1( kx

)2( kx

)( kx

)( ky

1ˆW

2ˆW

1z

1z

1z

1

Fig.3: TDNN structure


21

4- TDNN LEARNING ALGORITHM

The network shown in Fig.2 is classified as a network of multilayered feedforward neuralnetwork. The back-propagation algorithm is used to train this type of neural network. As shownin Fig. 2, a multilayer feedforward network has an input layer of source nodes and an output layerof neurons (i.e., computation nodes); these two layers connect the network to the outside world.In addition to these two layers, the multilayer perceptron usually has one or more layers of hiddenneurons, which are so called because these neurons are not directly accessible. The training isusually accomplished by using a backpropagation (BP) [2] algorithm that involves two phases:1) Forward Phase. During this phase the free parameters of the network are fixed, and the inputsignal is propagated through the network of Fig. 2 layer by layer. The forward phase finishes withthe computation of an error signal:

e[m]=d[m]–y[m] (2)

where d[m] is the desired response and y[m] is the actual output produced by the network inresponse to the input x[m].2) Backward Phase. During this second phase, the error signal e[m] is propagated through thenetwork of Fig. 2 in the backward direction, hence the name of the algorithm. During this phasethe adjustments are applied to the free parameters of the network so as to minimize the error e[m]in a statistical sense. To update weights the least mean square (LMS) learning rule is used:

w[m+1]=w[m]+ *x[m]*e[m] (3)

In TDNN, each filter coefficient adaptation uses its present coefficient value, w[m], to add tothe product of the step-size parameter, , tap input x[m] and error output e[m], to obtain anupdated version of the filter coefficient w[m+1]. Note that the only information needed to updatefilter coefficients are the tap input and the error term. The step-size parameter (learning rate) ispre-determined by the system engineer and is a decimal number between 0 and 1[9]. The back-propagation learning algorithm is simple to implement and computationally efficientin that its complexity is linear in the synaptic weights of the network. However, a majorlimitation of the algorithm is that it does not always converge and can be excruciatingly slow,particularly when we have to deal with a difficult learning task that requires the use of a largenetwork.

5- HARDWARE IMPLEMENTATION OF TDNN

It is the goal of neural network engineers to transfer the progress made into new hardwaresystems. They are intended to accelerate future developments of algorithms and architectures,and to make possible the use of dedicated neural networks in industrial applications.

Learning algorithm implementation` The learning algorithms used for modifying weights values using inputs and training data are

as an important part of the network system as the architecture itself. Implementation of learningin VLSI systems takes three forms; off-chip, ’chip-in-the-loop’ and on-chip learning. In off-chiplearning, weights values are calculated externally by software and are then downloaded to the


22

neural network which is then used only for recall. This is the easiest but least favoured method,since training times can be long. Off-chip learning does have the advantage in that it is easy tochange the learning algorithm simply by modification of software. It also allows the use offloating point arithmetic for the algorithms which may not be feasible on a neural network chip.’Chip-in-the-loop’ training chip may also be considered as an off-chip method since the trainingalgorithm is still run in software. However, in this case the neural network is used in the trainingloop which removes the need for a software model of the network itself, and compensates fordevice variability. The main drawback of this method is the communications overhead incontinually reading and writing data across the network/host interface. On-chip learning must beseen as the most desirable method, since it may open the way to stand-alone neural networkchips. The main advantage of running the learning algorithm in hardware is the gain in speed.The on-chip learning consumes large chip area for floating point calculations [10] that are usuallyrequired to update weights in backward phase. Thus, in this paper off-chip learning is used, andthe on-chip learning can be left to be a future work.

Numerical RepresentationIn general, neural networks have low-precision requirements, even though the exact

specification is algorithm and application dependent. Digital neurohardware can profit from thisproperty by using fixed-point arithmetic with reduced precision. Fixed-point implementations areless complex and less area consuming than floating-point arithmetic and therefore their use helpsto reduce system cost [11].

Implementation of TDNNs Activation Function`Direct implementation for non-linear activation functions is very expensive. There are

two practical approaches to approximate non-linear functions with simple FPGA designs. Piece-wise linear approximation describes a combination of lines in the form of y=ax + b which isused to approximate the non-linear function. Note that if the coefficients for the lines are chosento be powers of two, the non-linear functions can be realized by a series of shift and addoperations. Many implementations of neuron activation functions use such piece-wise linearapproximations. The second method is lookup tables, in which uniform samples taken from thecenter of non-linear function can be stored in a table for look up. The regions outside the centerof the non-linear function are still approximated in a piece-wise linear fashion.

`The TDNN consists of two types of activation functions classified as linear and non-linear. Hidden layer neurons have non-linear activation functions, while the output layeractivation function is linear (pureline function). Implementation of non-linear activation functionwas discussed in previous works [12]. The pureline linear activation function is y=ax anda=1(see Fig.4), therefore the function weighted sum input will be the activation function output.

)( xf

x

Fig.4: pureline linear activation function


23

The architecture of FIR networkThe FIR network implemented in this work is a TDNN of two layers (input and output

layers). The type of the output neuron activation function is pureline linear function. The designand all of the work are geared towards the implementation of the TDNN in a scalable fashion.The scalable design means that the network can become as large as practically possible thusproviding a structure for more complex application. In this implementation, the TDNNs aretrained off-chip and the convergent parameters then fed to the hardware for test and synthesis.The processing elements and the complete architecture of the proposed neural network is shownin Fig.5(a,b).The processing element does the algebraic equations of the electric model of theneuron, that is, the multiplication and sum required in the neuron's internal processing(see Fig. 5).

All PEs are identical. Each PE has one delay register, one weight storage element and onemultiplier. The number of multiplier depends on the number of tapes. Therefore, when thenumber of multipliers exceeds the number of FPGA embedded multipliers, the redundantmultipliers are built by exploiting the FPGA chip gates.

The multiplications are the most power dissipating arithmetic operations. Today's FPFAscontain a number of speed optimized signal processing building blocks, such as multipliers,RAM blocks or I/O structures with propagation delays in the range of a few nanoseconds.However, in this system we intended to design the TDNN to be as a part of larger system.Therefore any of embedded hardware multipliers can be left to be used by other parts of thesystem and a custom hardware multiplier can be designed. Based on that, and in order tominimize the area cost and the total power consumption and also to simplify theimplementations, the parallel Booth multiplier technique can be chosen. The simple serial byparallel booth multiplier is particularly well suited for FPGA implementation without carrychains because all of its routing is to nearest neighbors with the exception of the input. The totalarea of the implemented multiplier highlights the advantages over traditionally implementation ofthe multipliers specially with the increasing of multiplier operand sizes.

weightROMw(k)

delayregister

)( kx )1( kx

control

kPE

0PE 1PE PE

Pipeline Register

SIMD Adder Tree

Output Register

SIMD Saturation Module

)( kx

)( ky

Fig. 5: (a)` A block diagram of the processing element (PE) (b) Thearchitecture of FIR neural network

(a)

(b)


24

To minimize the time required to add all the PEs results, and to reduce the number of adders,the single add instruction, multiple data coming from each PE (SIMD) adder tree [4] is suggestedto be used. The SIMD saturation module is used to fix the bit resolution of the adder results.

The minimum word length for each weight and input elements is the one that shouldpreserve the same generalization ability that achieved from software. Since the input parametersword length for FPGA embedded multipliers are (18x18), and to exploit the full precisiongiven by this constrain, the word length for input and tap's weight are selected to be 18-bits.Therefore, each PE has a 18x18 bit multiplier.

6- RESULTS AND CONCLUSIONSThe FIR neural network mentioned in previous sections has been applied as a 1_D

adaptive FIR filter which is used to filter a 1_D signal from noise in real time. In this paper anFPGA VLSI architecture that implements the network is proposed. Xilinx Spartan-3E FPGA of500,000 gates is used for implementation. The FPGA digital hardware model has been designedusing Xilinx Foundation environment.

The choice of the order of the FIR filter is experimentally determined. The order thatmakes the FIR filter gives a reasonable signal to noise ratio (SNR) is adapted. There is no clearmathematical analysis to derive the quantities. Only through experiments may we obtain afeasible solution.

The software part of this paper is implemented on a personal computer works underWindowsXP. The algorithm emulation is carried by using MATLAB. The noise that added to aselected signal, has a mean of zero, a variance of one and standard deviation of one.

In Fig.6a, the original signal is used as a target to train the network, while the trainingsignal (Fig.6b) is the original signal added noise with a SNR equals to 10 dB. In Fig.6c andFig.6d, the filter output signals are shown when the filter order (no. of taps) equals to 5 and 10,respectively after 200 learning steps. In Fig.6e, the filter used to generate the signal in Fig.6d isconsidered but after 800 learning steps. In Fig.6f and Fig.6g, the filter output signals are shownwhen the filter order equals to 10 and the SNR of the input signals are 15 and 20 dB respectively.

In Fig.6h, a test signal of 10 dB SNR that is not used in training is considered as an input to theFIR network of 10 taps trained in 800 epochs. The output test signal is shown in Fig.6i. In Fig.6jand Fig.6k, the output test signals are shown when the input signals are of 15 dB and 20 dBSNR respectively.

The FIR filter has still worked as a low pass filter (LPF) if the noise added to the originalsignal is reasonable. For example, the original signal properties that shown in Fig. 6a areobviously presented in Fig. 6g which represents the restored signal after adding noise. While themain properties of the original signal are disappeared in Fig. 6c, Fig. 6d and Fig. 6e when theSNR=10dB (the signal shown in Fig. 6b). One also can see that the filter of higher order (no. oftap =10 in Fig. 6d) is better than that of lower order (no. of tap=5 in Fig. 6c). When the propertiesof the original signal is preserved (Fig. 6h) after adding the noise, the filter can work as a LPFeven if the SNR is low (SNR=10dB in Fig. 6i). Briefly, the filter performance is degraded whenthe SNR decreases, while the filter still works as a LPF even if the input signal is highly


25

corrupted with a noise. Also, it is concluded that the filter is adapted as a low pass filter for anytype of input signals even if the type of the input signal is different from the training signal.

(a) (b) (c)

(d) (e) (f)

(g) (h) (i)

(j) (k)

Fig.6: Training and testing signal with the noise influences (a) original target signal (b) theadded noise input signal, SNR=10dB (c) The output signal with canceling noise ofSNR=10dB, Tap=5,epochs=200 (d) The output signal with canceling noise ofSNR=10dB,Tap=10, epochs=200 (e) The output signal with canceling noise of SNR=10dB,,Tap=10, epochs=800 (f) The output signal with canceling noise of SNR=15dB, ,Tap=10,epochs=800 (g) The output signal with canceling noise of SNR=20dB, ,Tap=10,epochs=800 (h) The test input signal , SNR=10dB, ,Tap=10 (i) The test output signal withcanceling noise of SNR=10dB, ,Tap=10, epochs=800 (j) The test output signal withcanceling noise of SNR=15dB, ,Tap=10, epochs=800 (k) The test output signal withcanceling noise of SNR=20dB,Tap=10, epochs=800.

% %

%

%

amplitude (volt)

time (ms)


26

As mentioned earlier, the device chosen is the Spartan 3E, which contains 4656 logicelements, 20 embedded multipliers, and has maximum clock rate of 50MHz. Area and speed arethe two main measurements in evaluating the hardware performance of this filter. Fig. 7 showsthe varying filter orders vs. speed and area, respectively. Area is measured by number of slicesoccupied. Speed is measured by the maximum allowable clock frequency. These two factors arecompared between the system proposed in this paper and the system of Stratix FPGA devicepresented in [5]. Each Logic Element (LE) in both Spartan 3E and Stratix families consists of a 4-input lookup table and a flip-flop.

One can see from Fig.7 that the system presented of our paper consumes lower areaand can be operated in higher clock frequencies than the system presented in [5]. That is dueto the techniques used in calculation components. Also it is due the optimized VHDL codesthat used for modeling.

From Fig.7, one can see that the system can be operated in a maximum frequency butof course if the FPGA chip is able to work on this frequency.

Fig .8 shows the timing diagram for the calculations that performed inside the networkPEs. One can see that the PEs inputs (x_0,x_1,……………) that represent the input signalsamples are fully pipelined. The PEs outputs (y_0,y_1,…………….) are calculated in parallel.Also every clock cycle the network final output signal (z) is generated.

It is believed that a better result can be achieved if the numbers of tapes are increased.Also, more than one layer can be used in the TDNN to improve the filter performance. The LMSalgorithm that is used for weight adaptation can also be on-chip implemented, then, the speed ofthe learning process is greatly increased. These works can be left to be investigated as a future works.

Fig 7: (a) Speed vs. FIR filter order (b) Area vs. FIR filter order.

Fig.8: The timing diagram of the FIR network

0

5 0 0

1 0 0 0

1 5 0 0

2 0 0 0

2 5 0 0

S p a r t a n 3 E

S t r a t ix

Filter order

Logi

c

elem

ents

105(b)Filter order

Logi

c ele

men

ts

Spartan 3EStratix

0

2 0

4 0

6 0

8 0

1 0 0

S p a r t a n 3 E

S t r a t ixMax

imum

frequ

ency

(MHz)

Filter order

5 10(a)Filter order

Max

imum

frequ

ency

Spartan 3EStratix


27

7- REFERENCES:

[1] Figueroa M., Hsu D., and Diorio C. "A Mixed-Signal Approachto High Performance, Low-Power Linear Filters," IEEE Journal of Solid-State Circuits,vol. 36, pp. 816- 822, 2001.[2] Kihas D., Zeljko M., and Branko D., " Adaptive Filtering based on Recurrent NeuralNetworks", Journal of Automatic Control, University of Belgrade, vol.13, no. 1, pp.13-24, 2003.[3] Miguel Figueroa, Seth Bridges, David Hsu and Chris Diorio,"A 19.2GOPS, 20mW AdaptiveFIR Filter",Solid-State Circuits Conference ESSCIRC '03. Proceedings of the 29th European,pp.509-512,2003.[4] Meyer U., "Digital Signal Processing with FieldProgrammable Gate Array", Springer-VerlagBerlin Heidelberg New York, 2001.[5] Lin A. Y. and Gugel K. S. , " Feasibility of Fixed-Point Transversal Adaptive Filters inFPGA Devices with Embedded DSP Blocks" Proceedings. The 3rd IEEE International Workshopon Volume, Issue, 30 June-2 July 2003 pp. 157 –160, 2003.[6] Pirsch P., "Architectures for Digital Signal Processing", John Wiley & Sons, 1998.[7] Boozari M., " State Estimation and Transient Stability Analysis in Power Systems usingArtificial Neural Networks", M. Sc. Thesis, School of Engineering Science,Simon FraserUniversity, 2004.[8] Hush D.R. and Horne G., "Progress in supervised neural networks", IEEE SignalProcessingMagazine, vol. 10, no. 1, pp. 8-37. 1993.[9] Douglas S. C. and Meng T. H., "Linearized least- squares training of multilayer feedforwardneural networks", In Int. Joint Conf. of Neural Networks, pp. 307–312, 1991.[10] Jocelyn C., Eric C., and Steven P.,“VIP: An FPGA-based Processor for Image Processingand Neural Networks”, Fifth International Conference on Microelectronics for Neural Networksand Fuzzy Systems (MicroNeuro'96), IEEE, Lausanne, Switzerland, pp. 330-336.1996.[11] Li X., Moussa M., and Areibi Sh., “Arithmetic formats for implementing Artificial NeuralNetworks on FPGAs” Canadian Journal on Electrical and Computer Engineering, vol. 31, issue1 ,pp. 31-40,2006.[12] Tommiska M.T., “Efficient digital implementation of the sigmoid function forreprogrammable logic”, IEE Proceedings – Computers and Digital Techniques 150, no. 6, pp.403-411, 2003.[13]Morris Mano M., "Computer System Architecture", Prentice Hall, Inc., 1993.


Abdul-Jabbar: Design And Realization Of Circular Contourlet Transform

28

DESIGN AND REALIZATION OF CIRCULAR CONTOURLETTRANSFORM

Dr. Jassim M. Abdul-Jabbar* and Hala N. Fathee**

* Ph. D in Elect. Eng., Dept. of Computer Eng., College of Eng., University of Mosul, Iraq** M. Sc. in Computer Science, Dept. of Sport Science, College of Sport Educ., University of Mosul, Iraq.

Abstract In this paper, circular contourlet transform (CCT) is proposed, designed and

realized. As in the classical contourlet transform (CT), a double filter bank structure isalso considered in this work but in different manners. A circularly-decomposed filterbank is first used to capture the points of discontinuities in the image edges, and thenfollowed by a directional filter bank to obtain smoothed contours. The resulting CCTcontains a critically sampled filter bank that decomposes images into any power of two'snumber of directional subbands at multiple scales. The designed CCT is realized by 2-D lattice allpass sections with separable and non-separable 2-D functions of z1 and z2.The resulting structure preserves both modularity and regularity properties which aresuitable for VLSI implementations. Objectively, the performances of the realized CCTare tested and proved to be better than the classical CT in detail image preservation.The resulting subband images also indicate the superiority of the proposed CCT.

Keywords: Circular contourlet transform, Contourlet transform, Laplacianpyramid, Directional filter bank, 2-D lattice allpass sections, Multiresolution(multiscale & multidirection) analysis.

. - -

––

.

.

. z1 z2 .

VLSI .

.CCT.



29

I. INTRODUCTION

Although the wavelet transform (WT) is known to be a powerful tool in many signal andimage processing applications such as compression, noise removal, image edge enhancement,and extraction; wavelets are not optimal in capturing the two-dimensional singularities foundin images and often required in many segmentation and compression applications [1]-[3]. Inparticular, natural images consist of edges that are smooth curves which cannot be capturedefficiently by the wavelet transform. Therefore, several new transforms are required for imagesignals.

The contourlet transform (CT) is one of the new geometrical image transforms, whichcan efficiently represent images containing contours and textures[4]-[6]. This transform uses astructure similar to that of curvelets [7]-[10], that is, a stage of subband decompositionfollowed by a directional transform. In the contourlet transform, a Laplacian pyramid (LP) isemployed for the first stage, while directional filter banks (DFBs) are used in the angulardecomposition stage. A comparison between the wavelet scheme and the contourlet shows theimproved edge contours of the later [6],[11]. This is attributed to the grouping of nearbywavelet coefficients, since they are locally correlated due to the smoothness of the contours asshown in Fig. 1. Therefore, a sparse expansion for natural images can be obtained by firstapplying a multiscale transform, followed by a local directional transform to gather the nearbybasis functions at the same scale into linear structures. In essence, a wavelet-like transform foredge detection, and then a local directional transform for contour segment detection areapplied. The overall result is an image expansion using basic elements like contour segments,and thus are named contourlets and the process is called the contourlet transform (CT).

Circular split 2-D spectral schemes (circularly-decomposed frequency subspaces) areknown to give better performance than rectangular-support or diamond-support schemes whenit is desired to extract as low frequency information as possible in a 2-D low-pass filteringprocess. This also means that circular split schemes are preferred to extract as high frequencyinformation as possible in a 2-D high-pass filtering process [12].

Wavelet ContourletFig. 1 The successive refinement by the two systems (wavelet and contourlet) near a smooth

contour, which is shown as a thick curve separating two smooth regions.

In this paper, A circular contourlet transform (CCT) is proposed, designed and thenefficiently realized. A double filter bank structure is also applied. Using circularly-decomposed filter bank, A circular split scheme (CSS) is first employed to capture thepoints of discontinuities, and then followed by a directional filter bank (DFB) to obtain


30

smoothed structures. The resulting CCT decomposes images into directional subbands atmultiple scales and contains a critically sampled filter bank that decomposes images into anypower of two`s number of directions. The rest of this paper is organized as follows: Thecontourlet transform is described in section II. Section III contains the multiscale andmultidirection analysis of such transform. In section IV, the proposed circular contourlettransform is formulated, designed and realized using 2-D lattice allpass sections withseparable and non-separable 2-D functions of z1 and z2 (z1 and z2 are the 2-D complex spatialfrequencies in the discrete spectral domain). The quantitive and qualitative performances ofthe realized CCT transform are examined and compared with those of the classical CTtransform in section V. Finally, section VI concludes this paper.

II. THE CONTOURLET TRANSFORMThe contourlet transform consists of two major stages: the subband decomposition and

the directional transform. At the first stage, LP is used to decompose the image into subbands ,and then the second one is a DFB which is used to analyze each detail image. A flow graph ofthe CT is shown in Fig. 2. Example of the spectral split scheme achieved by such LP is shownin Fig. 3. 2-D filters may be employed in the realization of this stage. Example of thedirectionally-decomposed frequency split scheme achieved by such DFB is shown in Fig. 4 .2-D fan filters which serve as the building blocks of this DFB have wedge-shaped passbandspectral regions [10] as in Fig. 4.

Fig. 2 A flow graph of the contourlet transform. The image is first decomposed into subbandsby LP and then each detail image is analyzed by DFB.

Fig. 3 The spectral split scheme of LP filter . Fig. 4 An example of the directional bank frequency partitioning .


31

It is easily shown that these wedge-shaped regions correspond to directional components ofthe image [8], [9]. All 2-D filters employed are 2-D extensions of the 1-D filters previouslydesigned to satisfy perfect / aliasing-free reconstruction constraints [12]. The resulting 2-Dfilter functions are nonseparable in z1 & z2.

Since, the CT is formed precisely via a new multiresolution analysis framework that issimilar to the link between wavelets and filter banks [1], the new elements in this frameworkare multidirection and its combination with multiscale. With this insight, a double filter bank structure (see Fig. 5a) is used for obtaining sparseexpansions for typical images having smooth contours [5], [6]. In this double filter bank, theLP (rectangular-support scheme) is first used to capture the points of discontinuities, and thenfollowed by a DFB (directionally-decomposed split scheme) to link points of discontinuitiesinto smooth curves [4]. The overall result is an image expansion using basic elements likecontour segments, and thus are named contourlets. In particular, contourlets have elongatedsupports at various scales, directions, and aspect ratios. Thus allows contourlets to efficientlyapproximate a smooth contour at multiple resolutions just like the scheme shown in Fig. 5b.From frequency domain point of view, CT provides both multiscale and multidirectionaldecompositions.

(a) (b)

Fig. 5 (a) The contourlet filter bank: first, a multiscale decomposition into octave bands by LPis computed, and then a DFB is applied to each bandpass channel, (b) A typical contourletfrequency partition scheme.

III. MULTIRESOLUTION ANALYSIS

In the followings and for simplicity, the case with orthogonal filters, which lead to tightframes will be considered only. The more general case with biorthogonal filters can be treatedsimilarly. Multiresolution analysis is divided into the following two analysis:A) Multiscale analysis

This is the multiresolution analysis for the LP, which is similar to the one for wavelets.Suppose that the LP in the contourlet filter bank uses orthogonal filters and down sampling by2 in each dimension as shown in Fig. 6 (that means M = diag(2,2)= ). Under certain


32

regularity conditions, the lowpass synthesis filter G in the iterated LP uniquely defines aunique scaling function )()( 2

2 RLt that satisfies the following two-scale equation [1], [2].

.)2(][2)(2Zn

ntngt (1)

where g[n] is the impulse response of the lowpass synthesis filter G. Let

.,,222 2

, ZnZjntj

jj

nj (2)

Then the family 2, Znnj is an orthonormal basis for an approximation subspace jV at the

scale j2 . Furthermore, ZjjV provides a sequence of multiresolution nested subspaces

...,... 21012 VVVVV where jV is associated with a uniform grid of intervalsjj 22 that characterizes image approximation at scale j2 . The difference images in the LP

contain the details necessary to increase the resolution between two consecutiveapproximation subspaces. Therefore, the difference images live in a subspace jW that is the

orthogonal complement of jV in 1jV (see Fig. 7a) , or.1 jjj WVV (3)

(a) (b)

Fig. 6 LP scheme; (a) Analysis: Outputs are a coarse approximation c(n) and a difference d(n)between the original signal and the prediction P. The process can be iterated bydecomposing the coarse version repeatedly, (b) Usual synthesis.

(a) (b)Fig. 7 (a) Multiresolution Analysis: scale, (b) Multiresolution Analysis: direction

It is believed that the LP can be considered as an oversampled filter bank where eachpolyphase component of the difference image ][nd in Fig. 6, together with the coarse image


33

][nc ,comes from a separate filter bank channel with the same sampling matrix M = diag (2,2).Let 30,)( izFi be the synthesis filters for these polyphase components. These are

highpass filters. As for wavelets, a continuous function )()( ti , can be associate with each ofthese filters, where

.)2(][2)(2

)(

Zni

i ntnft (4)

where fi[n] is the impulse response of the highpass synthesis filter Fi(z).So, letting )()( ti in (4), be in the following form

2)()(, ,,

222)( ZnZjntt j

jiji

nj (5)

Then, for scale j2 , 2,30)(

, Znii

nj is tight frame for jW . For all scales,

2,30,)(

, ZniZjinj is a tight frame for )( 2

2 RL . In both cases, the frame bounds are equal to

1. Since jW is generated by four kernel functions (similar to multi-wavelets), in general it isnot a shift-invariant subspace. Nevertheless, a shift-invariant subspace can be simulated bydenoting

,30,)()( )(,2, itt injknj i

(6)where ik are the coset representatives for down sampling by 2 in each dimension, i.e.,

.)1,1(,)1,0(,)0,1(,)0,0( 3210TTTT kkkk (7)

With this notation, the family 2, Znnj associated to a uniform grid of intervals 11 22 jj on2R provides a tight frame for jW .

B) Multidirection Analysis

Using multirate identities [2], it is instructive to view an l-level tree-structured DFBequivalently as a l2 parallel channel filter bank (as in Fig. 8) with equivalent analysis filters,synthesis filters and overall sampling matrices. In Fig. 8, the equivalent directional analysisfilters are denoted as , and the directional synthesis filters as

llk kD 20,)( , which correspond to the subbands indexed as in Fig. 7. The corresponding

overall sampling matrices are proved to have the following diagonal forms [1]

)"("2220

02

)"(",202002

11

11

)(

directionverticalneark

directionhorizontalneark

Sll

l

ll

lk (8)

which means sampling is separable. The two sets correspond to the mostly horizontal andmostly vertical set of directions, respectively. From the equivalent parallel view of the DFB, itcan be seen that the family


34

,][ 2,20)()(

Zmkl

kl

k lmSnd (9)

obtained by translating the impulse responses of the equivalent synthesis filters )( lkD over the

sampling lattices by )(lkS , provides a basis for discrete signals in L2(z2). This basis exhibits

both directional and localization properties. In the iterated contourlet filter bank, the discretebasis (9) of the DFB can be regarded as a change of basis for the continuous-domainsubspaces from the multiscale analysis of the previous LP stage. Suppose that the DFB in thecontourlet filter bank utilizes orthogonal filters and when such DFB is applied to thedifference image (detail) subspaces, then the resulting detail directional subspaces )(

,lkjW

in the frequency domain will

Fig. 8 The multichannel view of an l-level tree-structured DFB,

be as illustrated in Fig. 9. The indexes j , k, and n specify the scale, direction, and location,respectively. Note that the number of DFB decomposition levels l is different at differentscales j, and is denoted by jl . Recall that jW is not a shift-invariant subspace. However, its

subspaces )(,lkjW are regenerative, since they are generated by a single function and its

translations.

Fig. 9 Multiscale and multidirection subspaces generated by the transform which isillustrated on a 2-D spectrum decomposition .


35

For a contourlet filter bank, the following properties hold [11], [13]:(1) If both the LP and the DFB use perfect-reconstruction filters, then the discrete contourlettransform achieves perfect reconstruction, which means it provides a frame operator.(2) If both LP and the DFB use orthogonal filters, then the CT provides a tight frame withframe bounds equal to 1.(3) The discrete contourlet transform has a redundancy ratio that is less than 4/3.(4) Suppose an levell j DFB is applied at the pyramidal level j of the LP, then the basisimages of the discrete contourlet transform(i.e. the equivalent filters of the contourlet filter

bank) will have an essential support size of jCwidth 2 and22 jljClength , where C

is a constant.(5) Using FIR filter realizations, the computational complexity of the CT is )(NO for N-pixelimages.

IV. THE PROPOSED CIRCULAR CONTOURLET TRANSFORM (CCT)

A) Formulation and properties

The proposed circular contourlet is a cascade of a circular split scheme (CSS) and a DFBas shown in Fig. 10a . Such structure decomposes images into directional subbands at circularmultiple scales. The DFB is a critically sampled filter bank that decomposes images into anypower of two`s number of directions. Due to this cascaded structure, the circular multiscaleand directional decompositions are independent of each other. One can decompose each scaleinto any arbitrary power of two's number of orientations and different scales can be dividedinto different numbers of orientations. Fig. 10b also shows the proposed frequency divisionof the circular contourlet transform where three of the four scales are divided into two, four,and eight directional subbands from coarse to fine scales, respectively. Figure 11 illustrates the subspace generation by CCT. In Fig. 11a , jV is a subspace,defined on a uniform circular grid. The difference image ( Wj subspace in the CSS) carries thedetails necessary to increase the resolution from jV to 1jV on an image approximation; index kruns to all l2 directions. Figure 11b illustrates the subspace generation by the directionaldecomposition and the increase in the resolution from to and . and

, in Fig. 11c, are the resulting subspaces from the applications of the directionaldecomposition on the details subspace Wj in the CSS.

(a) (b)Fig. 10 (a) The proposed the discrete circular contourlet transform implementation. (b) Its frequency partition scheme.


36

(a) (b) (c)

Fig. 11 Generation of subspaces by CCT; j subspace index; k direction; l all commands(a) CSS decomposition. (b) directional decomposition. (c) CCT subspaces.

B) Design and realization

The design of the circular split scheme (CSS) 2-D filters starts from the 1-D splittingorthogonal filers in Fig. 12a. In such figure and for perfect-reconstruction , it can be written as

1)(~)()(~)( 11 zgzgzhzh (10)

where

)()(~

)()()(~)()(

zhzgand

zhzgzh

zhzg

(11)

(10) can be evaluated on the unit circle )( iez as

1)()(22 ii eheh (12)

It is clear, that condition (12) is met for all , even at 2/ and leads to a LP-HPpower complementary filter pairs with perfect-reconstruction. It should be noted here thatconstraints (11) does not impose any phase conditions. Nevertheless, the approach will resultin zero-phase properties of the total system. Condition (12) can easily be written in thefollowing from:

1)()( 22 ii egeh ,for all (13)

Keeping the condition (13) in mind, h(z) and g(z) can be expressed as the sum anddifference of two all-pass sections, where

)()(21)( 10 zazazh (14a)

and

)()(21)( 10 zazazg (14b)


37

where )(0 za and )(1 za are stable all-pass functions.

Fig. 12 (a) 1-D analysis bank, (b) Lattice all-pass equivalent filter bank.

The structure in Fig. 12b represents the lattice all-pass equivalence of the analysis bankof Fig. 12a which satisfy (14a&b). It should be noted that that the lattice all-pass equivalentstructure of the synthesis bank is identical to the 2-band analysis bank and can also berepresented as the sum and difference of two all-pass sections. However, the design of thedesired circular split scheme (CSS) 2-D filter bank is an extension of the 1-D filter bank viaa suitable 1-D to 2-D transformation. It should be noted, here, that the circular decompositionscheme requires some nonlinear change of variables, such as [12]

)()( 22111 zfzfz (15a)and

)()( 122112 zfzfz (15b)

where

2,11

)( iforz

zzfii

iiii (16)

with 21 for circular shape decomposition. The choice of 1 and 2 values provides the appropriate bending for the supports to belook like circular shapes. On the other hand, the values of 1 and 2 are limited to

.2,11 ifori (17)to insure stability [12]. It should be noted that, (16) is nonlinear transformation called the digital spectraltransformation (DST) [12], [13]. Due to this non-linearity, these variations of variables cannotbe readily incorporated in the sampling rate alteration filters. Therefore, such changes can beincorporated in the analysis/synthesis filters themselves, resulting in 2-D all-pass sectionswhich are nonseparable, although however they can be efficiently realized as will be shown inthe following discussion: The circularly-decomposed filter banks can be realized, using the same 1-D structureshown in Fig. 12a, but with )(zh being transformed via the DST to ),( 21 zzh which can beformed as [12]

))(())(()()(),( 22112121 zfhzfhzhzhzzh (18)where 1z and 2z are as those given in (15a&b) and )( ii zf is as given in (16). It was proved previously in [14], that the application of this DST to linear-phase 1-Dfilter leads to 2-D filters which preserve the linear-phase characteristics approximately. Thus,it can be concluded that the application of this DST to the 1-D filter banks, in Fig. 12a,will lead to 2-D circular filter banks for perfect reconstruction of images. The equivalent filter


38

bank in Fig. 12b can be used. Since, it is realized in a lattice all-pass sections, then somesaving in computation can be gained , while lattice structure will provide the system with areduced sensitivity to finite word length of multiplier values [11]. The support of circularly-symmetric response of 2-D low- /high-pass splitting scheme isshown in Fig. 11a, the cutoff curve in this scheme is characterized by the circle c = ,where c is the 2-D cutoff frequency in rad. The pass-band is described by the

region c22

2121 0:, while the stop-band is described by the

region 22

2121 :, c .

The same 1-D structure shown in Fig. (3.3a) with h(z) and g(z) being 1-D filters chosen tobe realized as Haar orthogonal filters [ and is as given in (11)],while, their 2-D versions and , used for the realization of the CSS scheme,are identical to that of (18). After designing and realizing this 2-band circular-split scheme, asecond stage of DFB still have to be designed and realized. In such a stage, the design of the2-D DFB will also start from the 1-D splitting filters in Fig. 12b with the two 1-D all-passsections and being replaced by their 2-D counterparts. As in Fig. 13, a 3-level treedecomposition structure is used to serve for an 8-band directional decomposition. The Haarorthogonal filters are also utilized, here, and realized as a two 1-D all-pass sections in a latticestructure, with

(19a)and

(19b)

The corresponding 2-D lattice all-pass sections in the first level of the tree structure of Fig.13will yield a 2-band directional-decomposed split scheme. These sections can be derived as [ 11]

)()(),( 1210210210 zzazzazzA (20a)

and

)()(),( 1211211211 zzazzazzA (20b)

The transfer functions till the outputs of the first level of directional decomposition),( 210 zzDA and ),( 211 zzDA can be written in matrix form as

),(),(

1111

21

),(),(

211

210

211

210

zzAzzA

zzDAzzDA

(21)

The resulting 2-band directionally-decomposed filter bank can be realized as in Fig. 12b, butwith )(0 za and )(1 za being replaced by ),( 210 zzA and ),( 211 zzA given in (12a&b),respectively.

At the second level of the tree structure, a 4-band directional-decomposed split scheme isalso formed as in Fig. 13. It is realized just like the previously mentioned Haar lattice all-passstructure with )(0 za and )(1 za being replaced by


39

.1,0)()(),( 2121 iforzjazjazzB iii (22)

The transfer functions till the outputs of the second level of directional decomposition),( 210 zzDA and ),( 211 zzDA can be written as

.1,0,),(),(),(),(41),( 21121021121021 iforzzBzzBzzAzzAzzAD i (23a)

and

.3,2,),(),(),(),(41),( 21121021121021 iforzzBzzBzzAzzAzzAD i (23b)

The all-pass functions in third level of the tree structure of the directional decompositionstage (Fig.13) form the final requirement to accomplish an 8-band directional-decomposedsplit scheme. These functions are given by [11]

C j i (z1,z2) = Ai(z1,z2) , for i = 0, 1 (24)

with , for j = 1, , for j = 2,

, for j = 3, and , for j = 4.

Thus, the overall 8-band directionally-decomposed filter bank transfer functions are

DA0"(z1,z2) = [A0(z1,z2)+A1(z1,z2)][ B0(z1,z2)+B1(z1,z2)][C10(z1,z2)+ C11(z1,z2)],

DA1"(z1,z2) = [A0(z1,z2)+A1(z1,z2)][ B0(z1,z2)+B1(z1,z2)][C10(z1,z2) - C11(z1,z2)],

DA2"(z1,z2) = [A0(z1,z2)+A1(z1,z2)][ B0(z1,z2) - B1(z1,z2)][C20(z1,z2)+C21(z1,z2)],

DA3"(z1,z2) = [A0(z1,z2)+A1(z1,z2)][ B0(z1,z2) - B1(z1,z2)][C20(z1,z2)- C21(z1,z2)],

DA4"(z1,z2) = [A0(z1,z2) - A1(z1,z2)][ B0(z1,z2)+B1(z1,z2)][C30(z1,z2)+C31(z1,z2)], ..(25)

DA5"(z1,z2) = [A0(z1,z2) - A1(z1,z2)][ B0(z1,z2)+B1(z1,z2)][C30(z1,z2)- C31(z1,z2)],

DA6"(z1,z2) = [A0(z1,z2) - A1(z1,z2)][ B0(z1,z2) -B1(z1,z2)][C40(z1,z2)+C41(z1,z2)],

andDA7

"(z1,z2) = [A0(z1,z2) - A1(z1,z2)][ B0(z1,z2) -B1(z1,z2)][C40(z1,z2)- C41(z1,z2)]

Finally, this directional decomposition stage is cascaded with the predesigned 2-D circularsplit scheme (CSS) to form the total structure of the proposed CCT and the output of the ith

band is given by

Yi(z1,z2) = DAi"(z1,z2) X(z1,z2) for i = 0, 1, 2, …, 8. (26)

where X(z1,z2) is the scaled image from the CSS stage.

It can be seen that the resulting structure of Fig. 13 is a regular and a modular one. Suchproperties make it suitable for VLSI implementation.


40

Fig. 13 8-band directionally-decomposed analysis bank

V. ASSESSMENT AND COMPARATIVE STUDY

In this section, the performance of the proposed CCT transform is evaluated andcompared with that of the ordinary CT transform. A test image as the one shown in Fig. 14(a)is applied as an input to both CT & CCT transforms. The resulting directional details of theeight bands for both CT & CCT transforms are shown in Fig. 14(b). In such figure, it can beseen that due perfect circular and directional decompositions of high frequency bands in CCT,each band simulates the details in its direction with a better high-low frequency resolutionbecause of using CSS instead of LP. This indicates the superiority of CCT in preserving imagedetails. The objective performance of the proposed CCT transform is also evaluated viacalculating an assessment parameter which is called Deflection Ratio (DRi) at each directionalband i, for i = 0, 1, 2, …, 8. DRi is used here as a performance estimator. A proposed formulafor this deflection is given by [15],[16]

(27)

where represents the ith detail image pixels of the ith resulting CT or CCT bands. andR*C is its size. Also

(28)and

(29)

MVi is the mean value with SDi as the standard deviation, of the same ith detail image pixels. Itshould be noted that, the ratio DR should be higher at pixels with stronger reflector points andlower elsewhere. Table-1 illustrates these DRi values for different CT and CCT directionalbands. From such table, the value of DR in each band of CT is modified in the case of CCT,referring to accurate definition of image details in that direction. It is believed that, the use ofCSS instead of LP, is also the reason of such modifications. The previous properties nominate


41

the proposed CCT as a candidate instead of the classical CT to be utilized in those fields ofimage processing (such as denoising, compression and classification) which need better edgerepresentations.

(a)

(b)Fig. 14 (a) Original image, (b) It's directional details of the eight bands for both CT & CCT.

Table-1 Directional ratio DRi values for different CT and CCT directional bands

DIRECTIONAL BAND I=

0 1 2 3 4 5 6 7

DRI WITH CT 9.01X10-4 0.9179 0.0876 0.9118 0.0884 0.9114 0.0904 0.9093DRI WITH CCT 2.87X10-5 0.8970 0.1031 0.8968 0.1032 0.8968 0.1033 0.8967

VI. CONCLUSIONS

A proposed circular contourlet transform based on both circularly- and directionally-support decomposition structures has been designed and realized utilizing lattice all-passsections. The idea is based on using a circular-split scheme (CSS) followed by multiresolutionDFB with many levels of decomposition. The proposed circular contourlet transform has beendiscussed and an execution algorithm has been adopted for the calculation of such transform.The details of a test image) has been analyzed via this circular contourlet transform. Theresulting detailed images have been compared with the corresponding detailed images due tothe application of the classical contourlet transform. From objective measures point of view,the resulting assessment parameter, Deflection Rate (DR) has indicated the superiority of suchproposed CCT for perfect detail preservation. In addition to that, the resulting subband imagesfrom the proposed CCT are visually better where significantly more levels of detail areretrieved. The comparison also indicates that the application of the proposed circularcontourlet transform results in more continuous contours (edges). On the other hand, fromrealization point of view, it is believed that the resulting structures preserve both modularityand regularity properties which are suitable for VLSI implementations. Besides, since suchCCT is realized in a lattice all-pass sections, then some saving in computation can be gained,


42

while lattice structure will provide the system with a reduced sensitivity to finite word lengtheffect.REFERENCES[1] M. Antonini, M. Barlaud, P. Mathieu, and I. Daubechies, "Image compression usingwavelet transform", IEEE Trans. on Image Proc., Vol. 1,No, pp.205-220, April 1992.[2] M. Vetterli, J. Kovacevic, "Wavelets and subband coding", Englewood Cliffs, NJ: PrenticeHall, 1995.[3] S. Mallat, "A Wavelet Tour of Signal Processing", Academic Press, New York, 1998.[4] V. Velisavljevic, P. L. Dragotti, and M. Vetterli," Directional wavelet transforms andframes", Proc. IEEE Int. Conf. on Image Proc., 3:589-592, Sept. 2002.[5] M. N. Do and M. Vetterli, “Contourlets", J. Stoeckler, G. V. Welland (Eds.), BeyondWavelets, pp.1-27, Academic Press, 2003.[6] M. N. Do and M. Vetterli," The Contourlet transform: An efficient directionalmultiresolution image representation", IEEE Trans. on Image Proc., Vol. 14, No. 12, pp.2091-2106, Dec. 2005.[7] J. Rosiles and M. J. Smith,” Image denoising using directional filter banks", Proc. IEEEInt. Conf. Image Proc.-2000, pp. 292-295, 2000.[8] M. N. Do, ”Directional multiresolution image representations", Ph. D. thesis, EPFL,Lausanne, Switzerland, Dec. 2001. on http: //www. ifp.uiuc.edu/~minhdo/publications/thesis.pdf.[9] T. T. Nguyen and S. Oraintara, " A multiresolution directional filter banks for imageapplications,” Proc. Int. Conf. Acoust., Speech & Signal Proc., Montreal, QC, Canada, May2004.[10] T. T. Nguyen,” Multiresolution direction filter banks: Theory, design and applications”IEEE Trans. on signal proc., Vol.53, No. 10, Oct. 2005.[11] B. A. Taha," Directional wavelets; theory , design and application", Ph. D. Thesissubmitted to the college of Science, Dept. of Mathematics, University of Basrah, Nov. 2007.[12] J. M. Abdul-Jabbar, “Design procedure of two-dimensional digital filter and filter bank,”Ph. D. Thesis submitted to the college of Eng. , Dept. of Elect. Eng., University of Basrah,Sept. 1997.[13] S. Park, M. J. Smith, and R. M. Mersereau, ”Improved structures of maximally decimateddirectional filter banks for spatial image analysis,” IEEE Trans. on image proc., Vol. 13, No.11, pp. 1424-1431, Nov. 2004.[14] A. N. Willson and H. J. Orchard, ”Insights into digital filters made as The sum of twoallpass functions ,” IEEE Trans. Circuits Syst. -I: fundamental theory and applications, Vol.42, No. 3, pp. 129-136, Mar.1995.[15] M. Mastriani and A. E. Giraldez," Kalman’s Shrinkage for Wavelet-Based De-specklingof SAR Images", International Journal Of Intelligent Technology, Vol. 1, No. 3, pp. 190-196,2006.[16] M. Mastriani, A. E. Giraldez," Smoothing of coefficients in wavelet domain for specklereduction in Synthetic Aperture Radar images", The International Congress for GlobalScience and Technology (ICGST), International Journal on Graphics, Vision and ImageProcessing (GVIP), GVIP Special Issue on Denoising, pp.1-8, 2007. on www.icgst.com


http://:@www.icgst.com/

Ahmed: Effect Of High Temperature On Mechanical Properties Of Concrete Containing Admixtures

43

EFFECT OF HIGH TEMPERATURE ON MECHANICALPROPERTIES OF CONCRETE CONTAINING ADMIXTURES

AHMAD,A.H. (Professor) Abdulkareem,O.MCivil Engineering Dept. University of Mosul

ABSTRACT

This research work includes an experimental investigation to study the effect of hightemperatures on the mechanical properties of concrete containing admixtures. A comparativestudy was conducted on concrete mixes, reference mix without an additive and that with anadmixture. Concrete was exposed to three levels of high temperatures (200,400,600)° C, for aduration of one hour, without any imposed load during the heating. Five types of admixtureswere used, superplasticizer, plasticizer, retarder and water reducing admixture, an acceleratorand an air entraining admixture.

Mechanical properties of concrete were studied at different high temperatures,including: compressive strength, splitting tensile strength, modulus of elasticity and ultimatestrain. Test results showed a reduction in the studied properties by different rates for differentadditives and for each temperature, the decrease was very limited at temperature up to(200°C) but was clear at (400,600)° C.

Key words: concrete containing admixtures, high temperatures, mechanical properties.

, )(,)( ,

.ºC)600, 400, 200(

. .

.

)ºC200 ( ºC)400, 600.(



44

1. INTRODUCTION

Concrete acts as an unflammable construction material; however most of itsmechanical properties are changeable due to chemical and physical changes that may occurdue to high temperatures effect, such as concrete strength, modulus of elasticity etc., althoughthese effects occur in different manners, this causing a disadvantage in the properties of thebuilding materials (concrete and steel reinforcement) which may be controlled by severalmeans such as : modification of the relative codes of practice. Therefore; this research workaims to evaluate properties of the concrete containing (superplasticizer, normal plasticizer,retarder and water reducing, accelerator and air entraining admixture) when subjected to hightemperatures of (200,400,600) C.

In (2002) Tolentino et al. [1] analyzed the residual performance of Portland cementconcretes heat-treated at (600) C after cooling down to room temperature. Concretes withcharacteristic compressive strength at (28) days of (45) MPa and of (60) MPa were studied.The heat-treatment was carried out without any imposed load. Researchers measured theresidual compressive strength and modulus of elasticity. The geometry of the structure wasdescribed by mercury intrusion porosimetry and nitrogen sorption tests. They observed adecrease of residual compressive strength and modulus of elasticity, with the raise of heat-treatment temperature, as a result of heat-induced material degradation. The results alsoindicated that the microstructural damage increased steadily with increasing temperature.Based on the results of this experimental work they concluded that residual mechanicalproperties of concrete are dependent of their original non heat-treated values.

In (2004) Yüzer et al. [2], carried out a study on the effects of fire and extinguishingon the properties of concrete, mortars with and without silica fume were exposed to differenttemperatures, such as (100, 200, 300, 600, 900 and 1200) C and cooled slowly in the air andfast in water in two groups. Flexural and compressive strength tests were performed on thesamples which were cooled up to room temperature and changes in compressive strength incolor were determined by Munsell Color System. High temperature has caused damages indecrease in mechanical strengths at (600) C. Researchers observed that the changes in color'shue component and the compressive strength have similarities. Test results show that residualcolor changes in mortar can give an idea about the effect of high temperatures on mechanicalproperties of mortar during a fire.

In (2005) Savva et al. [3], have studied the influence of high temperatures onconcrete mechanical properties and properties that affect the measurement by non –destructive methods (rebound hammer and pulse velocity) of concrete containing variouslevels (10% and 30%) of pozzolanic materials. Three types of Pozzolans, one naturalpozzolan and two lignite fly ashes (one of low and the other of high lime content) were usedfor cement replacement. Two series of mixtures were prepared using limestone and siliceousaggregates. The W/b and the cementitius material content were maintained constant for allthe mixtures. Concrete specimens were tested at (100, 300, 600 and 750) C for 2 h withoutany imposed load, and under the same heating regime. At the age of (3) years, tests ofcompressive strength, modulus of elasticity, rebound hammer and pulse velocity were comeout. Results indicate that the residual properties of concrete as well as rebound and pulsevelocity versus heating temperatures are established. The above results are evaluated toestablish a direct relationship between non-destructive measurements and compressivestrength of concrete exposed to fire.


45

2. RESEARCH SIGNIFICANCE

The aim of this investigation is to study the influence of exposing to high temperatures onsome mechanical properties of concrete containing admixtures. The Iraqi standardspecifications have not refer to effect of temperature on concrete containing admixtures. TheBritish specifications and (ASTM) discuss the effect of temperature on concrete and theeffect of admixtures on concrete independent each other, but had not taken the effect oftemperature on concrete containing admixtures, therefore necessity of such study. It is expected that each admixture added to concrete has a different effect on themechanical properties, under the influence of high temperatures.

3. EXPERIMENTAL PROGRAM3.1. MaterialsCement:

Locally available ordinary portland cement produced in "Badoosh cement factory" according to the Iraqistandard specification (IQS, No.5, 1984) was used throughout the research work. The chemical andmineralogical analysis of the cement is shown in table (1).

Table (1): Chemical and Mineralogical Analysis of CementChemical Composition (IQS: 5/1984)(%) ASTM C150 (%)

CaO 62.76SiO2 21.50

Al2O3 6.20MgO 2.75 (5.00 maximum) 6.00 maximumFe2O3 3.25SO3 3.00 (2.80 maximum) 3.00 maximum

Mineralogical Composition (IQS: 5/1984) (%) ASTM C150 (%)C3S 46.24C2S 27.18C2A 11.56 (5.00 minimum)

C4AF 7.61

Water:Ordinary drinking (tap) water from Mosul area was used in all concrete mixes of this study.

Fine aggregate: River aggregate obtained from "Al-Khazir" near by Mosul area was used with finenessmodulus of (1.84) and a gradation compatible to the Iraqi standard specification (IQS, No.5,1980). The characteristics of fine aggregate is shown in table (2).

Coarse aggregate: River gravel available in the suburbs of Mosul area with a maximum aggregate size of(20mm) was used; compatible to the Iraqi standard specification (IQS, No.5, 1980). Thecharacteristics of coarse aggregate is shown in table (2).

Table (2): Properties of Fine and Coarse Aggregate

Aggregate Type Specific Gravity % AbsorptionSSD Oven DryFine 2.63 2.59 1.5

Coarse 2.65 2.63 0.8


46

3. 2. Fresh ConcreteThe concrete mix is prepared as follows:1) Design of the mix : Mix was designed according to British method (Department of Environment (DOE)), with aspecified compressive strength of (30 MPa) at (28) days and a slump of (100 mm) on thebasis that both aggregate (fine and coarse) were in a saturated surface dry state. The mixingproportions are (1:1.14:3.07/0.54) (cement: fine aggregate: coarse aggregate : w/b= w/cratio), with a cement content of (415 kg/m3). The additives used were in an amount asrecommended by the manufacturer and as shown in table(3):

Table (3): Dosage of Used Admixtures

Type of Additive

Superplasticizer(Rheobuild

800)

Plasticizer(Pozzolith

322N)

Retardingand WaterReducing(Pozzolith

100Ri)

Accelerator(ColloidalSilicates)

AirEntraining

(Polyrex SP32/1

+ IsorexR 310/1)

Dosage(mls/100kg of

cement)1000 327.5 195 437.5 1666.7

2) Casting of the specimens:After the concrete had been prepared according to the design outcome; (100x100x100) mmcubes and (100x200) mm cylinders specimens were cast three specimens for each test, (72)specimens were be tested at the age of (28) days. The specimens were cured in water for (28) days at room temperature (20±1) C. At the endof the curing period, the specimens were left to dry in the air for (2 hrs) prior to heating.

3. 3. Concrete Heating and Cooling Process

The concrete specimens were heated to different levels of high temperatures; using anelectrical furnace with a maximum temperature of (1200 C). The furnace is consisted of widechamber of a double metal containing auto-control thermal probes; with built inthermocouples. The temperature of the furnace increases by an average value of (5 C/min) atits primary stage up to (200 C), becoming faster to about (10 C/min) at the requiredtemperature. The concrete specimens were then placed inside the furnace for one hour at aconstant temperature; after that the specimens were left for (24 hrs) to be air cooled.

3. 4. Tested Parameters

The following mechanical properties were investigated on heat-treated specimens:

1) Compressive strength test was carried out on (100x100x100) mm cubes according toBritish standard specification (BS1881:Part116: 1983).

2) Splitting tensile strength test was performed on cylindrical specimens (100x200) mmaccording to (ASTM C496-71).

3) An electrical strain gauge (YL-10) of (10 mm) length was used to measure the strainchanges in a concrete cube due to changes of stress up to ultimate strength. The strain wasmeasured by a digital strain meter (TDS-301). Compressive stress-strain relation is thenpresented.


47

4. RESULTS AND DISCUSSION

4.1. Compressive Strength:Figure (1) shows the change of the residual compressive strength in concrete mixes at an ageof (28) days during the temperatures rise.

In general, the compressive strength of different concrete mixes is decreased byvarious proportions as a result of exposure to high temperatures. For reference concrete mixat (28) days, the residual compressive strengths are about (95%, 83%, 42%) at temperature of(200,400,600) C respectively. The residual stresses for concrete containing additives rangedfrom (75%-94%) at a temperature of (200 C). The highest stress was that for concretecontaining the plasticizer; while the mix containing air entraining additive maintained thelowest proportion (75%). At a temperature of (400 C), the residual compressive strength forconcrete containing additives ranged between (59%-88%). Retarding and water reducingadditive maintained the highest proportion of compressive strength (88%). The acceleratingadditive maintained the lowest proportion of (59%). At a temperature of (600 C) theseproportions ranged between (31%-50%). Concrete containing retarding and water reducingadditive is maintaining a highest residual strength, about (50%); whereas the lowest residualstrength (31%) is for that containing superplasticizer.

Test results show different losses in compressive strength at different temperaturesand different additives. The obvious loss begins at temperature after (300 C) whichrepresents the critical temperature at which the strength loss would be rapid and continueswith temperature rise [4]. It is well-known that the concrete consists of discrete aggregatedispersed in a continuous cement paste matrix, and that the transition zone between cementpaste matrix and the aggregate is considered to be a critical zone and evidently affectsconcrete performance exposed to high temperatures[1]. It is noticed that changes occurring inthis zone is responsible for the loss of compressive strength through the pore structure coarsefor the cement paste structure and aggregate , as a result of non –thermo identificationbetween the cement paste and the aggregate which consequently causes a bond failurebetween the cement paste and aggregate surface; in addition to the effect of the vaporizedwater pressure during the heating process [1], the chemical changes occurring in this zone –represented by the loss of free moisture, the increase in calcium content (C-S-H) and fromelongation of calcium hydroxide crystals spreading tightly in cement paste at temperaturesrise and so on[5]. Besides other chemical and physical changes such as shrinkage occurringin cement paste or mortar and aggregate expansion which leads to the appearance of micro-cracks in the concrete [6] .

It can be noticed that all concrete mixes containing admixtures, except that containingretarding and water reducing additive, tend to loose higher percentage of its compressivestrength at a heat- treatment compared to normal concrete.

4.2. Splitting Tensile Strength:Figure (2) shows the change in the rate of residual splitting tensile strength in concrete mixesat the age of (28) days with temperatures. It can be observed that the rate of the residualsplitting tensile strength for the reference concrete mix reached (95%) at a temperature of(200 C), reduced to (71%) at a temperature of (400 C), and became (47%) whentemperature rise to (600 C).

The residual splitting tensile strength in concrete mixes containing additives rangedbetween (75%-94%) at temperature of (200 C). The mix containing superplasticizermaintaining a higher proportion of this strength compared to the reference mix (94%);whereas the mix containing the accelerating additive maintaining the least proportion (75%).


48

At temperature of (400 C) the residual splitting tensile strength for concrete mixescontaining additives ranged between (42%-80%). The highest residual rate of splitting tensilestrength is for concrete containing plasticizer, compared with the reference mix that is about(80%), and the least value (42%) is for the mix containing air entraining additive.

At temperature of (600 C) the residual splitting tensile strength ranged between(25%-85%), the mix containing an accelerating additive maintained the higher rate, reached(58%), whereas the mix containing an air entraining additive maintained the least ratio (25%)of this strength. It has been found that the splitting tensile strength is more sensitive to heat-exposure than the compressive strength, for example, the rates of residual compressivestrength in the mix containing superplasticizer are (86%,69%, 31%),while the rates ofresidual splitting tensile strength in the same mix are (94%,77%,41%) at (200,400,600) Crespectively.

4.3. Modulus of Elasticity:The secant elastic modulus of the different mixes was calculated from the stress-strainrelation produced previously according to (ACI Code) [7].

Figure (3) shows the variation of the residual secant elastic modulus with hightemperatures. Figure (4) shows that the residual percent of the reference mix at (200 C)equals (57%) of elastic modulus at its initial value, then changed to (30%) at a temperature of(400 C) and (26%) at (600 C).

0

20

40

60

80

100

120

0 100 200 300 400 500 600 700

Temperature (oC)

Perc

ent R

esid

ual o

f Com

pres

sive

Stre

ngth

Reference MixMix Containing SuperplasticizerMix Containing PlasticizerMix Containing Retarding and WRAMix Containing Accelerating AdmixtureMix Containing Air Entraining Admixture

Figure (1): Change in the rate of theresidual compressive strength in concretemixes at (28) days with the temps rise.

Figure (2): Change in the rate of residualsplitting tensile strength in concrete mixes at(28) days with the rise of temp.

0

20

40

60

80

100

120

0 100 200 300 400 500 600 700

Temperature (oC)

Perc

ent R

esid

ual S

plitt

ing

Ten

sile

Stre

ngth



49

0

100

200

300

400

500

600

0 100 200 300 400 500 600 700

Te mperature (oC)Pe

rcen

t of U

ltim

ate

Stra

in

Reference MixMix Containing Superplas ticizerMix Containing PlasticizerMix Containing Retarding and WRAMix Containing Accelerating AdmixtureMix Containing Air Entraining Admixture

On the other hand, the values ranged between (10%-89%) for the mixes containingadditives at a temperature of (200 C). The mix containing air entraining additive maintainedthe highest proportion of (89%) compared to the reference mix; whereas the mix containingplasticizer additive maintained the least value (10%). At a temperature of (400 C), theresidual elastic modulus ranged between (27%-54%) for concrete containing additives. Themix containing retarding and water reducing admixture maintained the highest rate of (54%);whereas the mix containing a plasticizer, the residual changed to (27%). At a temperature of(600 C) the residual proportions ranged between (12%-35%); the mix containing retardingand water reducing admixture maintained the highest proportion (35%) compared to thereference mix; whereas the least proportion (12%) is maintained by the mix containingplasticizer.

The elastic modulus is related to its sensitivity of the aggregate stiffness more than tothe cement paste stiffness [8]. However; there are – in general – several factors affecting theconcrete elastic modulus exposed to high temperatures. These factors are [9]:

1- Strength of concrete.2- Type of aggregate and its elastic properties.3- Water-cement ratio.4- Type of cement.5- Sustained stresses.

On top of that, different types of additives may be taken as one of the factors affectingthe elastic modulus.

4. 4. Ultimate Strain:Figure (4) shows variation of ultimate strain with high temperature for concrete mixes at theage of (28) days. It can be found that strain increases gradually with increasing temperaturefor all mixes.

0

20

40

60

80

100

120

0 100 200 300 400 500 600 700

Temperature (oC)

Perc

ent R

esid

ual M

odul

us o

f Ela

stic

ity


Figure (3): The variation proportion of residual elasticmodulus of concrete mixes at an age (28) days with the

temperature rise.

Figure (4): The extent of ultimate strainvariability for concrete mixes at an age of (28)days with the rise of the temperatures.


50

Table (4) show the percentage increases in ultimate strain for mixes having differenttype of additives for temperatures (400,600) C. It can be concluded that the effect of hightemperatures on ultimate strain varied and depends on type of additives.

Table (4): Percentage Increases in Ultimate StrainTemp

Type

Reference

Concrete

Super

plasticizer

Plasticizer Retarding &

water reducing

Accelerating

admixture

Air entraining

admixture

400Cº 178 155 188 195 146 136

600C 212 394 507 248 180 198

No significant effect of additives on ultimate strain at a temperature of (200 C).While at (400 C) the increase in ultimate strain reduced from that of reference mix for theadditive, air entraining, accelerator and superplasticizer, while the others having small effect.The increase in ultimate strain relative to the reference mix is obvious at (600 C) in the twoadditives (superplasticizer and plasticizer), so that increase in ultimate strain reached (394%,507%) respectively.

4.5. Compressive Stress-Strain Relations:The compressive stress-strain readings were presented in table (6) for all concrete mixes atthe age of (28) days and for different temperature levels, each value was an average of threetested specimens.

The compressive stress-strain relations are represented for each mix and differenttemperatures by a suggested empirical formula, the data analyzed, using the statisticalanalysis program (SPSS 11.5). The general form of the empirical formula is given as in table(5), show the coefficient of equation (1) for the additives used:

= X1* X2 +X3* TX4 +X5* +X6* T ……(1)

Table (5): The Coefficient Of Equation (1) For Concrete Mixes Used

To study the effect of the additives used in this research work, the compressivestress – strain curves are presented using the experimental data and the numerical equations(1), and table (5) for temperature ( 20, 200, 400, 600 ) C° as shown in figures ( 5 To 10 ).

Additives

Parameters

ReferenceConcrete

SuperPlasticizer

Plasticizer

Retarding &

water

reducingAccelerator

AirEntraining

X1 25.857 65.8800 55.7800 215.3850 17.8350 23.028

X2 0.4795 0.6850 0.5865 0.9127 0.4244 0.4167

X3 -0.0889 -0.0905 3.9220 0.0930 -.0888 -0.0849

X4 -5.7540 -5.7699 0.9950 -5.7620 -5.7500 -5.7530

X5 -4.9600 -27.2960 -14.0900 -176.6560 -2.7700 -3.6960

X6 -0.0265 -0.0567 -3.8700 -.0458 -0.0213 -.0242

R² 0.9130 0.931 0.9600 0.942 0.9586 0.9690


51

Table (6): Compressive Stress-Strain Readings.Concrete

Type

(20)oC (200) oC (400) oC (600 )oC

Stress

(MPa)

Strain

(m/m x10-3)

Stress

(MPa)

Strain

(m/m x10-3)

Stress

(MPa)

Strain

(m/m x10-3)

Stress

(MPa)

Strain

(m/m x10-3)

Reference

Concrete

11.51

16.70

20.36

22.65

26.80

30.04

0.40

0.73

1.15

1.42

2.30

3.20

3.32

10.23

17.36

21.20

24.50

28.40

0.17

0.58

1.38

2.03

2.69

3.70

2.50

5.50

11.00

18.68

22.80

24.90

0.11

0.46

1.40

3.10

4.57

5.70

4.50

7.30

9.00

10.85

12.15

12.60

0.54

1.31

2.25

3.76

5.67

6.80

Super plasticizer

8.00

12.60

20.40

32.10

48.35

55.78

0.16

0.26

0.40

0.82

1.66

2.40

9.50

18.00

25.60

32.56

41.70

47.80

0.33

0.61

0.84

1.17

1.77

2.62

8.34

15.40

22.20

28.90

34.40

38.34

0.58

0.98

1.45

2.00

2.63

3.72

6.25

10.00

12.50

15.35

16.47

17.24

0.72

1.20

1.95

3.77

6.95

9.49

Plasticizer

15.35

23.75

33.45

44.80

55.10

58.62

0.19

0.41

0.69

0.11

1.85

2.25

12.21

16.63

23.00

29.20

42.23

54.90

0 36

0.53

0.73

1.00

1.51

2.48

6.00

12.60

20.20

28.40

34.80

38.84

0.56

1.06

1.63

2.45

3.28

4.22

10.00

17.30

21.60

23.60

25.22

26.29

1.76

3.38

5.31

7.17

9.46

11.43

Retarding and

Water Reducing

7.56

15.00

20.93

31.00

41.85

51.20

0.11

0.30

0.50

0.77

1.22

2.05

10.64

18.36

23.12

28.60

38.19

47.00

0.36

0.58

0.72

0.90

1.36

2.18

15.00

22.00

28.00

35.00

42.00

45.00

0.65

1.00

1.40

2.00

2.80

4.00

10.80

17.30

20.90

22.80

24.45

25.77

0.72

1.45

2.33

3.13

4.17

5.06

Accelerator

5.00

8.00

10.00

13.00

16.00

19.40

20.50

21.00

21.20

0.10

0.30

0.48

0.84

1.46

2.56

3.22

3.86

5.00

4.90

7.90

10.00

12.40

15.40

17.10

17.90

18.70

19.10

0.30

0.57

0.87

1.31

2.11

2.76

3.42

4.49

5.60

0.94

5.32

8.00

9.00

9.75

10.88

11.65

11.98

12.50

0.10

0.60

1.45

2.14

2.78

3.95

4.94

6.00

7.30

3.00

5.40

6.70

7.30

7.70

7.80

8.00

8.10

8.22

0.86

1.72

2.51

3.43

4.62

5.70

6.94

8.05

9.00

Air Entraining

11.36

16.64

19.77

22.50

24.87

26.18

26.92

0.46

0.73

0.93

1.47

2.35

3.10

3.48

8.23

13.13

15.68

17.64

18.72

19.35

20.19

0.35

0.71

1.06

1.64

2.33

2.93

3.59

6.82

10.62

13.45

14.96

16.03

16.57

16.96

0.58

1.03

1.60

2.44

3.26

4.00

4.73

3.96

7.25

8.99

9.98

10.73

11.88

13.19

0.69

1.35

2.01

2.77

3.64

5.16

6.95

* The results of table(5) are the average of three tested specimens for each.


52

Figure (7): The stress-strain relationshipsfor the mix containing plasticizer Figure (8 ): The stress-strain relationships for

the mix containing Air Intraining

Figure (5): The stress-strainrelationships for reference mix

Figure (10): The stress-strain relationships for the mix containing air entraining admixture

Figure (6): The stress-strain relationshipsfor the mix containing superplasticizer

0 2 4 6 8 10Strain e*103

Figure (6): The Stress – Strain relation for the Mix

containing Super Plasticizer

Figure (9): The stress-strain relationships forthe mix containing accelerating admixture


53

The trend of the compressive stress-strain relation for concrete containing acceleratingadmixture shows that this type of mix is more ductile than the reference concrete; The othertypes of mixes show a brittle concrete. All mixes become more ductile at higher temperature.

It is worth to mentioning that there are several factors affecting the behavior of(stress-strain) relationship including [9, 10]:

1- Aggregate-cement ratio.2- Curing conditions.3- Binder material type.4- Aggregate type.The type of additive may be added to the above factors.

5. CONCLUSIONS1- The existence of concrete additive in the concrete mix exposed to high temperatures

resulting variable changes in compressive strength, splitting tensile strength, elasticmodulus and ultimate strain, compared to the reference mix. These changes -ingeneral–varied from one additive to another for different temperatures; however itwas limited at (200 °C) and soon became evident at (400,600) °C.

2- The mix containing a plasticizer maintained the higher residual proportion of thecompressive strength at a temperature of (200 °C) ; while the mix containing theretarding and water reducing additive maintained the highest residual compressivestrength at temperatures of (400,600) °C.

3- The mix containing superplasticizer gives the highest residual splitting tensilestrength at a temperature of (200 °C). At temperature (400 °C) the highest value isobtained for mix containing plasticizer, and at (600 °C) it is found that the mixcontaining an accelerating additive is maintained the highest percentage of thisstrength.

4- The highest residual proportion of elastic modulus is obtained by the mix containingair entraining additive at a temperature of (200 °C) ; While at (400,600) °C thehighest value is obtained by the mix containing retarding and water reducing additive.

5- The mix containing an accelerating additive gave the highest proportion of an increasein the ultimate strain at a temperature of (200 °C); whereas the mix containingretarding and water reducing additive gave the highest increase at a temperature of(400 °C). The mix containing plasticizer gives the highest increase at a temperature of(600 °C).

6- It can be concluded that the effect of additives may be considered as one of the factorsaffecting the heat-treatment of concrete with other factors.

6. REFERENCES[1] E. Tolentino, F. Lameiras, A. Gomes, C. Silva and W. Vasconcelos, “Effects of High

Temperature on the Residual Performance of Portland Cement Concretes”, MaterialsResearch, 5 (2002), p. 1 – 11.

[2] N. Yüzer, F. Aköz and L. Öztürk, “Compressive Strength – Color Change Relation inMortars at High Temperature”, Cement and Concrete Research, (2004), p. 1 – 5.


54

[3] A. Savva, P. Manita and K. K. Sideris, “Influence of Elevated Temperatures on theMechanical Properties of Blended Cement Concretes Prepared with Limestone andSiliceous Aggregates”, Cement and Concrete Research, 27 (2005), p. 239 – 248.

[4] A. M. Neville, “Properties of Concrete”, Pitman Publishing Ltd, London ,(1975).[5] I. Janotka, and L. Bágel, “Pore Structures, Permeabilities, and Compressive Strengths of

Concrete at Temperatures up to 800ºC”, ACI Materials Journal, 99 (2002), p. 196 –200.

[6] G. T. G. Mohamedbhai, “Effect of Exposure Time and Rates of Heating and Cooling onResidual Strength of Heated Concrete”, Magazine of Concrete Research, 38 (1986), p.151 – 158.

[7] ACI Committee 318, “Building Code Requirements for Structural Concrete ACI”,American Concrete Institute, (2002).

[8] R. Ravindrarajah, R. Lopez, and H. Reslan, , “Effect of Elevated Temperature on theProperties of High Strength Concrete Containing Cement Supplementary Materials”,Center of Built Infrastructure Research, University of Technology, Sydney, Australia,(2002), p. 1 – 9.

[9] U. Schneider, “Concrete at High Temperature”, Fire Safety Journal, 13 (1988), p. 55 –56.

[10] I. Janotka, and T. Nürnbergerova, , “Effect of Temperature on Structural Quality ofHigh-Strength Concrete with Silica Fume”, Institute of Construction and Architecture,Slovak Academy of Sciences, Bratislava, Slovak Republic,

(2003), p. 1 – 8.


Mahammed: Finite Element Analysis of Unreinforced Masonry Walls

55

Finite Element Analysis of Unreinforced Masonry Walls

Mohammed S. MohammedAssistant Lecture

Civil Engg. Dep./College of Engg. /Mosul University

Abstract

In the present work nonlinear finite element program written in fortranlanguage to simulate the behavior of masonry wall under the action of monotonicloading has been developed. The masonry is modeled as a two-phase material, treatingbricks and mortar joints separately, thus allowing for nonlinear deformationcharacteristic and progressive local failure of both bricks and mortar joints. Theinfluence of the mortar joint is taken into account by using an interface cap model as apart of a rational unit-joint model able to describe cracking, slipping and crushing ofthe material. The capabilities of the program have been examined and demonstrated byanalyzing two different types of masonry wall. The accuracy of the analytical resultswas assessed by comparing them with the experimental results and shown to be good.Key words: Interface element, Masonry Wall, Nonlinear behavior.

–/

. . )Interface Cap Model (

. .



56

1-Introduction In recent years, there has been an interest study in the mechanics of unreinforced

masonry structures, with the aim to provide efficient tools for better understanding of theircomplex behavior. Mortar joints usually present a lower strength than masonry units, soexplaining the existence of plane of weakness along which cracks propagate to failure.Therefore, two main approaches have been developed for the constitutive description ofmasonry, usually known in the technical literature as macro-modeling and micro-modeling[1].In macro-modeling masonry, no distinction between the individual units and joints is made,and masonry is considered as a homogeneous, isotropic, or anisotropic continum. As macro-modeling of masonry is advantageous when the global behavior of the structure is important.The influence of the mortar joints acting as planes of weakness cannot be addressed.The alternative micro-modeling approach, expanded units are modeled with continuumelements, while the behavior of the mortar joints and unit-mortar interface is lumped asdiscontinuous line interface elements [2]. In this research micro-modeling has been adoptedin preference to the macro-model.The behavior of masonry wall under in plane loading has been the subject of manyinvestigations. Dhansekar etal [3] proposed nonlinear finite element model for solid masonrybased on average properties derived from biaxial tests on brick masonry panels.Page[4] presented a method that accounts the nonlinear behavior of masonry , where themasonry is considered as a two-phase material. Ali and Page [5] also used the method tostudy the nonlinear behavior of masonry subjected to concentrated loads. However all thesemodels include only tensile (brittle) and shear failure (brittle or elastic/ideal plastic) of thejoint. A number of plasticity-based continuous –interface models have been developed tomodel the tension and shear behavior of masonry –mortar joints [6].Lourenco [7] recently used both micro-modeling and macro-modeling to represente masonrywall. In this model the gradual softening behavior in the model for interface element wasused and the elastic model was used to represent behavior of the brick of wall.In this search, nonlinear bidimensional finite element models are used to simulate fracture inmasonry structures. Masonry consists of bricks, which is modeled with eight quadratic planestress elements. Interface elements are used to simulate the joints. The plasticity modelproposed by Lourenco [7 ] is used to formulate a modern algorithmic plasticity concepts.These include implicit Euler backward return mapping schemes and consistent tangentoperators .Including a correct handling of the corners. The model is formulated in the contextof non-associated plasticity. The analysis is carried out with a special arc-length control thatautomatically search for the largest relative displacement in the interfaces. Numericalimplementation of the model is evaluated by a comparism between numerical results with theexperimental results for the case of masonry wall with in plane loading.

2-Finite element model of the present study

In the finite element analysis conducted here, masonry is treated with micro-model, inwhich the units of brick and joints are modeled individually with different type of elements.The masonry units are modeled with smeared crack elements, which account for both tensileand compressive fracture of the units, while the mortar joints are modeled with interfaceelement to account for the inherent planes of weakness to include all the basic types of failuremechanisms that characterize masonry, see Fig. 1.


57

(d)

(b) (c)

(e)

(a)

v i,P y i

u i,P x iGeneral node,i

Parabplic shape anddisplacement variation

Fig.(1) Masonry failure mechanisms: (a) joint displacement; (b) joint slipping; (c) unitdirect tensile cracking; (d) masonry crushing; (e) unit diagonal tensile cracking.

2.1 Masonry unitsIsoperimetric plane-stress element with eight-noded and smeared crack pattern is

used to model the behavior of masonry units as shown in Fig. 2.

Fig.( 2) Typical 8-node isoperimetric element

Material nonlinearities due to cracking of concrete, plastic flow or crushing of the unit incompression are considered. The compressive failure and tensile fracture of masonry aregoverned by a von Mises failure surface with tension cutoff as shown in Fig.3, in which 1

and 2 are the principal stresses, mf and tf are compressive and tensile strength ofmasonry, and of determines the initial yield surface which is also governed by the VonMises criterion where assumed in this research to be mf5.0 .Before the tension cutoff surfaceis reached, the material is assumed to be elastic-plastic, of which the plastic behavior isrepresented by 2J plasticity as soon as the stress state reaches the initial yield surface. Thematerial exhibits a strain-hardening behavior when the stress state is between the initial yieldsurface and the final failure surface. Strain softening occurs once the final yield surface isreached. The von Mises failure criterion can be expressed as follows.

0)(22 peJ

(1)In which 2J is the second invariant of the deviatoric stress, and e and p represent theeffective stress and effective plastic strain respectively. A crack is initiated when the


58

maximum principle stress reaches the tensile strength, in direction normal to the maximumprinciple stress at each Gauss point. The cracked element is assumed to be nonlinearorthotropic material. The crushing type of unit is a strain-controlled phenomenon .A simpleway is used by converting the yield criterion in stresses into the yield criterion directly interms of the strain. The details of the plasticity model and smeared crack model can be foundin Ref. [10].

mf tf

mf

tf

of

of1

2

Fig.( 3) Yield and failure surface

2.2 Mortar joints Interface element with six node as shown in Fig.4 permit discontinuities in the

displacement field and their behavior is described in terms of a relation between the traction,and relative displacement , u , across the interface.

Fig.( 4) Interface element

The linear elastic relation between these generalized stress and strains can be written in thestandard form as:

eD (2)Where T, , sn

e KKdigD , and Tsn uu , ,with n and s denote the normal and

shear components, respectively. The elastic stiffness matrix eD can be obtained from theproperties of the two masonry components (unit and mortar) and the mortar thickness of thejoint .Due to the zero thickness inherent to the interface element formulation, the size of theunit has to be expanded by the mortar thickness, mh , in both direction (vertical andhorizontal) as shown in Fig. 5. Due to relative dimensions of mortar and unit, it is assumedthat the elastic properties of the unit remain unchanged.

Y,v

X,u


59

Fig.(5) Suggested modeling strategy. Units (u), which are expanded in both directionsby the mortar thickness, are modeled with continuum elements. Mortar joints (m) andpotential cracks in the units are modeled with zero-thickness interface elements.

The normal and shear stiffness required to define the material property matrix of the element,can be represented by the following expressions [7].

)(.

mum

mun EEh

EEK ,)(

.

mum

mus GGh

GGK (3)

Where uE and mE are the young’s modules, uG and mG are the shear moduli , respectively, for unit and mortar and mh is the actual thickness of the joint.The stiffness values obtained from formula do not correspond to a penalty approach, whichmeans that overlap of neighboring units subjected to compression will become visible. Thisfeature is, however, intrinsic to the interface elements formulation and is independent of thevalues of normal stiffness, even if it is clear that the amount of penetration will be higher withdecreasing interface stiffness. The interface model includes a compressive cap where thecomplete inelastic behavior of masonry in compression is lumped. This is aphenomenological representation of masonry crushing because the failure process incompression is, in reality, explained by the microstructure of units and mortar and theinteraction between them. For dry joint of masonry wall, when the mortar not used betweenunits, the stiffness must be represented by the following expressions.

brickwall

n

EEh

joK11

1int, ,)1(2

int,int,

joKjoK n

s (4)

Where h is the height of the block, wallE is the young’s modulus of the wall and brickE is theyoung’s modulus of the brick and is the Poisson’s ratio.The elastic domain is bounded by a composite yield surface that includes tension, shear andcompression failure see Fig.6. This model was developed by Lourenco [ 7 ]. This model hasbeen developed within the flow theory of plasticity. The composite yield surface is definedby three yield functions, where softening behavior has been included for all modes(tension,shear and compression modes) .


60

Fig.(6) Adopted model for interfaces. an “interface cap model”. [8]

The stress rate vector can be expressed as:)( peee DD

(5)And the following yield functions have been adopted: Tension criterion: )(),( tttt KKf(6) Coulomb friction criterion: )(tan),( ssss KKf(7) Compression cap criterion: )()(),( 2/1

ccT

cc KPKf(8)Where represents the friction angle and P is the projection diagonal matrix

ssnn CCdiag 2,2 with ssC and nnC a set of material parameters. st , and c are theisotropic effective stress of each of the adopted yield functions, ruled by scalar internalvariables st KK , and CK . The evaluation lows of these internal variables are giving by:

cc

PT

cststnt KuKuK ,,

(9) And the plastic strain rate vector reads:

gP

(10)Where is the plastic multiplier rate , g is the plastic potential function. Associated flow ruleis assumed for tensile and cap modes and anon-associated plastic potential sg is adopted forthe shear mode with a dilatancy angle and cohesion C , given by

ccstt fgCgfg ,tan,(11)In the particular case of dry masonry joints, the tensile strength and cohesion of the joints areassumed to be equal to zero

.


61

2.2.1 Integration of the elasto plastic equation:The return mapping algorithm [9] in strain is driven and basically consists of two

steps, first the calculation of the elastic trial stress , also called the elastic predictor, andscond the return mapping to the yield surface, the plastic corrector.The stress update for each individual yield surface is obtained as [10]:

Pnnn

ennnnn DD 11111 (

(12)This equation can be recast as:

1111

nn

Trialnn

gD

(13)Further more, since 1n and 1nK can be expressed as function of 1n , the yield function istransformed into nonlinear equation of one variable 0)( 1nf , which is solved locallyusing the Newton-Raphson method. Considering that only one surface is active.The consistent tangent stiffness matrix for each individual surface is obtained according to:

gHh

HgHH

ddD

T

T

ep

..

...

(14)Where

2

2

11 gDH n ,

1n

KKff

1n

KKfh

(15)

In the described composite yield criterion, the intersection of the different yield surfacesdefines two possible corners, see Fig. 7, composed by the tensile and shear modes or by theshear and cap modes, due to the intersection between cap and tensile modes is numericallyprevented from occurring.

Fig.(7) Composite yield surface [7]

In this model, tensile and shear softening are coupled because, physically, both phenomenaare related with the degradation of bond between the unit and the mortar and due to physicalreasoning, shear and cap modes are assumed to be uncoupled, since phenomena that rule the


62

hardening/softening of each mode seem to be only lightly related. Thus equations (9) 2, 3remain unchanged.The plastic strain rate in the corner is given by a linear combination of plastic strain rates ofyield surfaces 1 and 2

212121

ggppp (16)

At this corner, quadratic combination for the hardening parameters rate isadopted, which read,

2

22

11 )(t

IIf

Ifc

fC

GG

K ,

2

12

22 )(t

IIf

Ifc

fC

GG

K (17)

Where IfG is fracture energy (from tension test) for mode I and II

fG is fracture energy (fromshear test) for mode II.The stress update is then given by

11,2

11,111,21,111

21),(n

nn

nTrialnnnnn

gDgD

For the corner regime, the Euler backward algorithm [10] can be simply expressed in a

system of two nonlinear equations on the variables 1,1 n and 1,2 n

0.0),( 1,21,11,1 nnnf (19)0.0),( 1,21,11,2 nnnf

These equation are Solved by a Newton-raphason procedure. The Jacobian matrix is given inLourenco [8].For multi-surface plasticity an expression equivalent to Equation (14) can alsobe obtained. The reader is referred to Lourenco [8], where such expression can be found.

2.2.2 A nonlinear Solution Technique

The global nonlinear equations of equilibrium are solved using an incremental-iterative technique performed under displacement control by using arc-length method.iterative techniques use the standard and modified Newton-Raphson method. An automaticload incrimination scheme is included. As shown in the previous section the theory of multisurface plasticity is used to define the material behavior of interface element. UnconditionallyElure backward algorithms are derived for all models of the cap model. The Eluer backwardreturn mapping is solved using a local Newton-Raphson method for nonlinear equations ofinterface element only. A rail and error procedure to solve the return mapping is used. Foreach step, the following algorithm is performed [11]:1- Compute elastic trail stresses )( o . Check for plastic behavior; If 0),( 11 nn Kf , the integration point is elastic, update stresses and strain, Exit; If 0),( 11 nn Kf , the integration point is plastic see Fig.8;

2- If the integration point is plastic; Iterations J=0… n

jj

j

gDff

)/()/(()(

jp gD )/(

(18)


63

n

t r i a l

. )(11

i n cin

. )(21

i n cin

. )(31

c o rin

t r i a l

n. )(1

1i n ci

n

. )(21

c o rin

Cornerzone

Individual surfacezone. )(1

1i n ci

n. )(3

1c o ri

n

Incorrect return mapping

correct return mapping

Yield surface

pjj 1

Until ToleranceKf nn ),( 11

if a converged state is found compute consistent tangent operator.

Fig.(8) Return mapping algorithm (different case).

Numerical ExamplesConfined masonry wall

A two masonry shear walls with same property and dimension (J4D, J5D) wascarried out by Raijmakers and Vermeltfoort [8]. The width /height ratio (L/H) of shear wallsis 990/1000 (mm/mm); the walls were built up with 18 courses of bricks, from which 16courses were active and 2 were clamped a stiff steel beam, Fig.9. The brick dimensions are210*52*100 3mm and the mortar joints are 10 mm thick. The vertical load ( 23.0

mmNP )

was applied on the top and their resultant was kept constant during the complete horizontalloading procedure. The stiff steel beam did not allow rotations of the top and wassubsequently pushed within increasing horizontal force. The micro-properties for thedifferent materials according to Lourenco [8] are given in table.1 and table.2.

Table.1 Elastic properties for the bricks and joints.Brick Joint

E nK sK16700

2mmN 0.15

82

3mmN

36

3mmN

Table .2 Inelastic properties for the joints.Tension Shear Cap

tfIfG C tan tan II

fG mf ssC

0.252mm

N0.018

2mmNmm

tf4.1

2mmN

0.75 0.0 0.1252mm

Nmm10.5

2mmN

9.0


64

Fig.( 9) Loads Walls: (a) phase 1 - vertical loading;(b) phase 2 - horizontal loading under displacement control[8].

The crack patterns for the walls tested are shown in Fig. 10 while the finite element meshsystem is shown in Fig. 11.

Fig.(10) Walls Experimental crack Fig.(11) Finite Element Meshes ofWall - -patterns for different tests[8].

Fig.12 present the horizontal load-displacement curves for the top steel beam for the wallstudied. In this figure, the experimentally obtained results are compared with the numericalanalysis. This gives a good impression about the numerical implementation because it ispossible to trace the response of the structure through initial cracking, failure loads behavior.The comparison with the experimental failure loads is shown , good agreement is found sincethe difference between predicted and observed result is less than 3%.

Fig.13 shows the deformed shapes of the finite element models at 0.75 mm and 3.0 mm,respectively. This figure shows opening and slip along the mortar joints. the crack starts inthe middle of the wall under increasing deformation, and progresses in the direction of thesupports and ,finally a collapse mechanism is formed with crushing of the compressed toesand under the steel beam at the top.


65

0 . 0

5 0 .0

1 0 0 .0

1 5 0 .0

0 . 0 1 .0 2 .0 3 .0 4 .0

H o r iz o n t a l d is p la c e m e n t [ m m ]

Hor

izon

tal f

orce

[kN

]

E x p e r i m e n ta lN u m e r i c a l

Fig.(12) Load-displacement diagram.

Fig.( 13) Deformed mesh for JD wall at horizontal displacement equal to: (a) 0.75mm; (b) 3.0 mm.`

Unconfined masonry wallThe analyzed wall has a thickness of 200mm, with square panel 1000*1000 2mm and

made of stone blocks laid without mortar between them for two masonry wall with sameproperty and dimension tested by Oliveira [12] as shown in Fig.14.in addition to the deadweight, the walls were first loaded by a vertical load of 30 kN applied at the top, after whicha horizontal load was progressively applied at the top beam. The experimental crack patternsfor the tested walls are shown in Fig.15. In the particular case of dry stone masonry joints, thetensile strength and cohesion are assumed to be equal to zero. The material property aregiven in table.3 and table.4.

(a) (b)


66

Table.3 Elastic properties for the bricks and joints.[12]

Table .4 Inelastic properties for the joints.Shear Cap

tan tan IIfG mf ssC

0.62 0.0 0.1252mm

Nmm

6.02mm

N2.0

Fig.(14) Adopted geometry for the dry stone masonrywalls and schematic loading arrangement[12].

Fig.(15) Experimental crack patterns for different tests [12].

Brick Joint

E nK sK15500

2mmN 0.15

5.87

3mmN

2.45

3mmN


67

0 .0

5 .0

1 0 .0

1 5 .0

2 0 .0

2 5 .0

0 .0 5 .0 1 0 .0 1 5 .0

H o r iz o n t a l d is p la c e m e n t [ m m ]

Hor

izon

tal f

orce

[kN

]

E x p e r im e n t a l s w 3 0 .1E x p e r im e n t a l s w 3 0 .2N u m e r ic a l

Fig.(16) Comparison of curves relating the horizontal force (H) with thehorizontal displacement for walls SW30.1 and SW30.2.

Fig. 16. Illustrates the load-displacement diagram from the tested wall and the numericalresults, up to a displacement of 15 mm. the agreement between experimental and numericalresponses can be considered satisfactory.

Globally, the analysis captures well the experimental behavior of the walls, as illustrated inFig.17 together with the global load-displacement response, a comparison in terms of thedeformed mesh and failure pattern is necessary to appraise the quality of the numericalanalysis. At initial horizontal load, see Fig.17 it is possible to observe that separation of theblock through diagonal cracks gradually progresses from the bottom courses to the top,finally overturning failure mechanism is found with a complete diagonal crack through headand bed joints.

Fig.(17) Deformed mesh for SW wall at horizontal displacement equal to : (a) 0.75 mm; (b) 3.0 mm.

(a) (b)


68

Conclusions:This study presents an efficient finite element analysis technique which shows a

great versatility in analysis complex discontinuities in the analysis of masonry wallsstructures by use of interface elements with a constitutive model entirely established on thebasis of the incremental theory of plasticity to simulate the actual behavior at the interfacebetween contacting materials.A comparison between numerical and experimental result was also given. It was shown thatthe finite element method model was able to predict effectively the behavior of masonrystructures, with both confined and unconfined masonry wall, as well as sufficiently accuratecollapse load values.

Reference

1-Milani, G., Lourenço, P.B., Tralli, A., “ Homogenised Limit Analysis of Masonry Wall,Part I: failure Surface” ,Computers & Structures, 84(3-4),2006, pp. 166-180.

2-Asteris, P.G., Syrmakezis, C.A. ,“Strength of Unreinforced Masonry Walls UnderConcentrated Compression Loads” , Practice Periodical on Structural Design andConstruction, ASCE, Vol. 10, No. 2, 2005, pp. 133-140.

3-Tzamtzis, A.D, Asteris P.G., “Finite Element Analysis of Masonry Structures: Part I-Review of Previous Work”, Proceedings, Ninth North American Masonry Conference,South Carolina ,June 2003.

4-Page, A.W., “Finite Element Model For Masonry Structure”, J. Struct. Division ASCE,Vol.104, No.8, 1978, pp.1267-1285.

5-Ali, S. and Page, A., W.,” Finite Element Model for Masonry Subjected toConcentrated Loads”, J., Struct. Division ASCE, Vol.114, No.8, 1988, pp.1761-1784.

6-Al-Chaar,G.K and Mehrabi , A.B., “Constitutive Models for Nonlinear Finite ElementAnalysis of Masonry Prisms and Infill Walls”,ERDC/CERL TR-08-19.Champaign,IL:Engineer Research and Development Center-Construction EngineeringResearch Laboratory,2008.

7-Lourenco, P.B., “Analysis of Masonry Structures with Interface Elements Theory andApplication”. Report Nº 03.21.22.0.01, Delft University of Technology, Delft, TheNetherlands, 1994.

8-Lourenco, P.B., Rots, J.G., “A Multisurface Interface Model for Analysis of MasonryStructures” J. Eng. Mech., ASCE ,Vol. 123,No.7, 1997, pp. 660-668.

9-Chen, W. F., Lan , Y. M. and Sotelino, E. D. ,“The Strain-Space Consistent TangentOperator and Return Mapping Algorithm for Constitutive Modeling of ConfinedConcrete” International Journal of Applied Science and Engineering, Vol.1, No.1, 2003,pp.17-29.

10-Zienkiewicz,O.C. and Taylor ,R.L.,”The Finite Element Method for Solid andStructural Mechanics”, Sixth edition,2005.

11-Alfaiate.J. V., de Almeida, J. R. and Gago, A. S., “On the Numerical Analysis ofLocalized Damage in Masonry Structures” In 2nd International Conference onStructural Engineering and Construction (ISEC-02), Franco Bomtempi (Ed.), 769-774.,Roma, 2003.

12-Oliveira, D.V., “Mechanical Characterization of Stone and Brick Masonry”, Report00-DEC/E-4, Universidade do Minho.Gumaraes,Portugal, .2000.


Mawjoud: Investigation of Handoff Algorithms for GSM Mobile Cellular Networks

69

Investigation of Handoff Algorithms for

GSM Mobile Cellular Networks

S. A. MAWJOUD H. A. Al-TAYYARElectrical Eng. Dept.

University of Mosul

AbstractOne of the main feature of wireless cellular network is to achieve continuous

(uninterrupted) services using handoff when mobile subscribers cross the boundaries ofcells in the coverage area. Handoff calls are usually given higher priority than new callsinitiated.

Various algorithms are investigated using simulation and the results obtainedshow that the received signal strength with hysteresis and threshold in the serving cell(RSS-HTser.) and the received signal strength with hysteresis and threshold in the newcell (RSS-HTnew.) are the two methods which are closely representing the cellular systemenvironment since they contain additional constrains in Handoff (HO) execution.Simulation are carried out by varying the governing parameters including the effects offading on the received signal strength, averaging of signal strength, hysteresis andthreshold, window size on average signal strength, and the standard deviation whichrepresent worsening signal fading.

Keywords: Cellular networks, Signal strength, Hysteresis, Handoff algorithms.

GSM

) ( .

.(RSS-HTser)

(RSS-HTnew) .

.



70

1. Introduction:

The challenging needs in a wireless cellular system is good performance in diverseenvironments. Efficient methods to execute handoff (HO), when mobile stations (MSs) aremoving between adjacent cells or in the intended coverage area, otherwise loading increaseson the network resulting in calls terminations.

Handoff is the process of changing the channel (frequency, time slot, spreading code, orcombination of them) when a mobile station (MS) crosses the boundary of the serving celland enters a neighboring cell. Handoff is classified in broad categories to: hard handoff(HHO) and soft handoff (SHO). HHO in characterized by "break before make", for whichserving channels are released from the serving cell and new channels are allocated from theneighboring cell, SHO is characterized by "make before break" (i.e. two or more cellscommunicate with a MS) before handoff is executed. HHO is implemented in Global Systemfor Mobile Communication (GSM) , while soft handoff is used in Code Division MultipleAccess (CDMA) [1].

In this paper, HHO is investigated.

2. The Mobile Radio Channel:

The mobile channel places fundamental limitation on the performance of wirelesscommunication systems. The radio channels are extremely random and do not offer easyanalysis. Modeling the radio channel is the most difficult part of the radio system design andis characterized in a statistical fashion. The channel experiences two effects[1][3]:

Short-term Fading:This fading is mainly caused by the multipath propagation of radio waves that are

reflected or diffracted on obstacles such as buildings, vehicles etc. The transmitted signalreaches the receives with time delay signals and on paths with different lengths. Dependingon the phase position of the signals of the individual paths, this results in interference of thereceived signal at the receiver.

Fades are more or less at fixed locations in a given environment. The distances offading minima are frequency-dependent at about half a wavelength, thus in GSM 900 atapproximately 15 an and GSM 1800 at approximately 8 cm.

Long-term Fading:This fading is caused by shadowing, such as by buildings, therefore it has a greater

distance of occurring in build-up urban areas. In GSM this type of fading occurs at about 12-60 m.

3. Reasons For Handoff (HO):There are many reasons for HO execution. Usually every MS tries to utilize the radio

channel with the best carrier to interference ratio (CIR) by monitoring the signal strengthusing the MS and the BS of the radio channel. Handoff execution is performed using the BSor the MS by the followings:

Radio Link-type HO:This handoff occurs due to the mobility of the MS in the coverage area and depends on:

Number of MSs in the cell.


71

Number of new calls in the cell.Number of calls transferred from a serving cell to a neighboring cell due tohandoff.Number of terminated calls.Number of calls transferred to adjacent cells.Dwell time in the cell.

Network Management HO:The network executes handoff in case of imbalance in traffic intensity between

neighboring cells and try to achieve the best balance of channels and other resources requiredbetween these cells.

Service-Related HO:It occurs due to degradation in the quality of service (QoS).

4. Problems Encountered With Handoff:Handoff does not always occurs in the correct and successful way due to the limitations

facing this process.

4.1 Ping-Pong Handoff:It occurs when an MS reaches the boundary of the serving cell and entering the

neighboring cell and then returns back to the serving cell [4] [5]. Hysteressis margin (thedifference in signal strength level in the neighboring cell and the serving cell at whichhandoff is initiated), and the average window length are the two basic factors used to avoidping-pong HO to happen. The effect of ping-pong can be reduced by increasing the hysteresislevel to overcome signal fading in shadow region, or using appropriate average windowlength of the signal to overcome the time the MS spend in a shadow region. Certainlimitations must be taken into account:

The value of high hysteresis will limit the ping-pong happening also the numberof handoff executions, but it will increase delay in HO at cells boundaries which isnot accepted practically.The increase in average window length of signal strength causes handoff to beslow and not executed in the appropriate time resulting in lost call.

It is important to known the averaging window (either rectangular or exponential ofvariable weights). The ping-pong cannot be avoided completely even when using appropriatevariables (average signal strength and hysteresis), but can be reduced to acceptable level.

4.2 Number of Handoff:Usually microcell radius is about 1 km radius or less is appropriate in urban areas in

order to increase system capacity [1] [2]. In such a case the mobile station may cross cellboundaries (depending on MS speed) and therefore many handoff's are needed. Each handoffrequires the provision of channels for the MS from the BS the mobile is entering itsboundary. This will add an extra burden on the network [5][6] and cause call dropping orhandoff failure, also increased number of handoff's requires modified handoff algorithms fora required quality of service (GoS).

4.3 Corner Effect on Handoff:Mobile stations moving in microcells experience handoff in line of sight (LOS-HO) and

non line of sight handoff (NLOS-HO). In the latter case the handoff will be difficult. Theproblem of MS's moving in microcells is when there is a sudden change at street corners orstreets junction. A sudden large drop in signal level (20 to 30 dB) occurs. The corner effect is


72

due to the loss of LOS component from the serving BS to the MS, which demands a fasterhandoff to a new BS otherwise call dropping occurs.

5. Types of Handoff [8] [6]:Network controlled handoff (NCHO): In this type of handoff, the network willdecide the handoff by measuring the signal strength at the MS and is performed bythe mobile switching center (MSC). This type is used in the 1st generation advancemobile phone service (AMPS).Mobile Assisted handoff (MAHO): In this method, the MS perform the signalstrength measurements and the mobile switching center (MSC) or the base stationcontroller (BSC) control the handoff.Mobile Controlled Handoff (MCHO): The MS completely control and execute thehandoff. This method is suitable for microcell system. The method is of thehighest degree of decentralization and the benefits of decentralized handoff is thefast decision making. It is used in Digital European Cordless Telephone (DECT).

6. Detection of the Necessity of Handoff:The need of handoff is specified by the measurement of signal strength or the

measurement of carrier to interference ratio (CIR) which is considered as an important valuein a cell and at a certain location. Low value of CIR will force to change the use of thepresent channel between a mobile station and a base station [1]. The mobile radio channel is afading channel and this makes the initiation of handoff decision difficult. This effect maycause many unnecessary handoff's (e.g. ping-pong), therefore the need of effective, adaptiveand fast method for handoff to deal with fast and temporal(i.e. change of radio channel environment) [9]. To deal with such channel environments, theaverage signal strength is adopted in order to reduce the effect of short term signal fading [3].

Figure (1) illustrates the signal variation of the MS from BS1, to BS2. The averagesignal strength received by the MS from BS1 decay as the MS moves further away from BS1,at the same time the average signal strength received by the MS from BS2 increases as theMS moves toward BS2.

Fig. 1 Received signal strength from twoneighboring base stations [Ref. 13].


73

6.1 Received Signal Strength (RSS):Handoff occurs when the signal strength of a neighboring base station received by the

MS exceeds that from the serving base station [1].

6.2 Received Signal Strength with Threshold (RSS-T):Handoff occurs when the signal strength of the serving base station is at a level below

threshold and the signal strength of the neighboring base station is higher than the signalstrength of the serving base station [10].

If the threshold is relatively large as at point T1 in figure 1, the case will be exactlysimilar to the RSS in section (6.1), and handoff will occur at point A in figure 1. If thethreshold is less than that for example at T2 in figure 1, the handoff is delayed until the curveof signal strength intersects the threshold level at point B. If handoff is executed at T3 thedelay will be more so the MS will travel large distance in the neighboring cell and MSconnection is till with the serving base station, which results in degradation in the quality ofconnection with BS1 and results in call failure.

6.3 Received Signal Strength with Hysteresis (RSS-H):Handoff occurs when the signal strength of the neighboring base station is higher than

that of the serving base station by a hysteresis value (h) as in figure 1 at point C. This methodreduces multiple HO (ping-pong HO), and handoff occurs even when the signal strength ofthe serving base station is strong enough to serve the MS in that cell, so RSS-H results inunnecessary handoff. To avoid unnecessary handoff the received signal strength withhysteresis and threshold of the serving base station is used.

6.4 Received Signal Strength with Hysteresis and Threshold of Serving BS (RSS-HTser):In this case handoff to a new cell occurs only when the level of the received signal by

the MS from the serving base station (BS1) decreases to a level lower than the threshold andthe received signal from the new base station BS2 is higher that from BS1 by a certainhysteresis as in figure 1 at point D.

6.5 Received Signal Strength with Hysteresis and Threshold of the New BS (RSS-HTnew):An unintentional handoff to the wrong cell may sometimes occurs. To reduce such

wrong handoff, handoff is delayed until the received signal strength received by the MS froma neighboring station is of enough value, to achieve the threshold of the intended new basestation with the RSS-H algorithm, and this enhances system performance according to thefollowing:

When choosing a correct threshold for the new BS, this will reduce the number ofunnecessary handoffs to the new cell when the cell signal strength is inadequate.When threshold value is high and appropriate, the number of unintended handoff,(wrong cell) is lowered.

7. Handoff Performance Parameters:Handoff is classified into:

Hard Handoff (HHO) is employed in GSM [1].Soft Handoff (SHO) is used in Code Division Multiple Access (CDMA).

In HHO, communication between the serving BS and the MS is disconnected by theserving BS cell before communication between the MS and the neighboring cell BS starts(i.e. communication of the MS is with one BS at a time).


74

BS1

To evaluates the efficiency of handoff operation, it is necessary to define theperformance of the measuring parameters of the HHO method [11] which are:

The number of unnecessary handoff's which occur when the previous connection is ofsatisfactory performance (no need of handoff).The number of unsuccessful handoff that occur in the cell of which the MS receivesinadequate signal strength.Expected number of handoff: is the number of handoff's which occur when the MS istraveling from serving cell to a neighboring cell.Crossover point: is the distance between the MS and the BS of the serving cell atwhich the probability of connection between the MS and the serving BS becomes 0.5,and at the same time the probability of connection of the MS with the neighboring cellBS reaches 0.5. The crossover point is considered to be an important measure of delayin handoff operation.Expected Average Signal Strength (EASS): is the average signal strength of theserving base station and at which the handoff operation is initiated when the MStravel from one cell to another. This parameter is considered as an indication ofdelayed handoff.The effect of window size on the expected average signal strength (EASS).

8. Simulation of Handoff Algorithms:A two cell mode have been considered in [12] [13].Considering a three cells model [14] shown in figure 2.

The model network has three base stations BS1, BS2 and BS3, the distance between anytwo cell centers is D meter with the mobile moving at constant speed along the straight linepath.

The signal strength received by the MS from the three base stations can be written as:dudLogkkda 21 … (1)

dvdDLogkkdb 21 … (2)

dw2D

33d

2DLogkkdc

22

21 … (3)

Where: a(d), b(d) and c(d) are the received signal strength by the MS (in dB) from BS1,BS2 and BS3 respectively.

Fig. 2 Three neighboring cells model

BS2

BS3

BS1D meters


75

D: is the distance between any two cells centers in meters.d: is the distance moved by the MS from BS1 towards BS2 in meters.k1: is the signal strength of BS1 and is taken as 1W (0 dB).k2: represent the path loss and is equivalent to ten times the path loss exponent (n)

which depends on the propagation environment and take a value (3 – 4) in urban area.u(d), v(d) and w(d) represent shadow fading, which follows Log-normal distribution is

represented by zero mean, stationary Gaussian random process and of a standard deviation( ) in dB.

The values of the parameters used in the simulation are assumed as follows:k1 = 0 dB, k2 = 30, D = 2000 m, correlation distance = 20, averaging constant

= 30 and a standard deviation representing shadow fading.The terms u(d), v(d), and w(d) can be generated using white Gaussian noise generator,

then passing it to first degree filter [11].The final equations representing shadow fading terms:

1du20/1expdg10/1du 1 … (4)1dv20/1expdg10/1dv 2 … (5)1dw20/1expdg10/1dw 3 … (6)

Where g1 (d), g2 (d) and g3 (d) are the signals resulting from the noise generators usedfor the serving, new wrong cells (BS1, BS2 and BS3) respectively.

Figure 3 represents the received signal strength from BS1 and BS2 given by equations 1and 2.

Figure 4 represents the received signal strength received from BS3.Which is given by equation 3.

When dealing with random numbers in system's simulation (e.g. generation of randomnumbers of log-normal distribution as path loss, or exponential distribution to representduration of a cell … etc. ) requires executions of simulation for many times (1000 to 2000)which means increasing the number of iterations and taking the average results to reach thestable state.

Fig. 3 Received signal fromBS1, and BS2 with shadow fading.

Fig. 4 Received signal strengthfrom BS3 with shadow fading.


76

Assuming that there is no direct radio waves between the BS and the MS, theprobability distribution of the envelop of the composite signals is a Rayleigh distribution, andits probability density function is given by:

2

2

2r

2 errP r > 0 … (7)

Where r is the envelop of the fading signal and is the standard deviation. Byincreasing the value of indicates the worsening environment (increased fading) as shown infigure 5.

9. Averaging Process of Signal Strength:It is well known that taking the average signal strength is an important operation in

simulation models in various mobile cellular system. The two fundamental indications in theexecution of handoff are the average number of handoff and the delay in handoff [12] andboth are affected by taking the average signal strength.

In the handoff algorithm used in this paper, the average signal strength is calculated inaccordance with the simulation program from the two basic cells in the RSS-HTser model (orfrom the three neighboring cells in the RSS-HTnew model) for one meter distance during themovement of the MS from BS1 location to BS2 location. The results of this model depends onthe distance and not on the MS speed [23].

In the case of practical window, the average signal strength was taken as exponentialwindow representing the process of taking the average signal strength from BS1, BS2 and BS3as follows [11]:

dadfda … (8)dbdfdb … (9)dcdfdc … (10)

dav/dexpdav1df … (11)

Where dc,db,da are the average signal strength from BS1, BS2 and BS3

respectively. f(d) is the impulse response of the filter used in calculating the average signalstrength. dav is the distance constant for the windowing process when taking the averagesignal strength, it is also called the window size of taking the average signal strength, it isvalue is taken to be 30 meter and is considered an appropriate value in the simulation model[11-14].

dxxdadav/xexpdav1da

0

… (12)

dxxdbdav/xexpdav1db

0

… (13)

dxxdcdav/xexpdav1dc

0

… (14)

The final equations representing the average signal strength received by BS1, BS2 andBS3 are:

dadav/1exp11dadav/1expda … (15)dbdav/1exp11dbdav/1expdb … (16)

dcdav/1exp11dcdav/1expdc … (17)


77

Figure 5 represents the average signal strength from BS1, BS2, while figure 6 is theaverage signal strength received for BS3.

Fig. 5 Received signal strength from BS1, and BS2 at differentstandard deviation ( ) representing shadow fading.

Figure 3 and 6, 4 and 7 show the difference between the average signal strength fromBS1, BS2 and BS3 before and after taking the average signal strength also in figures 5 and 8.

Fig. 6 Average received signal strengthfrom BS1 and BS2.

Fig. 7 Average received signal strengthfrom BS3.

Fig. 8 Average received signal strength from BS1 and BS2 at different standard deviation representing shadow fading.


78

10. Simulation Models for Handoff Process:

The methods mentioned in section 6 which are: RSS, RSS-T, RSS-H, RSS-HTser, RSS-HTnew of which the last two [RSS-HTser and RSS-HTnew] are considered in the simulation dueto resembling the mobile cellular environment and contain extra constrains in HO.

10.1 RSS-HTser Model: This model is explained in section 6.4 and the two cells model (figure 2 without BS3) is

used in the simulation. The signal strength received by the MS for a certain location whenmoving from BS1 cell towards BS2 cell is calculated using equations 1 and 2.

The signals are filtered to calculate the average signal strength in order to reduce theeffect of shadow fading on the signal strength. Then comparison is drawn and if the handoffconstrains are met the handoff is executed. The operation is repeated as the MS moves to thenext location (1 meter distance) from BS1 to BS2.

Finally the number of handoff occurrence are calculated from the simulation and thenre-executed for large number of times and calculating the number of handoff to exclude theeffect of the random variables on the results.

10.2 RSS-HTnew Model:This method assures the reduction of unintentional handoff to the wrong cell. Using the

model of three cells instead of two cells model used in RSS-HTnew as in figure 2. In thismodel ping-pong handoff is taken into consideration when HO occurs between BS1 to BS2and BS3 to BS1.

11. Comparisons of the RSS-HTser and RSS-HTnew Models:

The comparative performance of the RSS-HTser and RSS-HTnew models in the handoffprocess some results are presented [11].

Figure 8 represents the variations in average number handoff's versus hysteresis levelfor the two models. The average number of handoff in the RSS-HTnew is less than that in theRSS-HTser. for the same threshold level of – 85 dB.

Figures 9 to 12 are for constant threshold level of – 85 dB for comparison purpose.Figure 9 shows the average crossover point versus hysteresis level in theRSS-HT model is higher (longer time delay of HO) than that of RSS-HTser. Figure 10represents the relation between crossover point with hysteresis level for RSS-HTser and RSS-HTnew, it shows that the values of crossover periods of RSS-HTnew is higher than RSS-HTserbut it is acceptable due to large decrease of the average number of handoff.

Fig. 11 shows comparison between average number of handoff when using exponentialand rectangular average signal window, while Fig. 12 shows comparison between calldropping probability when using exponential and rectangular average signal window (davexplained in section 9) and cell dropping probability when using exponential and rectangularwindow size, which clearly indicate the improved results of the average signal strength usingexponential window. In addition, the rectangular window size deals with the signal strengthwhen taking its average with equal weights, while the exponential window deals with thesignal strength with different weights. It is clear that the received signal strength versus inlevel by large amount due to shadow fading and that is why exponential window is used.


79

12. Conclusions:From the results obtained the following conclusions can be drawn.

- From the methods of detection of necessary handoff, the use of threshold level for thenew cell (RSS-HTnew gives less number of handoff for the same hysteresis level and

Fig. 9 Average number of handoff versushysteresis level

in the RSS-HTser and RSS-HTnew models.

Fig. 10 Crossover point versus hysteresis levelin the RSS-HTser and RSS-HTnew models.

Fig. 11 Comparison between average number ofhandoff when using exponential and rectangular

average signal window.

Fig. 12 Compassion between the probability of lost callwhen using exponential and rectangular average signal

window.


80

of a constant threshold level but it presents an acceptable delay in the execution ofhandoff.

- The hysteresis levels used in the simulation model decrease the average number ofhandoff but on the account of increased delay in handoff.

- The process of taking the average received signal strength from the base stations willdecrease the number of handoff by decreasing the effect of shadow fading on signalstrength.

- When taking the average signal strength, it is clear that the exponential window givesimproved results over rectangular window. Increasing window size results in decreasein the average number of handoff's and decrease in the probability of lost calls(probability of decrease in the received signal level 10 dB below threshold level) andincreased delay (increase in crossover points and decrease in EASS values).

- Increasing the values of standard deviation of the shadow fading (worsening ofpropagation environment will increase the average numbers of handoff and increasethe probability of lost calls also causes handoff to occur at crossover points near theboundary of the cells concerned.

References:[1] D.P. Agrawal and Q.A. Zeng, "Wireless and Mobile Systems".Thomson Books / Cole 2003.[2] T. S. Rappaport, "Wireless Communications", 2nd edition Prentice-Hall, 2002.[3] B. H. Walke, "Mobile Radio Networks", 2nd edition, John Wiley and Sons Ltd, 2002.[4] A. Markopoulos, A. Markopoulos, P. Pissaris, S. Kyriazakos, A. Koutsorodi and Prof. E. D.

Sykas. "Performance Analysis of Cellular Networks by Simulating Location Aided HandoverAlgorithm", Fifth European Conference; Mobile and Wireless Systems Beyond 3G EuropeanWireless. Feb. 2004.

[5] K. Jaswal, "Handoff Issues in a Transmit Diversity System", M. Sc. Thesis, TexasA and M University, Dec. 2003.

[6] A. K. Alhafith, "Radio Resource Management Using Location Estimation", M. Sc. Thesis,University of Mosul, Jan. 2005.

[7] "An Introduction to Handoff in Mobile Cellular Communications", People.Deas.harvard.edu/~jones/cscie129/nu-Lectures/Lecture 7/ Cellular/handoff-html.

[8] R. Sensharma, "Communication Networks Architecture", CDA, Fall 2002.[9] M. Ylianttila, "Vertical Handoff and Mobility System Architecture and Transition Analysis",

University of Oulu, 2005.[10] S. S. C. Rezaei and B. H. Khalaj, "Gray Prediction Based Handoff Algorithm", International

Journal of Information Technology, Vol. 1, No. 3, 2004.[11] H. A. AL-Tayyar, "Study of Handoff Schemes in Cellular Systems", University of Mosul, 2006.[12] N. Zhang and J. M. Holtzman, "Analysis of Handoff Algorithm Using Both Absolute and

Relative Measurement", IEEE Trans. On Vehicular Technology Vol. 45, No. 1, 1996.[13] P. Marichamy, S. Chakabarti and S. L. Maskara, "Overview of Handoff Schemes in Cellular

Mobile Networks and Their Comparative Performance Evaluation", IEEE Proce. VehicularTechnology Conference (VTC99)-Fall, Amsterdam, The Netherlands, 1999.

[14] P. Marichamy, S. Chakrabarti, and S. L. Maskara "Performance Evaluation of Handoff DetectionSchemes". IEE Proc. Vehicular Technology Conference, 2003.


Younis: The Effect Of Drawing Ratio In Deep Drawing Process On Thickness Distribution …..

81

The effect of drawing ratio in deep drawing process on thicknessdistribution along the cup

Dr.A.D.YounisUniversity of Mosul

AbstractIn the present study the effect of drawing ratio in deep drawing process on the thicknessdistribution along the cup (wall, base and nose) has been performed. Obviously, thedrawing ratio is the one of the most important parameter has been adopted to design thedrawing die. Both experimental and numerical models were carried out on variousdrawing ratios (1.484, 1.589, 1.739, 1.908, 2.12 and 2.332).Tthe simulation results showedthat the best drawing ratio is 1.484, which gives small difference between maximum andminimum thickness distribution along the cup. To examine the simulation results,experimental tests were performed one of the drawing ratios which shows the samebehavior and pattern approximately. Keywords:ANSYS9, Deep Drawing Drawing ratio.

..

) , ((Mild Steel) . "

" " :1.484 ,1.589 ,1.739 ,1.908 ,2.12 2.332 . 1.484

. , "

.



82

NOMENCLATURE

mmInstantaneous outside radiusRmmRadius of cuprmmOriginal thickness

otMPaRadial stress

rMPaInstantaneous stress

f

MPaThickness stresst

MPaPrinciple stresses321 ,,

MPaMean instantaneous stressfmi

MPaUltimate stressu

degreeSmall angleDrawing ratio

1-Introduction

Deep drawing is one of the most important processes for sheet metal forming. It is the basefor the mass production of part pieces for many different applications, such as lighter casingsor parts of automobile bodies ….etc. Deep drawing may be defined as: it is a process in whicha blank or work piece is usually controlled by pressure plate, forced into and/or through a dieby means of a punch to form a hollow component in which the thickness is sub stantially thesame as that of the original material [1].It is important to assess the limitations on the drawingratio or reduction that can be accomplished successfully (i.e., without plastic instability) in adesign stage so that minimum number of draws to achieve the required reduction can be used.Limiting drawing ratio (LDR) is defined as the ratio of the largest blank radius that can besuccessfully drawn (i.e., without failure) to the punch radius. In deep drawing process, thelimiting drawing ratio depends on the characteristics of the material, die and punch design andfriction condition [2].The drawing ratio must not exceed a maximum value, in order to prevent cracks at the bottomof the cup.When the friction between the drawn part and the punch is low, then failures will occur in thebase of the part. If the friction between the part and the punch is high, the base of drawn partwill be increasingly stressed with increasing friction in the can body so that the failure zonewill be moved to the body of the drawn can. In order to ensure a safe production process, it ispreferable to select a drawing ratio that is rather modest and less than the maximum possiblevalue [3].

2-Theoretical consideration

The drawing ratio (ß) is an important numerical value for cylindrical draw parts indetermining the required number of drawing steps. The drawing ration is the ratio of thediameter of the initial blank form to the diameter of the drawn part.Figure (1) shows a schematic view of a tool set up for a first draw.Neglecting friction, the equilibrium condition in the radial direction in figure (1) Can be written as [1]:


83

ddrrd rr ot - rdr ot + t2 ot dr 02

sin d ………. (3-1)

Replacing2

sin d By2

d because d angle is small

02

2 otororororor tddrtrdtdrddtrddtdrdtrd

Neglecting products of differential termsdrdr rddt ro drdt to 0ot

0drrddr trr

drrd trr

rdrd trr ……………. (3-2)

The Tresca yield criterionf31 …………………. (3-3)

At the onset of plastic flow. Substitutingr1 And t3 so that

ftr

The Tresca criterion predicts values which on the average are about 10%. Tresca equation,a correction factor is introduced:

ftr 1.1 ……………… (3-4)From eqns. (3-2) & (3-4)

rdrd fr 1.1 ……………. (3-5)

The radial stresses are obtained by integration of eqns. (3-5).r

Rfr r

drdr

1.10

rR

fmir ln1.1 ………………. (3-6)

Where LDR=R/r

By neglecting the friction effect and blank holder force and by approximately from tensiletest:

ufmi 3.1LnLDRur 43.1

: max To calculate the limiting drawing ratio Ultimate tensile strength of the sheet ur max

LnLDRuu 43.1Ln LDR=0.7LDR=2.02 …… this results as the same which calculated by Hosford [4]

Fig. 1 Deep drawing Tool [1]


84

3-Numerical modelThe finite element method has become a powerful tool of numerical solution wide range ofengineering problems. In this method of analysis a complex region defining a continuum isdiscretzed into simple geometric shapes called finite elements. In the finite element method itwill often be possible to improve or refine the approximate solution by spending morecomputational effort. The solution region is considered as built up of many small,interconnected sub regions called finite elements [5].Three type's elements are selected:1-Visco106 to represent the blank.2-TARGET169 to represent the tool.3-CONTACT171 to represent thee contact between blank and tool.

In this work the commercial FEM code (ANSYS 9) are used to simulate the process of deepdrawing operation.Cup forming was created and the numerical results were comparing with the experimentalresults.

4- Experimental workWith the use of the same conditions as those set for the finite element simulations. Theexperiment was carried out using the INSTRON testing machine which has a capacity ofmachine 100 KN the crosshead speed of the testing machine was kept constant at 10 mm/min.A typical cylindrical cup drawing process was chosen for detailed analysis in deep drawingprocess with draw beads. The cup (49.9mm) outer diameter with corner radius .6mmrd and(22mm) height is axisymmetric and the blank from which it is formed has a diameter (82mm),the punch (47.1mm) diameter with corner radius .2mmrp , a thickness of (1mm) mild steel.This cup without flange, and completely drawn into the die shown in figure (2).

Figure (2) the sample of completely drawn cup5- Results and discussions

In this paper the variable study the size of blank and the effect has been explained anddiscussed upon the deep drawing during process, using six different sizes of blank. Therelations between the distance from cup center with the strain, stress and thickness were also


85

discuses, also the load relation with the punch displacement was considered. Comparisons ofnumerical and experimental results were carried out for only one case (one size of blank).Figure (2) shows the completely drawn cup from the experimental process, it can be seenfrom this figure there is no wrinkling and the difference between the maximum and minimumthickness is very small, so that good uniform distribution thicknessAlong the cup has been achieved.Figure (3) represents the relationship between draw force (punch load) and displacement incase of the theoretical and experimental results(which calculated by the INSTRON testmachine directly). It can be seen that both curves have the same pattern.Figure (4) shows the relationship between the thickness distribution and distance from the cupcenter, these curves are similar and has no significant change between theoretical andexperimental results (which calculated by the Tip Micrometer).Figure (5) shows the relationship between thickness and distance from the cup center for allthe drawing ratios considered cases. It can be observed that the best results is founded atdrawing ratio of 1.484, which gives minimum variation between maximum and minimumthickness distribution (approximately 0.05) along the cup, so that this value is less than theLDR which calculated in the theoretical consideration.

Thickness Distribution

-0.2

-0.15

-0.1

-0.05

0

0.05

0.1

0.15

0 10 20 30 40 50 60

Distance from cup center(mm)

Thic

knes

s St

rain

(mm

)

B=1.589B=1.739B=1.908B=2.12B=1.484B=2.332

LOAD CURVE

0

1

2

3

4

5

6

0 5 10 15 20 25 30 35 40 45

DISPLACEMENT (mm)

PUN

CH

LO

AD

(TO

N)

THEO.

EXP.

Fig. (3) Theoretical and experimental load curve Fig. (4) Theoretical and experimental Thickness distribution

Thickness Distribution

-0.08

-0.06

-0.04

-0.02

0

0.02

0.04

0.06

0 10 20 30 40 50


Thic

knes

s St

rain

(mm

)

THEO.EXP.


86

Effective Stress

0

2040

60

80

100120

140

160

0 10 20 30 40 50 60Distance from cup center(mm)

Effe

ctiv

e St

ress

(MPa

) B=1.589B=1.739B=1.908B=2.12B=1.484B=2.332

Effective Strain

-0.020

0.020.040.060.08

0.10.120.140.160.18

0 10 20 30 40 50 60


Stra

in(m

m)

B=1.589B=1.739B=1.908B=2.12B=1.484B=2.332

Figure (6) represents the relationship between the effective strain and distance from the cupcenter, which gives the maximum value at drawing ratio of 1.908. It can be pointed out thatthe difference values of the effective strain is high near the center radius (under punch)because the action equal biaxial tension and will gradually decrease as move away towardsthe edge.

Finally, figure (7) represents the relationship between the effective stress and the distancefrom cup center. In all considered cases, the behavior is observed to be uniform and similarapproximately. Where the effective stress show low value and almost constant under thepunch base, because there is no forming would take place under the punch base. After that theeffective stress increases on the cup wall until reached the maximum value at the end of thecup wall.

References

1. Kurt Lange, Hand Book of metal forming, university of Stuttgart 1985.2. Prakash S., N. Venkata and G.K.L., on multistage deep drawing of axisymmetriccomponents, India institute of technology, May 2003.3. Schuler GmbH, metal forming Hand Book, springer-veriag Berlin Heidelberg 1998.4. William F. Hosford, mechanical behavior of materials, Cambridge University press 2005.5. Rao, the finite element method in engineering, Robert Maxwell, San Diago state university,USA 1982.

Fig. (5) The effect of drawing ratio on the thickness distribution

Fig. (6) The effect of drawing ratio onthe effective strain

Fig. (7) The effect of drawing ratio onthe effective stress


:

. .

1.: ) (

.2.:

.3.:)

( : .A4 .

.35 . .

.

. .(Windows 95) Time

New Roman Arabic Transparant . )16 ( )14 ()12 . (

singlespace .

..4. )3- (15

.5.

.

)3- . (6. .

)3- . (

7. :) ( .

(ABSTRACT) 150 . .

4

. .

.) (

.8. [ ]

: ) (

.9.

. .

10. .11.3.5

.12.i.)30000 (

/ /. .

ii. )3- (

)20000 () ( )2000 (

)4 ( .13 . :

/-http://www.alrafidain.engineering-coll-mosul.comE-mail: alrafidain@ engineering-coll-mosul.com

http://:@www.alrafidain.engineering-coll-mosul.com/

1842010

1.

.,,.

1.

10 .

. ..

2.

22.

3.

33.

.,.

4.

44.

.

5.

51.

.

6.

58.

.,.

.7

80 .. ,.

.8

:

1

...

.

.

)150x160x1000 ()0,0.5,0.75,1.0. (

.)150x300( .

) ( , .

Experimental Study of the Behavior of Fiber Reinforced ConcreteBeams under Repeated Loads

Samier S. ShmasMuna M.Abdullah

Dr.Bayar J. Al –Sulayfani

Dept.civil engineeringDept.civil engineeringDept.civil engineering

Abstract:Concrete behavior subjected to repeated loads is differs than the one subjected to staticloads. Repeated loads caused crushing in some part of concrete due to loading andunloading process. Concrete behavior under static loads is affected by adding steelfibers, which improved many properties especially concrete tensile strength. Theseimprovements are studied in this research under the effect of repeated loads, by testingsimply supported fibrous reinforced concrete beams with dimensions (1000x150x160)mm, reinforced with different percentages of steel fibers (0.0, 0.5, 0.75, and 1.0%).Repeated loads were applied to the beams through two points and for many cycles up tofailure. The physical properties and compressive strength of the concrete used wasfound by casting standard cylinders (150x300) mm for the different percentages offibers. All the results show improvements in beams behavior due to fiber addition to theconcrete under repeated loads, by increasing the deflection, strain, ductility and energydissipation due to increasing of added fibers percentages.

Keywords : Repeated load, Reinforced Concrete, Beams, Steel fiber

2009/4/222009/10/22


2

.

:-

SinhaHsu[1]1964

.1965 Agrawal [2]

SinhaHsu[1] )Bilinear (

.Al-Sulayfani[3])1986 ( ,

. )- ()- ( .

.2001ThammanoonDepongpan[4]

, )Material Deterioration (

, )40 % ()Monotonic Loads. (

2001Hyo-Gyoung KwakSun-Pil Kim[5] )- ()- (

)- ()Menegotto and Pinto 1973(. ,

)- (.

.

)Aspect Ratio ()30-150 .(:

1972LankardSnyder[6].

.1986NakamuraSakai[7]

)Beam Theory ( - .

.

:

3

:

.

:) () (

.

-:

1-: .

2- : .

3-) :( )BS882:1983 ([8].

4- ) :() ()10 ( )BS882:1983 (

[8].5- :

)150x300()75-100 ():::()1:2:4: 0.5.(

6- :)12 ()6 (.7-:%)0,0.5,0.75,1.0 .( )32.(

)1(

:

)25 (.

)0,0.5,0.75,1.0 % (

3000/ .)24 ()28 (.)1: 2 :4 /0.5) ( :

: : ( )23.(


4

)1 (

)2 (

:

5

:

)strain gage (

)4 .(

:1-)150 x 300 (

)1:(

)1 (

)()(0%65.536.5

0.5%70.539.20.75%75.542

)4 (

PP

900 mm

300 mm300 mm300 mm

160

mm

150 mm

)3 (

160

mm


6

1.

:1.

.2. .

) 6 ,7,89 ([9]

)strain capacity(.2-)160x150x1000 (:

)kN(

)5 (

)(

%

36

37

38

39

40

41

42

43

0 0.25 0.5 0.75 1

0.5%

)kN

(

)mm()7(-

)0.0%(

)mm()6(-

)kN(

0.0%

:

7

)6 ,7,89()deflection ( .

.

.)10 (

)80kN(.

)(8–

)kN(

)(mm

)0.75%(

) (9–

)(mm

)kN

(

)1.0%(

(kN

))

(mm

)(11.

%

P=60 kN

P=40 kN

P=80 kN

P=100 kN

Cycle 5

)(m

m

Cycle 6Cycle 5Cycle 4Cycle 3Cycle 2Cycle 1

(% )

)(10)80 kN(


8

)11 (

.

)Ultimate ()Yield( .)12 (

.

.

)) 6 ,7,89()13(

.

:

1.)127.2 %0.5%208.8 %%0.75243 %1(%

2..

3.)103.2 %0.5%104.16 %%0.75106 %1.(%

4..

5.)390 %0.5

%485 %%0.75580 %1(% .

)(12 )%(

)m

m/m

m(

)13 (

(%)

)kN

.mm

(

:

9

:

1. Sinha, B.Y and Hsu, Cheng –Tzu , "Stress-Strain Behavior of Concrete under Cyclicloading ",ACI Materials Journal , Vol. .95, No. 2 , March-April 1998,pp.178-193

2. Agrawal,G.L.,Tulinnd Gerstle,K.H,"Response of Doubly Reinforced Concrete Beamsto Cyclic Loading ",ACI Journal ,Proceeding, Vol. 66,No.9,sep 1969 ,pp.832-835.

3. Al –Sulayfani, Bayar J.," Contribution A L'etude Comportement Des Ossatures EnBeton Arme Sous Solicitations Cycliques Analysis Non-Linear Globale", Docteur DeL' Université De Nantes, Sepecialite Genie civil, No.87-St,1986.

4. Thammanoon Denpongpan, "Effect of Reversed Loading on Shear Behavior ofReinforced Concrete", A Dissertation Submitted To Kochi University of TechnologyIn Partial Fulfillment of Requirements For The Degree of Master of Engineering,January, 2001.

5. Hyo - Gyoung Kwak and Sun – Pil Kim, "Nonlinear Analysis of R.C. Beam Subjectedto Cyclic Loading", Journal of Structural Engineering,vol.127,No.12, Des. 2001,PP.1436-1444.

6. Synder, M. and Lankard, D.R., "Factors Affecting the Flexural Strength of Steel FibersConcrete ", ACI Journal, proceeding, Vol.69, No.2, Feb. 1972,pp.96-100.

7. Sakai, M. and, Nakamura ,N., "Analysis of Flexural Behavior of steel fiber ReinforcedConcrete "proceeding ,RILEM Symposium RC 86 on Development in fiberReinforced cement and concrete,Vol.1, RILEM Technical Committee 49-TFR,July1986 ,pp.27-34.

8. British Standards Institute, B.S:1983"Aggregates from Natural Sources for Concrete".

9. Shah, R., Mishra,S., "Crack and Deformation Charachteristics of SFRC Deep Beams",IE (1) Journal CV ,Vol. 85 ,May ,2004 ,pp.44-48.

–

:

10

. .

) / /(

192

): (

.7284:2:1/)0.45.(

, .

:-

Influence of Aggregate and Cement Source on the CompressiveStrength of Concrete Mixes

Asst. Lecturer. Yaseen Ali Salih

(Civil Eng., Eng. College, Tikrit University)

Abstract

In this research, (192) concrete cubes were taken from different concrete mixesincorporating coarse and fine aggregate from four different sources with two types ofcement; Iraqi cement & Turkish cement. The compressive strength of these cubes wasexamined to know the most suitable materials to product good concrete for use in theprojects of Tikrit University. The results indicate that, the aggregate from Al-Mosulsource gives compressive strength higher than the aggregate from other sources. TheTurkish cement gives compressive strength higher than that given by Iraqi cement.

2009/1/142009/10/6


11

: "

" .

" "

" .

.) () (

) , , () () (.

) ( ".

.

:--:"

"

[1]. ) ( ) .(

,)1 ( . ")5 (1984.

-:"70-75 %

. ) ( ) .(

.[1] .

.) () (

,

[2]. ) , , (

451980.[3])2 (

.)12 (.

:

12

)1 (*..)5(1984

)2/(265251250)(708045

)(4.56.010)/2(16.315.215

%0.92.74 %1.31.151.5

%0.830.88 --------%4.84.08

%02.52.8%4.03.316

%023 -------- %4.42.35

)* //(

)2 () (*

)(

...451980

)1()2( )3(

)4(

9.5100100100100100100100100

4.7595.694.387.593.590-10090-10090-10090-1002.3675.280.390.179.660-9575-10085-10095-1001.1858.365.280.357.530-7055-9075-10090-100

0.60028.245.665.743.515-3435-5960-7980-1000.30012.321.430.114.45-2010-3015-4515-500.1506.44.35.33.80-100-100-100-150.0752.31.23.21.60-50-50-50-5

)* //(

)3 ()*(...45

1980%0.0130.0180.0210.0240.5

%0.6720.7120.7910.942 )* //(

)4 () (*

)(...451980

50.0100100100100100-10037.510096.298.110095-10019.067.355.165.270.135-709.518.626.312.29.910-404.751.42.1000-50.07500000-0

)* //(


13

)5 ()*(...45

1980%0.00510.00430.00580.00650.1%

%0.0730.0610.0840.0965%

)* //(

)1 ()(

)2 ()(

0102030405060708090100

0.010.1110

Perc

ntag

e Pas

sed

By

Wei

ght

Diameter (mm)

Coarse Aggregate from RamadeCoarse Aggregate from MosulCoaese Aggregate from KirkukCoarse Aggregate from Tuz

Perc

ntag

e P

asse

d by

Wei

ght

:

14

:-

)192 ()150×150×150 (

.)3 (

.)67 ( 67283.

4:2:10.45.)52 (1972[3].

)3 (

)()(

)()


15

)7 (

)(

)6 (

)(

S1A1S2A2S3A3S4A4S5A5S6A6S7A7S8A8S9A9S10A10S11A11S12A12S13A13S14A14S15A15S16A16S17A17S18A18S19A19S20A20S21A21S22A22S23A23S24A24S25A25S26A26S27A27S28A28S29A29S30A30S31A31S32A32

: -

. [2] .

. .

)45 ( ) , , , ( .

.

:

16

.)6 .()1984( [1]

.

)4- ( )4- (

)4- ( )4- (

)4 (

0

5

10

15

20

25

30

A1A2A3A4

Samples

Com

pres

sive

Stre

ngth

MPa

Cubic Age 7 Days Cubic Age 28 Days

0

5

10

15

20

25

30

A5A6A7A8

Samples

Com

pres

sive

Stre

ngth

MPa


0

5

10

15

20

25

30

A9A10A11A12

Samples

Com

pres

sive

Stre

ngth

MPa


0

5

10

15

20

25

30

A13A14A15A16

Samples

Com

pres

sive

Stre

ngth

MPa



17

)5- ( )5- (

)5- ( )5- (

)5 (

0

5

10

15

20

25

30

A17A18A19A20

Samples

Com

pres

sive

Stre

ngth

MPa


0

5

10

15

20

25

30

A21A22A23A24

SamplesC

ompr

essi

ve S

treng

th M

Pa


0

5

10

15

20

25

30

A25A26A27A28

SamplesC

ompr

essi

ve S

treng

th M

Pa


0

5

10

15

20

25

30

A29A30A31A32

Samples

Com

pres

sive

Stre

ngth

MPa


)6 (

1. 2. 3. 4.

:

18

0

5

10

15

20

25

30

S1S2S3S4

Samples

Com

pres

sive

Stre

ngth

MPa


)78 ( ) , , ,() () () ( . "

" "

"728.

)7- ( )7- (

)7- ( )7- (

)7 (

0

5

10

15

20

25

30

S5S6S7S8

Samples

Com

pres

sive

Stre

ngth

MPa


0

5

10

15

20

25

30

S9S10S11S12

Samples

Com

pres

sive

Stre

ngth

MPa


0

5

10

15

20

25

30

S13S14S15S16

Samples

Com

pres

sive

Stre

ngth

MPa



19

)8- ( )8- (

)8- ( )8- (

)8 (

)45 ,7 ,8 (.

)9 ()10 (728 ") () .(

)1 (

0

5

10

15

20

25

30

S21S22S23S24

SamplesC

ompr

essi

ve S

treng

th M

Pa


0

5

10

15

20

25

30

S17S18S19S20

Samples

Com

pres

sive

Stre

ngth

MPa


0

5

10

15

20

25

30

S25S26S27S28

Samples

Com

pres

sive

Stre

ngth

MPa


0

5

10

15

20

25

30

S29S30S31S32

Samples

Com

pres

sive

Stre

ngth

MPa


:

20

[1] . " . C3S

C2S " "28[1] .

. )910 ( "

[2]

,.

)9 ( )10 ( 28 7

:-:-

1-

.2-) (

) (.

3-) (

0

4

8

12

16

20

A2,A18A6,A22A10,A26A14,A30

SamplesC

ompr

essi

v St

reng

th M

Pa

0

5

10

15

20

25

30

A2,A18A6,A22A10,A26A14,A30

Samples

Com

pres

sive

Stre

ngth

MPa


21

) , , () (.

4-

4:2:1.

5-) (

:-

)1(- , ,"" 1984 558.

)2(- , ,"

"-2 ,132006 .)3(- , ""

198481.)4(-1985

(5) - Neville A.M.," Properties of Concrete", Pitmum,1978.(6) - Orchard D.F.," Concrete Technology :Properties and Testing of Aggregate ", Vol. 3, Applied Science,1976.

:

22

[email protected]/

.

)1966-1985 ()1986-2005 (Cena et. Al.)5-24 %(

.P/E .

.

. .

30 .% .

)56-67 %( )36-45 .%(.

:P/E.

Affect Mosul Dam Lake of climates in Mousl city before and afterconstruct

Lelian Yaqoob MattiAbstract

The research studies the impact of Mosul Dam Lake on the climate of Mosul city,and comparison of the climate before and after the lake establishment. Monthly valuesof various climatic elements were analyzed for a period of twenty years before theestablishment (1966-1985) and twenty years after it(1986-2005).The analysis shows thatthere is an increase in cooling capacity calculated, according to the equation Cena et. al.after the establishment and at a rate ranging between (5 -24)% in most months but notin summer months and September. It is found that the value of P/E is also increased inthis period compared with that before the establishment. this value increases in wintermonths compared with the other months with low this percentage to reach zero duringthe summer months. In addition, there is an increase in the quantity of rainfall duringwinter months and March which highly affects wheat and barley growth in the region,whereas the comparison between the two periods concerning the amount of rainfall inspring and Autumn months is very simple. Quantity of evaporation increases graduallyafter January and reaches its highest values in June and July then it starts decreasing..There is no difference between the two periods except for quantities of evaporation awhere a noticeable increase in June and July by 30% in the second period after theestablishment. Furthermore, a significant increase in wind acceleration is also observedafter the establishment especially during November, December, January, February andMarch,. this increase is ranged between (56-67%) while during April, May and October,it is ranged (36-45)%.the summer and September, the rate of increase is minorcompared to other months.

2009/6/142009/9/29

mailto:lilian_yaqup2004:@yahoo.com


23

. ..

. )(36º30N,37º55123.5 . 3852300.11,11

33085)2(.

.)450-1500(/ 800 / )-50º (30 %

80 %2,1)3(.223

. .

.

.

.(5)1982)1 (.

)1 (.

)12()Gena et.al.1966 (

0.96)6(

H= (0.412+0.087v) (36.5-t) ………….1

:

24

: -Hm cm-2 sec-1

V m.sec-1

tCº

36.5Cº0.4120.087.Landsberg)1.(

)1 ()13(

H5<Hot10-5Pleasant15-10Cool22-15Cold30-22Very Cold

30>Extreme Cold

)7( . " "

)7(. .

0.52.5/7.6/)8(

)9( .

) ( .

P/E )10 ()1966-1985 (

)1986-2005.(.

)1966-1985 ()1986-2005 (

)2( .

.Genaetal.

P/E.


25

)2 ()1966 -1985 (

)(

(mm)6061.9624113.20001.212.35658(mm)2746871312292352793842441416025

)(m/s0.510.70.720.911.31.421.621.330.990.610.390.41

%817670674328232630456782

Cº1113182531.53843.541.33728.219.512

Cº1.32.65.310.71518.52321.81712.16.72.9

Cº6.157.811.6

517.8523.2

528.2533.2

531.552720.1513.17.45

Cº5.36.19.31316.818.919.519.31713.7106.1

mw/cm2175253331415512601579528443312218171

mb1021102

5101

51010100

8999.8100

01001100

3101310151020

)3 ()1986 -2005 (

)(

(mm)627179.54420.33.100017.15870(mm)2743901322353354023392501436027

)(m/s1.21.61.71.652.11.81.81.661.180.960.991.2

%807570644830283032506385

Cº11.51519273240.244.541.3392819.812.1

Cº1.52.95.810.9161923.222.917.512.673.1

Cº6.58.9512.418.952429.633.8532.128.2520.313.47.6

Cº56.38.912.91818.419.519.816.517105.9

mw/cm2

182.5256365458515620600560489336225168

mb10231029101810151012100199810011007101910201025

:

26

1-)4 (

)5-24 %(

)3.()2 ()1 (

landsberg:

)Cool ( .)Pleasant (.

)Hot (.

)4 ( H

1966-1985

13.8513.5711.799.166.963.881.2442.614.487.6010.4313.00

1986-2005

18.07415.1813.499.7497.433.921.512.494.258.0311.5114.92

%

23.3710.612.666.331.0217.62-4.82

-5.415.359.3812.87

)2 (1 : 2 :

3 :4 :

02468

101214161820 1985-1966 2005-1986

1

2 3 4 3 2


27

0

0.5

1

1.5

2

2.5

m/s

1985-1966 2005-1986

)3 (

2- P/E)5.(

)1-2.5 (

. )4-100 % (

)3. (

. )5 (

. .

)5 (P/E

P/E1966-1985

2.221.350.7130.310.060000.0050.0870.9332.32

P/E1986-2005

2.31.650.8830.330.0860.00930000.120.972.59

%

3.4818.1819256.0630.231000010027.53.8110.42

3-:1-.

:

28

0

10

20

30

40

50

60

70

80

90

mm

1985-1966

2005-1986

)4 (

)13-22( .% .

.

)4 (

)6 (

. )1966-1985(

.

.

)6 (.

)1966-1985(

60ab

61.9a

62a

41d

13.2e

0g

0g

0g

1.2f

12.3e

56c

58bc

)1986-2005(

62c

71b

79.5a

44e

20.3f

3.1g

0h

0h

0h

17.1f

58c

70b

abcdefghp<0.05

2-


29

0

50

100

150

200

250

300

350

400

450

mm

1985-1966 2005-1986

)5 ( .

)30 .%(

)5 (

)7 (

. .

.

.

)7 (

)1966-1985(

27j

46i

87g

131f

229d

235c

279a

384a

244b

141e

60h

25j

)1986-2005(

27j

43i

90g

132f

235d

335b

402a

339b

250c

143e

60h

27j

abcdefghijp<0.05

3-

)3 ()3 ( .

:

30

0

5

10

15

20

25

30

35

40

45

50

C

1985-1966 2005-1986

. )56-67 %( )36-45 .%(

.

35%76 % 100 %80 % .

.

)8 (

.

.

.

)8 (

)1966-1985(

0.51f

0.7ef

0.72def

0.91de

1.3bc

1.42ab

1.62ab

1.33b

0.99cd

0.61ef

0.39f

0.41f

)1986-2005(

1.2e

1.6d

1.7c

1.65cd

2.1a

1.8b

1.8b

1.66cd

1.18e

0.96f

0.99f

1.2e

abcdefp<0.05

4- ) ()9 (.


31

0

10

20

30

4050

60

70

80

90

%

1985-19662005-1986

0

100

200

300

400

500

600

700

mw/c

m2 1985-1966

2005-1986

980985990995

10001005101010151020102510301035

mb

1985-19662005-1986

)6 (

)7 (

)8 (

:

32

)9 (

1- )5-24 %( .

2-P/E.

3-)Cool (

.)Pleasant (.)Hot (.

4- .

5-.

6- ..

7- ) (.

1- " "1988.2- " "1989.3- ""2001 .4- "

"1988.5- "1918-2005.6-).1980 .(

.7-).1983 .(.8-).185.(.9-.).1980 .(.

10-).1973 .(.11-).2001 .(

52.12-Cena, M., Gregorezuk, M. and Wojcik,G. (1966).An attempt of formula determination forcomputation of biometeorological cooling power in Poland, Roezniki Nauk Rolnizych, 119D,pp442-446.13- Landsberg, H.E. (1972). The assessments of human bioclimate ,Alimited review ofphysical parameters, WMO-Geneva- Switzerland, No.331.

–

:

33

.––

99.2%

. ) + (10%

50%10%.

11:1, 10:1, 9:15 .)(2000)1 (.

30% .

)knock (.

The Effect Of Adding Ethanol To leaded Gasoline on ThePerformance of Spark Ignition Engine

Dr. A.R Habbo Mr.H.S HammodiUniversity of Mosul University of Mosul

College of Eng. College of Eng.Mech. Eng. Dept Mech. Eng. Dept

AbstractIn this study the effect of adding pure ethanol (99.2%) to leaded gasoline on the

performance of spark ignition engine have been investigated. All tests were carried outusing ethanol- leaded gasoline blends ( E10, E20, E30, E40 and E50) at variouscompression ratio ( 9:1, 10:1 and 11:1) for different ignition time ( 0º TDC- 30º BTDC),engine speed of 2000 rpm and an equivalence ratio ( 1).

The experimental results showed that blending leaded gasoline with ethanolslightly increased the torque , specific fuel consumption and exhaust gas temperatureand in particular when E30, E40 and E50 blends are used. However, results also showsa significant reduction in exhaust emission for high percentage ethanol-gasoline blend,i.e for E30, E40 and E50. In addition to all that blending gasoline with ethanol allowsengine to operate at high compression ratio without knock occurrence.

Keywords:- Spark ignition engine; Alternative fuel ; Exhaust emission

2008/11/202009/10/8


34

ANNATDCCA°BSFCg/kW hBTDCCA°

COC2H5OH

CA°E0E1090% +10%E2080% +20%E3070% +30%E4060% +40%E5050% +50%LCVkJ/kgMBTN.m

NrpmPbkWSgfTN.mtSec.

TDCCA°TELUHCWOT

= (A/F) st / (A/F) act.(A/F) act / (A/F) st.=

1..

]3,2,1.[

2.

)KNOCK(

)C2H5OH()CO()HC(

]6,5,4.[

3.

:

35

]7.[

4.M.AL-HASAN]1[99%)25%-0%(2.5%

.)8.3%, 9%, 7%, 5.7%(.

)46.5%, 24.3%(7.5%.

5.YÜCESUHÜSEYIN SEDAR]4[)60%, 40%, 20%, 10%, 0%()13:1-8:1(

)5000-2000(.8:111:1

60%11:15000.

.

6.TOLGA TOPGÜL]6[.99.5%

)60%, 40%, 20%, 10%, 0%()10:1, 8:1()10°-36°BTDC(2000

)1.()MBT(.

7.HAKAN BAYRABTAR]7[.

.

8.93%)21%-1.5%(1.5%

)8.25:1, 7.75%()10°BTDC(1500.7.5%

16.5%.

9.YÜCESUHÜSEYIN SEDAR]8[.

)60%, 40%, 20%, 10%(.

)ANN()ARTIFICIAL NEURALNETWORK.(

)MBT(

10.)1.1.()ANN(.

11.FIKRET YÜKSEL, BEDRI YÜKSEL]9[8:1.


36

.)2850-1200(:

)100%, 75%, 50%, 25%.(

60%.

:

)Faryman A30 ( . )11:1-5:1(

)30°BTDC to 10°BTDC( )carburator .()1.(

)1 (

)1 (.

)Multigas mod 488 ( .)CO%vol. ()CO2% vol. (

)UHC ppm vol.()O2% vol. ( . )Lambda (

)/1 .( .

1.25

:

37

)1(

SpecificationItemSpecificationItem

30°BTDC to10°ATDC

Ignition timing (petrol)Faryman A30 marinediesel engine

Engine type

5:1 to 18:1Compression ratio1Number of cylinder

CarburetorFuel system (petrol)95mmCylinder bore

Fuel enjectionFuel system (diesel)82mmStroke

Water-cooldCooling system4-strokeCycle

2 litersEngine oil (sump)582cm3Swept volume

10 litersFuel tanks1000 to 2500 rev/minSpeed range

1 literCooling water reservoir7KwMaximum power

30°BTDC to10°ATDC

Ignition timing (petrol)50 N.mMaximum torque

:99.2%

)E50, E40, E30, E20, E10, E0 ( .

IROX 2000– .

.)2 (.

)2 (Fuel Type

E50E40E30E20E10E00.75830.74690.74590.73780.73180.7282SPGf

1019995898376RON

:9:1

)Micrometer (E0 3000)10-5 (

2000)WOT .(

)0°TDC () ( )=1 .(

16

)Hanhart (

.


38

)CO% vol. ( )HC ppm vol..(

) ()5°BTDC()5°CA (

)30°BTDC( )E0 (

.10%

.

10:1 .11:1 .

:

1.)SPARK TIMING(

)MAXIMUM BRAKE TORQUE MBT(.

2.)4,3,2()9:1,10:1,11:1()E60, E50, E40, E30, E20, E10, E0.()2(

9:120)20°BTDC.(

3.)3(10:1)MBT()E20, E10, E0()15°BTDC(

)E50, E40, E30()20°BTDC()KNOCK()20°BTDC()E20,

E10, E0()E50, E40, E30(

)MBT(

.

4.11:1)4.()4(

)E0()0°TDC(E10

)5°BTDC(.20%)E20()10°BTDC(

)15°BTDC(E30.E50, E40

)20°BTDC(.

5.

)C2H5OH(

:

39

)EXPANSION STROCK(.

)7,6,5 ()BSFC .(

)MBT()Brake power(

:

Pb = )1......(............................................................)(1060

23 kwNT

mf = )2.....(............................../)(360010)(16 3

hkgt

Sgml f

BSFC = )3...(................................................../)(103

kwhgPb

mf

)LCV ()LCV (

)2 (

Ignition Timing (oCA, BTDC)

0 5 10 15 20 25 30 35

Torq

ue (N

m)

18

20

22

24

26

28

30

32

34

36

38

E0 E10 E20 E30 E40 E50

CR=9:1

N=2000 (rpm)

)3 ( Ignition Timing (oCA, BTDC)

0 5 10 15 20 25 30 35

Torq

ue (N

m)

26

28

30

32

34

36

38

40

E0 E10 E20 E30 E40 E50

CR=10:1

N=2000 (rpm)


0 5 10 15 20 25

Torq

ue (N

m)

28

30

32

34

36

38

40

42

E0 E10 E20 E30 E40 E50

CR=11:1

N=2000 (rpm)


40

. .)4 .(

)10,9,8 (

.

)heat loss (

.


0 5 10 15 20 25 30 35

BSF

C (g

/kw

h)

200

220

240

260

280

300

320

340

360

380

E0 E10 E20 E30 E40 E50

CR=9:1

N=2000 (rpm)


0 5 10 15 20 25 30 35

BSF

C (g

/kw

h)

200

220

240

260

280

300

320

340

E0 E10 E20 E30 E40 E50

CR=10:1

N=2000 (rpm)


0 5 10 15 20 25

BSF

C (g

/kw

h)

200

220

240

260

280

300

320

340

E0 E10 E20 E30 E40 E50

CR=11:1

N=2000 (rpm)


0 5 10 15 20 25 30 35

Exha

ust T

emp.

(C

)

600

620

640

660

680

700

720

740

760

780

E0 E10 E20 E30 E40 E50

CR=9:1

N=2000 (rpm)

)9 (Ignition Timing (oCA, BTDC)

0 5 10 15 20 25 30 35

Exha

ust T

emp.

(C

)

600

620

640

660

680

700

720

740

760

780

E0 E10 E20 E30 E40 E50

CR=10:1

N=2000 (rpm)


0 5 10 15 20 25

Exha

ust T

emp.

(C

)

620

640

660

680

700

720

740

E0 E10 E20 E30 E40 E50

CR=11:1

N=2000 (rpm)

:

41

)13,12,11 ()CO (

)MBT (

)Dissociation ( .

)( ) (

E0.

)16,15,14 ( .

E09:1

11:1, 10:1

)Crevices volume (

)11 (


-5 0 5 10 15 20 25 30 35

CO

(%)

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

E0E10E20E30E40E50

CR=9:1

N=2000 (rpm)

)12 (


-5 0 5 10 15 20 25 30 35

CO

(%)

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

0.20

E0E10E20E30E40E50

CR=10:1

N=2000 (rpm)

)13 (

Ignition TIming (oCA, BTDC)

-5 0 5 10 15 20 25

CO

(%)

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16 E0E10E20E30E40E50

CR=11:1

N=2000 (rpm)


-5 0 5 10 15 20 25 30 35

HC

(ppm

)

0

20

40

60

80

100

E0E10E20E30E40E50

CR=9:1

N=2000 (rpm)


42

:

)50% - 10% (

:-1- .2-)BSFC (

.3-

30% .

4- )CO (

)HC (.5-)CO ()HC (

.

:

1. M. Al-Hasan, "Effect of ethanol-unleaded gasoline blends on engine performance andexhaust emission", Transaction of PERGAMON, journal of Energy conversion andmanagement, vol. 44, P.P. 1547-1561, 2003.

2. Maher A.R. Sadiq Al-BAGHDADI, "A simulation model for a single cylinder four-strokespark ignition engine fueled with alternative fuels" Transaction of TUBITAK, Journal, vol.30, PP. 331-350, 2006.

3. E. ZERVAS, X. MONTAGNE, AND J. LAHAYE, "Emissions of regulated pollutants froma spark ignition engine. Influence of fuel and air/fuel equivalence ratio", Journal ofenvironmental science & technology, vol. 37, No. 14, PP. 3232-3238, 2003.

4. Hüseyin Serdar Yücesu, Tolga Topgül, Can Cinar, Melih Okur, "Effect of ethanol-gasolineblends on engine performance and exhaust emissions in different compression ratios",


-5 0 5 10 15 20 25 30 35

HC

(ppm

)

0

20

40

60

80

100

E0E10E20E30E40E50

CR=10:1

N=2000 (rpm)


-5 0 5 10 15 20 25

HC

(ppm

)

0

20

40

60

80 E0E10E20E30E40E50

CR=11:1

N=2000 (rpm)

:

43

Transaction of ELSEVIER, Journal of applied thermal engineering, vol. 26, PP. 2272-2278,2006.

5. M.A. Ceviz, F. Yüksel, "Effects of ethanol-unleaded gasoline blends on cyclic variabilityand emissions in an SI engine", Transaction of ELSEVIER, Journal of applied thermalengineering, vol. 25, PP. 917-925, 2005.

6. Tolga, Topgül, Hüseyin Serdar Yücesu, Can Cinar, Atilla Koca, "The effects of ethanol-unleaded gasoline blends and ignition timing on engine performance and exhaustemissions", Transaction of ELSEVIER, Journal of Renewable energy, vol. 31, PP. 2534-2542, 2006.

7. Hukan Bayraktar, "Experimental and theoretical investigation of using gasoline-ethanolblends in spark ignition engines", Transaction of ELSEVIER, Journal of Renewable energy,vol. 30. PP. 1733-1747, 2005.

8. H. Serdar Yücecu, Adnan Sonzen, Tolga Topgül, Erol Arcaklioglu, "Comparative study ofmathematical and experimental analysis of spark ignition engine performance used ethanol-gasoline blends fuel", Transaction of ELSEVIER, Journal of applied thermal engineering,vol. 27, PP. 358-368, 2007.

9. Fikret Yüksel, Bedri Yüksel, "The use of ethanol-gasoline blend as a fuel in an SI engine",Transaction of ELSVIER, Journal of renewable energy, vol. 29, PP. 1181-1191, 2004.

–

:

44

--

) ()1 (

.

(550)°)12 (

.

"Effect of annealing temperature on the formability of low carbon steel sheet"

Ahmed S.AbdulazizAss.Lecturer

Mosul University-college of engg.-Mech.Dept.

AbstractThis research include a practical study on the effect of pre-annealing in differentconditions and the number of forming steps on the formability of low carbon steel sheetof thickness(10)mm which is widely used due to popular industrial uses especially in thecold pressing processes. A particular specimens were annealed at different temperaturesthen a set of tests were applied on these specimen. It was noticed that the annealingimprove the ductility of the specimens especially the temperature that equal to 550°Cwith (12) minute soaking time which gave the best ability of cold pressing for thespecimen . Also it was found that use of two steps of cold pressing with annealingbetween improve the formability of the specimen as compared with the single step ofcold pressing.

Key words :annealing , steel , forming , cold pressing , ductility.

2008/7/242009/10/1


45

:Introduction

[1][2].

) ((inter pass anneal)

[3]

-.

= k n…………………………………………………………………….……..…..(1)

: : :

k :(strength coefficient)n :[4]

)(0.2-0.5[5].

[1]

)40 (%[2]) ([1].

[6](FEM) [4]

[8]

. [10] (IFSteel)(0.6,0.9,1.2&1.6)

)1.6 ( .[11]

(HCP) (BCC) (FCC)

.) ( )1 (

.

:

46

:Materials And Experimental Work

:

1-:) ()1(

.

2-: (ASTM 370A) (100)

)1 (.

3-:)500°650° ([2] (electric

muffle furnace).)1 (.

)1 :(//°

------As receivedA12550B12600C12650D

4- :(wolpert)

.

5-: (universal brooks inspection equipment)(HRF)

)1/16 (

(60kgf).

6-:

(wolpert)

(1%)

.


47

:Results And Discussion

(Microsoft excel)

:

1-: (A) )1 ()0.3()2 (

) ()A/ (

(B) ) (

.(2)(B)

(B).(3) (B)

(550°C) ) (

[9]

(C,D) .

)2:(ABCD

%26%52%50%42%

050

100150200250300350400

0 0,05 0,1 0,15 0,2 0,25 0,3

(Mpa)

-100

0

100

200

300

400

500

0 0,1 0,2 0,3 0,4 0,5 0,6

(Mpa)

)1 :((A))2:((B)

(MPa)(M

Pa)

:

48

0%

10%

20%

30%

40%

50%

60%

A B C D

0

5

10

15

20

)

DCBA

)3:( )4:(

2-:)3 (

) ( (A)(B)

.

)3:(ABCD95929087

3-:)4 () ((A)

(B)(B).)4 (

(FLD)

)... (

.

)4:(ABCDB1

/12.1615.3414.7314.9018.3


49

4-:) 550°C12min ((B1)

(B)

(B) (B1))20 (%(B1))A (

)50 (% (B)(A)(B1)

.)5 (.

0

5

10

15

20

ABB1

)

)5:(

(A,B,B1) (A)) (

) ((B) (C,D))(B,B1

.

:Conclusions:

:1-.2-)550-650(°.3-)550(°)12()1 (

.4-)3 (

.

)(

:

50

:References:

[1]:.""1986.[2]:R.A.Higgins,"engineerimg metallurgy part I", the English universities press ltd.,1973.[3]:EP4A.IA Mats&Min.Sci-CourseE,"Mechanical behavior of solids".[4]: Taylan Altan,"the importance of n value in sheet metal forming", stampingjournal,oct.2001.[5]: Work hardening from wikipidia,the free encyclopedia/internet/30 april2007..[6]: Yun ling,"uniaxial true stress – strain after necking" Amp journal of technology vol.5June 1996[7]:Z.Marciniak ,"Mechanics of sheet metal forming" Butter worth-Heinemann,oxford,2002. [8]:M.Aghaie-Khafri,"the effect of preheating on the formability of AL-Fe-Si alloy sheet",journal of materials proceeding technology,2005(38-43).[9]:Dr.Ravi Kumar," formability of galvanized interstitial tree steel sheets", journal ofmaterial processing technology,2006(225-237).[10]:R.Narayanasamy,"Evaluation of limiting strains",materialand design,2007(1555-1576).[11]:Donald R.Askeland,"The science and engineering material", Stanley Thornes publishersLtd.,United Kingdom,1998.

–

:

51

/ /

. .

Matlab ,

..

Starting of the Single Phase Induction Motor using StaticSwitched Capacitor

Yasser A. MahmoodAssist. Lecturer / Elect. Eng. / University of Mosul

AbstractThis paper provides a method of changing the effective value of the capacitor for

starting a single phase induction motor starting. This capacitor is connected in serieswith the auxiliary winding. Its effective value can be controlled by static switches inparallel with the capacitor through a series of pulses. The focus was on finding ways ofobtaining the best effective capacitor during starting condition through computersimulation software using (Matlab). By using this method, only one capacitor is used forboth the starting and running conditions and a similar starting performance can beobtained when compared with the conventional method using two capacitors. Thecomputer simulation results are validated by building a laboratory model andcomparison of results.

Key words: single phase induction motor, switched capacitor, variable capacitor,starting conditions.

2009/6/82009/9/15


52

1-. .

]1[ ,

. .)75% (

]2,1[)90 (]4,3.[

, ,

.

,]5,2[.

. . .

]4[ ,]5,3 [

]7,6.[

2-.)Matlab \Power System Blockset v7.6 (

)software packages ( .

..

)1](6,5,3 .[ .)1-b.(

)(a) Back-to-Back thyristor ( ,

.90o180o .

)2 ()3 ()RLC.(

.

ÇáãÝÇÊíÍÇáÓÇßäÉ

R LC

GT Mosfet SC

)1 :(RLC.

(a)(b)(c)

:

53

)(b) MosFet ( .

)1 kHz ()duty cycle ()1%100% ()4,5 .(

. .

. ,)(c) Back to back GTO ( .

)(c) GTO (]6,5[.

)6 .(

..

, ,

. . ,

]3[ .

. , , .

,)7.(

20 40 60 80 100 120 140 160 1800

0.5

1

1.5 x 10-3

)2 :(.

)3 :(.

0 10 20 30 40 50 60 70 80 90 1000

1

2

3

4

5

6 x 10-3

)5 :()duty cycle(

20 F.

)4 :(.

Dutycycle


54

3-.

. , ,

)8.(

)RLC circuit ( ,

]3[,)9.(

Vc

)8 :(.

)7 :(

18 F.

0 20 40 60 80 100 120 1400

0.5

1

1.5

2

2.5 x 10-3

)6 :(.

0

0.5

1

0

1

2

0 20 40 60 80 100 120 140 1600

0.01

0.02

)9 :(

)F(

%

)(

:

55

)10 (

. ..

4-. , .

.)12,11 .(30 F .

)0.85 ()11( ,)50 (

)0.35 ( ,)12 .( , .

.

0

1000

2000

-505

1015

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1-50

0

50

)\(

).

(

)()11 :(30 F0.5

.

0

1000

2000

-10

0

10

20

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1-50

0

50

)12 :(30 F0.550o.

)(

)A

()\

()

.(

)A

(

)10 :(.


56

5-.)13 (

)15o( , , .

, ,.

)14 ()15o (

Matlab)13 .( .

)13 :(15o

-500

0

500

0.25 0.26 0.27 0.28 0.29 0.3 0.31 0.32 0.33 0.34 0.35-5

0

5

-10

0

10

)A(

)A

()

V(

)(

(a)

(b)

(c) )14 :(a (b (c (15o.

300

0

300 -

)V

(

7.5 -

7.5

0

)A(

-4

4-

0

)A(

0.016msec

:

57

6-:

)Motor Starter ()( ,

.)periodically ()synchronously ( .

. .)850 ()350 (

.

7-.

1.T. Wildi. “Electrical Machines, Drives, and Power system” Pearson Prentice Hall, sixthedition, 2006.

2.S. Ghosh " Electrical Machines " Pearson Education, 2005.3.S. Sunter, M. Ozdemir, and B. Gumus, " Modeling and Simulation of a Single-Phase

Induction Motor with Adjustable Switched Capacitor " 9th International conferences onpower electronics and motion control –EPE- PEMC 2000.

4.R. Rabinovici and Z. Keller " New Electronic Starter for Single Phase Induction Motors" IEEE Transaction on Magnetics, vol 32, No 5, Sept 1996.

5.E. Muljadi, Y. Zhao, T. Liu and T. Lipo " Adjustable ac Capacitor for a Single-PhaseInduction Motor " IEEE Transaction on industry application, vol 29, No 3, May 1993.

6.T. Lettenmaier, D. Novotny and T. Lipo " Single-Phase Induction Motor with anElectronically Controlled Capacitor " IEEE Transaction on industry application, vol 27,No 1, Jan/Feb 1991.

8-.

Single phase induction motor:- RLC circuit:-r1m = 0.78 , L1m= 0.0032H, R= 5 , L = 0.6H, C =20µFr2' = 1.6 , L2' = 0.00318H,r1a = 3.52 , L1a = 0.0087H,a = 1.66,

–

:

58

.

/ /

.

:.

Tradition Employing in Contemporary Arabic Architecturethe Architect Rasem Badran " Specialty of "

Luma Abdalwahhab Al-Dabbagh Dr. Asma Hasan Al-DabbaghLecturer / Dept. of Architecture Engineer / Dept. of ArchitectureCollege of Eng. /Univ. of Mosul College of Eng. /Univ. of Mosul

ABSTRACT Many of modern studies attached special importance to the phenomenon ofemploying tradition in Architecture generally, and in Arabic region especially, becauseit considered as a tool for reviving societies, as it has rational and human principleswhich could be employed, therefore it was the problem area for this research, trying torecognize it first, and come out with theoretical framework by scrutiny in previousstudies in this context second, it appears that items of theoretical framework relatedwith conceptual principles which architects believes toward tradition employing,formulations and degree of employing, and finally employing mechanisms . Theresearch appliance for framework items in practical study was aiming to test itscientifically, and it chosen the Architect Rasem Badran to show his specialty.

The findings show that Badarn has fixed conceptual principle in tradition employingwhich was the evolving interpretative one, which related with certain values to the restof theoretical framework items. The conclusions proved the research hypotheses, andconfirm thinkers viewpoints about Badran especially, as he respects and dignifiestradition, therefore architects could apply Badran's manner in their new products, butthe conclusions also rises many questions about preciseness of previous classificationsdealing with conceptual principles for the architects and there belonging to it .

Keywords: Contemporary Arabic Architecture, Tradition Employing, Badran.

2009/6/22009/10/13


59

1.:

)1987211-212(

)2002114 ("

)2001200 (

)2001. (

)2001202 ()2001212 (

)198151. (

.

2.:

)907. (:

- " ")199876 (

)1994 ()198150. (

- )1987 (

)2000158 ( )199811 ( )199111. (

-

)2002111 -113. (:

)

(.

.

:

60

3 .:1.3. " "1981:

)50-51(

)51. (.

2.3.. " "1994

:. :. :.

..

3.3 ." Arab architectural debate on identity " 2008

)46 ()51(

)54. ()25 ()9 (

)7 ()25 (

)20 ()7 (

)25 ()13 (

..


61

4.3. " :... "2006

)4-5(

)5(

" "

) ...6 -7( .

.

5.3. " "1987

)211 ()212(

)216(

)218. (.

6.3. " "1997

)34 ( ..

7.3.. " "2000

)166(

:

62

)160 ()164. (

.

8.3 .. " " :20022007

..

9.3 . " "2004

)484()485()198730 (

)485 ()200213(

)485 ( )199925(

)486 ( )200210 (

)486 ()487. (

.

.

4 . :1.4 .::

, , ,.

: : , ,

) ,1997,34 (


63

) 1987216. (

: :),1994(

" , ,

," " , ,)

,1991 ,52-53(, ,

),2000,158.(

:: , ,

, , ,

)2002112( ,:

.),2000,158 ( , , , ,

,) ,2001,204.( .),2001,125()_ ) (

,2000,154(, ,

),1997,34(. . ,

,),2000,164-165. (

, , , , ,

,.

2.4.::1.2.4 . :

, , ,

),1981 ,51. ( , , , ,

, ,

,)1994( ,)

,1978,220-250(),2000 ,157( . , , ,

, , ),1994(.

:

64

,) ,2001,200-201( ,

, ,) ()(

),2000 ,156. (2.2.4.: ,

,)Badran199347( ,SOM

)200848.(3.2.4 .: , ,

)Badran199347( ,SOM)2008

48. (

)1( / ) :(

// / / / / / /

/ //

/ / / / / / /


65

3.4 .::1.3.4. :

),2000,157( ,),2000,160( ,

)200813.(2.3.4. :

,

)2007104. (4.4 .::

,,.....

...... )

1(.5 .:

: )(.

6 .:

)200732(

)2007( )200732 (:

- / / /1997.- / / /1992.- )( / //1997.- / / .- /.

)1 ( .

)1996107(*

)2()21(.

:

66

)22.(

.1

/ /

2 .

*)Nominal Scales :( ) (

.


67

3 .

4 .

5 .

:

68

)1 ()2 () (

1 .

/ /

* ).199834(.

* ) .199811(

. ) .199813(

.

2 .

*

*)Steele199143-45(

.

3 ..

) .Davidson , Serageldin199585(.

4 .

*

) .200496( *

) .2007(.

5 . :

.)199962(

.5100%

)3 (1

1.1 )-( : )(

)Steele2005133( )(

)) (199834.()context : (

) ,1998 ,11 .( : :

) ,1998 ,12(.


69

)4 (22.1 )-( :

,. * ,

40 ) ,1998 ,13 *() .Steele2005135(.

.

/

)5 (3

3.1 )-( :

: ,

) . ,1998 ,12( . .

/

)6 (4

4.1 )-( : )(

, ,).,1998 ,13( .

:X)Steele2005136( , ,

,.)Steele2005141( ..

:

70

)7 (55.1 )-( :

) : ( , .

, , ). ,1998 ,12( .

.

)8 (1

1.2 , : )(

,.

.

.

.)Steele199145(.)Steele199141(.

.

//

)9 (2

2.2 , :) (

, ,. ,

.)Steele199145(

.

.)Steele199141(

//


71

)10 (3

3.2 , :

) ,(.

..

)Steele199141(.

.

/

)11 ( 44.2 , : )

(.)Steele199141(

.

)12 (11.3 :

, . ,.

. .)Davidson199585()Davidson

199586( ,.)Badran1988159( .

.

:

72

)13 (22.3 :

) ( )()Davidson199589( ) ( , 9x9 ,

)Davidson199588() (.

/

)14 (33.3 :

Davidson,1995,p85)(

,

.)Davidson199592(

.

)15(44.3 :

. .)Davidson199592(

..)Badran1988159(.

//


73

)16 (1)(1.4)( :

) ,2004 ,98( ,

) . ,2004 ,99(.

)17(2)(

2.4)( : )(

, ,)Additative (

)Sub Divative ( , ).,2004,99-100(

. ),2004,99(

.

// /

)18(3)(3.4)( :

.

.

. / /

) . ,2004 ,100.(.

/ /

:

74

)19(1

1.5 :/ ) ,(

) , )– (

, ) ( ) .,1999,62(

.

///

)20 (22.5 :

)/ / ( )–– (

) .1999,62(.

/(

)21(3

3.5 : ) (

: )

( ,.

,).,1999,62(.

/


75

7.:1.7 . ::

100.%2.7 . ::

1.2.7 . ::63 %37 %.

2.2.7 . : :25 %19 %16%13%

9 %6. %3.2.7 . : :

55 %45. %3.7 . ::

1.3.7 . ::58 %37%5 %.

2.3.7 . : :53 %47. %

4.7 : :70 %30 . %

:

76

)5(1522111233324041

)4(1311121231133132

)4(12221112223144

)3(1121111211112132

)3(11221111231222

5701285232450311911711091670.0%0.0%100%37

%0%

63%

25%

19%

6%

9%

6%

13%

16%

0%

9%

55%

45%

58%

37%

5%

53%

47%

70%

30%

)22 (


77

8 .:

)20066-7 ( )2007(

)2001200. (

)200733(

.

)2007()200825 (

"

) "200733. (

9 .:

.

.

10 .:-.-.-.

1.1994 " "–1-31994comegypt.-www.cpas.

2. " " " "1987 .

3. " ".4. " "20015. " " " "–-2004–01.6. " " " ""

1981.

:

78

7. " "1999.8. " " " :

"2002.www.Omranet.com.9. " " " :

2007.www.Omranet.com.10. "

"1542007.11. " "

200212. " "

––200113. " ) ( )

( "23122004.14. " "

1952000.15. "– "

–1419-1998.

16. " " :–2/3 /2001.

17. " "6433 /1998.18. ""

199619. " :"2005

www.ar.qantara.de .20."

"198721. " " Deutschland /– /5/2007.

22. " : " " " " "16-202006.

www.Omranet.com23. " "

2002.24. " " .6

141161991.25./–1420 /– vol.19,No110,1999.26. " "– " "

1997.27. Asfour, Khaled, "Arab Architectural Debate on Identity - Historic Overview", MisrInternational University – Cairo, 2008,WWW.architecture_identity.DE27. Badran , Rasem ," Historical References and Contemporary Design ", " Theories andprinciples of design in the Architecture " , 1988 , The Aga Khan Program for IslamicArchitecture , WWW.ArchNET.org28.Badran , Rasem "A report reflecting personal viewpoint about agha khan 1989 awards"Architectural scientific journal, No.7, 1993.

http://:@www.Omranet.com/


http://:@www.ar.qantara.de/


http://:@WWW.architecture_identity.DE/

http://:@WWW.ArchNET.org/


79

29.Davidson , C.C. , Serageldin ,I. , " Great Mosque and Old City Center Redevelopment " ,"Architecture Beyond Architecture " , 1995 , Aga Khan Award for ArchitectureWWW.ArchNET.org.30. Steele , James , " RECENT WORK BY RASEM BADRAN " , " MIMAR 41 :Architecture in Development " , 1991 , Concept Media / Aga Khan TRUSTFOR Culture, WWW.ArchNET.org .31. Steele , James " The Architecture of Rasem Badran : Narratives on people" , Thames &Hudson, 2005 .

–

http://:@WWW.ArchNET.org./

http://:@WWW.ArchNET.org/

: (*)]1[

80

(*)]1[

.

– –

:

.

.

Climatic control indices in properties of forming the urbanresidential façade

Dr. M.H. Aljwadi H.A.H. Alsangari

Abstract:The basic urban residence pattern changes in Iraq during the previous century

have formed a foundation for a basic problems in its facade ability to provideenvironmental privacy and specially in its role as a physical filter to provide the thermalcontrol inside the house.

This research is trying , through the benefit of analyzing the residential realityproduct ,to criticize the indices of the precedent local studies which try to solve thisproblem and its application difficulties in the residential reality. and offer its realindices to improve the climatic control system through the residential façade, whichcould be implied in the future housing projects, integrated with the other resident needsin façade.

Key words: housing façade, residential needs

__________________________________________ (*) :

2008.

2008/9/252009/9/30


81

:.

) ( .

) ( .

.

: :

.:

.

.

.

.

.

.

.:

.

:1-.2-.3- .4-:5-:6-.7-

.

::- .

--]285 .[

: (*)]1[

82

. :

1 ::

. . .

] .283 -85 [

.2 ::

]283 [ .

.

::-) (

. .

] .34 [

:: .

].41 [

.:

1 :: .

]57 [

. .

)1 ( ) ( .

2 ::

] .310[


83

)1 :(

][

3::

]671 [: : :

].663[

].72 [.

.)3196-197 ()

()7-19 %(.. .

::) (

.)2 (]83-4 [

.

)2 :( .]93 -4[

: (*)]1[

84

::

..

. ::

.)3 (]9

81-84.[

)3:(

]1096[

4 ::] .412 [

.5 ::

.).(

.

].1040 [)1 -2 (

].1041 -43[

::-

. .

.

)4 (

:]115[


85

)4:(

][

1- :.

2- : .

3- : .

4- :.

5- : .

6- : .

.

::

)1 (.

::

: :. :)(

.

: (*)]1[

86

.

)1 (] [

123456/

::-

:

1- .

2-.

3- .

)60 ()10 (% ) (

.] )1[(.

.

::)

( :

1:-:

) ( . .

)47 (.

)2.(


87

)2 :(

82.55%77.92%84.52%78.27%

) (

)5( .

.

)5 :(

2:-:- .

.

3:-:-)1833 (2 .

) ( ) x .()(

)6 ( . ) (

)3 .( ) (

.

: (*)]1[

88

)6 :() (

)3 :() (

)(26.126.827.826.426.78

) ( .) ()7 (

) ( ) () ( .

)7 :() (

) ( )29 (%

. )8.(

.

)(

)(

%


89

) ()21 (% )8.(

)8 :() (

)9 .(

.

4:-:-)

( ) ( .)36.7

(%)9 (%)63.7 (%

)12.2 (% . )37.5 (%

.

.

)9 (

)(

: (*)]1[

90

5:-:-)6.7 (% )10 .(

)10 (%

.6:-:-

)20 (

.

7:-:- .

.

) () (

.

.

. .)11 (

.

)10 : (

)11 : (


91

8:-:-

. )12 (

)13.(

)13 :(

.

)(

) .()37 (%

)28(% .)]28 (% [

) (

. . )29 (% .

.

::-

)12: (

: (*)]1[

92

:

: ) (

) ( . ,

:1-

.2- .

3- .

: :

1- ) (

.

2- .

.)( ,

. ,:

-: .

-: ,

.-:

.

::

, .

,

. ,

.: ,

, ) ( ,

) ( .


93

::-

. :

::1-

.2-) ( ) (.3-

)

.( ::

1-

)

.2-

) (

.3-) (

.

.

:

1- "" , ,2008.

2-Rapoport Amos "House Form and Culture" Printice-Hall INC. EnglewoodCliffs N.J. 1969.

3-." " 1-28

2002-.4-. ""

200230-33 -.

5-.. " "

176-1902001.

: (*)]1[

94

6-""2003.

7-AL-Jawadi &Al-Noah" The Best Window Proportion for Latitude 30-36 N"

Proceeding of the HousingConference Miami Beach Pergamon press 1979 U.S.A.8-."

"131994.9-. "

"8 2200181-100–.

10-.""431989.

11-Malhis, Shatha ."The Multiplicity of Built Form Manifestation: Situa- -ting TheDomestic Form Within Interwooven Syntactic And Symiotic Domains". 4 th.International Space Syntax Symposium ,London , 2003.

)1 ( ** :

-.

-. ------------------------------------------------------------------------------------------

:

19901990-20002000

:

----------------------------------------------------------- -------------------------------:


95

\

\

:

–

: (*)]1[

96

J. Al-Rafidain Engineering Vol.18, No.4 (2010)

Documents

Transcript of J. Al-Rafidain Engineering Vol.18, No.4 (2010)