OPTIMIZATION OF A FLEXIBLE MANUFACTURING SYSTEM (FMS… · Steven, 2005). The next manufacturing...

1

Int. J. Mech. Eng. & Rob. Res. 2012 Prashobh Karunakaran and Mohammad Shahril Osman, 2012

OPTIMIZATION OF A FLEXIBLEMANUFACTURING SYSTEM (FMS) FOR HIGH

TECHNOLOGY PRODUCT PRODUCING FACTORYVIA THE MODULARIZATION OF PRODUCTION

SYSTEMS

Prashobh Karunakaran1* and Mohammad Shahril Osman1

*Corresponding Author: Prashobh Karunakaran,[email protected]

This research is timely because many High Technology Product Producing Factories (HTPPF)are still following the trend of factories that produces the likes of canned food or even carswhere all machines are jointed from start to end (Towill, 2010). These HTPPF have not realizedthat improvements to product must be done everyday to keep market share. And to performresearch continuously, buffers in-between production machines is a necessity. The absence ofthis has caused a number of hard disk manufacturers to close down, especially in Japan andTaiwan (notably Sony, NEC, Matsushita and Trace), which are leaders in joining up productionlines from start to end. It is due to this that currently 90% of hard disks are produced by WesternDigital and Seagate, with Toshiba having 10% market share.

Keywords: Buffer, Modularity, High Technology Product Producing Factory (HTPPF)

INTRODUCTIONHTPPF can best be run with buffers in-between production machines to enableresearch initiatives. The main reason whyresearchers prefer to perform hypothesis teston the production line rather than on adedicated research line is the accepted factin the production of hard disks, for example,

ISSN 2278 – 0149 www.ijmerr.comVol. 1, No. 1, April 2012

© 2012 IJMERR. All Rights Reserved

Int. J. Mech. Eng. & Rob. Res. 2012

1 Faculty of Engineering, University Malaysia Sarawak (UNIMAS), Kota Samarahan 94300, Sarawak, Malaysia.

upon start-up of a line there is a yield drop andthis yield picks up after a while. Therefore ona research line that is not always running, ittakes from an hour to a few hours before theyield stabilizes. This discourages researchersfrom performing tests on the research line. Ina production line where the yield is already ataround 98%, a small change in say a chemical

Research Paper

2


in one particular machine in the line can moreeasily be examined. Also by using productionlines as test beds many lines can be used, sosampling of research can more easily be done.Thus there is a consistent data base to validateresults. Research lines are still available formajor changes especially where mechanicalchanges to machines are involved, but forquick validation of results, the runningproduction line is used (Ting, 2012). Whiledeveloping new strategies for increasingresearch in HTPPF, other major problems wereencountered, one being the bottleneck causedby the Defect Detection System (DDS) andthe other being the wastage in the currentsystem of customer ordering products,factories manufacturing them and how it isfinally delivered to customers. This researcherhad fourteen years of working experience inthe Western Digital hard disk factory inKuching, Malaysia, where research is doneeveryday to improve the Key QualityCharacteristics (KQC) of hard disks.

LITREATURE REVIEWThe review of various types of flexiblemanufacturing was done including how thecurrent factory setup was originally devised.Manufacturing started with CustomManufacturing. In Custom Manufacturing, acarpenter made furniture, a cobbler madeshoes and a tailor made clothes. These peoplehad all the skills required to make the entireproduct. Today, special projects like the ASIMOrobot, spacecraft, special furniture and specialclothing are still custom manufactured. Asdemand grew these skilled craftsmen couldnot meet the increasing demand of the peopleso Intermittent Manufacturing was developed.This is a set of small factories where products

were made in batches like a batch of tenbrooms at a time. The parts for a product aremade at a time. Say ten wooden rods areturned from the raw wood. Then ten bottomsof the brooms are cut to shape from the rawwood. Then the ten bottom of broom are drilledwith holes. Then the holes of the ten bottom ofbroom are stuffed with the plastic brushes tomake the final broom. Continuousmanufacturing was first developed in the FordMotor company. There was a continuousdemand for the Ford Model T cars so aconveyor system was developed whereworkers added parts to the cars at eachstation.

This was the beginning of the humandrudgery of performing repetitive tasks likemachines in factories. Instead of manufacturingsay only ten brooms at a time, a continuousoutput of brooms are manufactured. At eachworkstation a particular task is performed tomake the broom. Workers at each station aretrained to accomplish their task as quickly aspossible. Today most consumable productsare manufactured in this way except that humanare often replaced by robots (Kalpakjian andSteven, 2005). The next manufacturing systemis called Flexible Manufacturing System (FMS)where computer and communicationtechnology is increasingly utilized. In a FMS,products are manufactured in lots but theprocess is continuous and highly automatedand generally controlled by a central computer.FMS is a group of computer-controlledindependent machines linked by materialhandling systems that are able to produce awide variety in types and quantities ofproducts. All the machines are controlled by acentral computer running a Human Machine

3


Interface (HMI) software which normallyincludes a SCADA system. Often there are nohumans even at the HMI computer. ManyHTPPF management mistakenly believe thatis the ultimate trend and work towardautomating everything without buffers, thushampering research initiatives.

Since factories are increasingly beingautomated, details of the latest availablerobots and capabilities were also studied. Theavailable configurations of robot arms areSCARA (four axes robots – SelectionCompliance Assembly Robot Arm), Verticalmulti-joint (six axes robot or articulated),Cartesian coordinates (two axes also knownas Linear Module). Cylindrical coordinates (twolinear axes plus one rotating axes) and the lastbeing Spherical coordinates (one linear axisplus two rotary axes). The Adept Cobra i600is a SCARA robot. It is the first Adept robotwith the amplifier unit built into the robot arm(beyond a heat sink at the back of the robot).Robots can change End-Effectors which arelike changing fingers according to the type orshape of the product it will pick up. The TeachPendant is a critical portion of the robot. TheTeach Pendant is used to move the end-effector to the product being processed andthen this location is stored as a variable (likeX) with the instruction: “here X”. Later to moveto this particular location, the instruction usedis: “move X”. The teaching of the location ofthe product must be extremely accuratebecause sometimes the robot arm have tomove to this particular point to pick up theproduct for 24 h a day and for many monthsafter the initial teaching.

Major developments in factory machineryespecially to prepare for future product

generations are slow due to the huge capitalinvested in existing HTTPF machines and thelarge amount of capital such industriesgenerate. The view that a process that is stillgenerating big income requires no update iswhat kills big companies. The scale at whichsuch industries can be is depicted by themostly Shenzhen based Foxconn which utilizesabout a million workers to create revenue of$94 bn in 2011 (Foxconn Reports Record HighDecember Revenues). With such big moneyand number of employees, the resistance tomajor changes to cater for future generationof products is much greater. A general view isthat big companies with highly educated peoplecannot make major mistakes. A classicexample to counter this is IBM. IBM inventedthe hard disk and was the main hard diskmanufacturer in 1994 when it started to changethe hard disk media from Al-NiP to glass. IBMwas so confident that this is the direction totake that they shut down all their Al-NiP plantswhile the rest of the hard disk industry stuckwith Al-NiP. The reason why glass was thoughtto be superior was that it was better able tohandle head crashes onto the disk while a headcrash onto the Al-NiP will cause a ripple justlike throwing a stone into a calm lake, thereforedestroying lots of data (Zhang et al., 2011).The problems is, there is a huge data base onmetallurgy developed since mankind startedmaking swords, while glass is a relatively newmaterial, therefore while Al-NiP factories wereable to increase data capacity, IBM struggled.The Al-NiP factories also later developedsystems that prevented the heads from evercontacting the disk via various sensors placedwithin the hard disk to detect externalmovements. Another major issue was thatglass can not dissipate heat from disk as it

4


spins below the head, while the AL-NiP diskcan dissipate it via the hub which holds thedisk; heat destroys magnetic data storage.Eventually this inventor of hard disk had to selloff the entire division (Christensen, 2010).

METHODOLOGY ANDRESULTSData was collected of the rate of researchactivities for the Western Digital (WD)factory in Johor, which is fully automatedwithout buffers, versus the WD factory inSarawak which has humans operating thebuffers in-between automated productionmachines. In the current hard disk industry,companies are still bent on full automation(without buffers) neglecting the data depictedby the chart in Figure 1; the exception beingWD and Seagate which in recent years arefollowing the Sarawak factory model, withhuman buffers in-between machines,because automated buffer systems has notyet been designed. It is not that the SarawakWD factory has superior research because

of the human filled buffers but because it hasbuffers stoppable by researchers. The aimof this work is to automate this buffer toenable researchers to easily perform theirtests.

Empirical observations were also made atthe Western Digital factory in Sarawak for thereasons why researchers in HTPPF prefer theactual production lines to test their hypotheses.These are:

• Faster turnover in research—a line ofproduction machines need a whole team oftechnicians to operate and maintain.Researchers will find it hard to justify to thefinancially oriented top management forhaving this team because they could be idlefor long periods in-between researchactivities. The research team technicians willhave lack of work compared to the rest ofthe production staff and this could be frownedupon especially during performance reviewswhich will eventually lead to employeesdeserting the research teams.

Figure 1: A Comparison of Research Initiatives Generated by Two Factories,the Western Digital Factory in Sarawak and Johor, Malaysia

Year

2004 2005 2006 2007 2008

500

400

300

200

100

0

Nu

mb

er

of R

es

ea

rch

Init

iati

ves

Johor

Sarawak

5


• The current process must be stabilizedbefore introducing a change. An empiricallyobserved situation is that unfailingly, a newlystarted production machine will always havemore defects, especially mechanicaldefects and particles than a line which hasbeen running for a few hours. Engineers infactories are still baffled by this empiricalobservation which goes against logicbecause machines which just went throughpreventive maintenance (PM) have newparts and a thorough cleaning of themachine is always done after PM. The mostprobable cause is not the mechanical partsbut the fact that the stirred up chemicals,filters, and other Operation SuppliesRequired Materials (OSRM) in a machinethat has been running for some time is moreconducive for the process than chemicalsthat have been stagnant for a day or more.One of Western Digital’s (formerly Komag)factory in Japan, concluded that even if theinner circumference of the pipes areclogged with coagulated chemicals, theprocess is still stable unless these pipes aredisturbed by the PM which dislodgesparticles from the clogged up chemicals.The PM for the pipes involves pumpingeither hydrogen peroxide (H

2O

2) or nitric

acid (HNO3) (depending on if the process

chemical is acidic or alkaline respectively)in the pipes for a duration of one to six hours(minor or major PM respectively). It is forthis reason that PM on the pipes was notdone for this Japanese factory and insteadall the pipes are changed after a certainspan of time.

• In HTPPFs, most production materials andOSRM like filters, chemicals and liquid

abrasives require constant stirring toprevent settlement or coagulation. This isusually harder and expensive to maintainin a seldom running line, like a researchline. So starting up a research productionline to test out a small change in theproduction process requires a day ofwasted production before the processstabilizes.

• More samples of the test can be carried out.A researcher can test out an idea on sayfive production machines and take anotherfive lines as controls. With more samplesthe conclusion can be more accurate. Theoverall change in the KQC of the resultingproduct due to the incremental researchchange on the product is often very smalland deciphering a conclusion may provedifficult. So the more samples and controlsthere are, the easier the research objectivesare achieved.

Considering all these potential problems ofrunning a newly started production line,researchers tend to prefer performing their teston production lines which are continuouslyrunning everyday for 24 h. Thus in conclusion,buffers between production machines inHTPPF are a necessity because it is easierto get management approval for researchinitiatives, it enables maintenance to be carriedout more easily, throughput improvements ofa particular machine within a long line isenabled and multiple product are more easilyproduced.

Four research problems were worked upon,the first being a system of automated buffersto replace the current manually operatedbuffers. Buffers are necessary in HTPPFs to

6


facilitate researchers in testing out theirhypotheses. In a long production line withoutbuffers, stopping machines has two problems.Firstly approvals will be needed from variouslevels of management because of the impactto production in stopping a big connected lineof machines. Secondly products will bedamaged as they are left idling in chemicalfilled lines for long periods of time. Currentlybuffers in HTPPF are manually operated. Inthe hard disk industry, cassettes of disks aremanually placed onto carts; where humanmade damage to product is rampant (evenwhole cassettes have fallen off the cartespecially during night shifts), and there is notraceability of the product. Traceability isimportant in HTPPF because in case a defectwas detected at the final Defect DetectionSystem (DDS), the disks with defects must besegregated and removed from production. Itis because of these factors that the buffershould be automated.

The first design was an automation of thebuffer in-between HTPPF machines (Figures2 and 3). Specifically the disks are placed inRFID coded cassettes and stacked on threelayered tunable shelves. The reason for threelayers is the standard dimension of floor to roofin a hard disk factory and the size of an off-the-shelve, six-axis factory robot. For this bufferto have more layers the robot arm must becustom built. The robot moves up and downwith the help of a ball screw. Moving robots upand down on a ball screw system has neverbeen done before (Binding, 2010) becausefactory robots have facilities going up to it.Currently most factory robots have a CPU andan amplifier to amplify the signals coming outof it before it goes into the robot arm. But it is

Figure 2: The Design of the Buffer Systemin-Between Machines (Solid Works

Software Sketch) Which will FacilitateResearch Initiatives in HTPPF

Figure 3: Side View of the AutomatedBuffer

7


possible to build-in both the CPU and theamplifier into the robot arm (Kalpakjian, 2005).The facilities required are power wires, signalwires and pneumatic pipes. Wireless poweris a recent advance developed at MIT(Massachusetts Institute of Technology)(Gozalvez, 2007). Basically a normal 50 Hzpower transformer already has wireless powertransfer between two coils separated by apaper. If the frequency is increased to themegahertz range the separation of coils canbe up to many meters. Wireless signals haslong been around as in cell phones but hasnot been implemented for factory robots, mainlydue to concerns of hackers penetrating thefactory robots. Therefore to communicatesignals wirelessly to factory robots, securitysoftware must be sourced for what is already

available or specifically developed. For thepneumatic system within the robot arm, CleanDry Air (CDA) supply can be sent to the robotarm by installing a small pump with a filter atthe robot arm itself.

Once the automated buffer has beendesigned, there is a need to design a betterway to stop the production machines.Researchers need to be able to stopproduction machines upon getting an inspiringidea without causing damage to products.Currently when a researcher stops theproduction line it is quite messy, with productsbeing placed in temporary storage containers(if wet) or in cassettes on carts while theresearcher makes the changes. This currentsystem requires lots of human handling andtherefore damage to the products. So a novel

Figure 4: Basic DPCL Wiring Schematic to Enable Machine Stop (MS). The Bigger BoldLetters Represent Changes

CTC Contactor

N LineCTC

Input # 18, 19, 20Detect MSButton I, II, III

L Line32

30

29

31

Main 100 psi AirValve

Main PowerContactor 120 V

Relay Triggeredby CTC

Ethernet Hub

Motor Controller

Servo Motors

Soap Purge

DI Purge

60ps

60ps100ps

Blow Dryer27

28

24 V power supply to power sensors and air valves,because CTC cannot supply enough amps. Positive

from power supply negative from CTC

8


machine stopping method was designedspecifically to cater to the needs ofresearchers within HTPPF. Figure 4 shows themodifications needed in one factory machineto cater to MS. The designed improvedstopping device called Machine Stop (MS) hasthree modes of stopping. If Button i pressed,the machine will stop with memory, i.e.,stopping just like an EMO (EmergencyMachine Off) does, but machine has memoryof last positions before stopping. This isenabled with the help of the recently availableencoders with battery backup. Thus the robotwill not need to do a “homing” upon startup. IfButton ii pressed, the stopping will occur afterfinishing the current process. Say a machinehas nine processes and the researcher maywant to stop it after the second process to testan idea, so the button ii must be pressed duringthe second process. If Button iii is pressed, allthe existing products from the machine will becleared before stopping.

With an automated buffer and a proper wayto stop machines designed, the bottleneck ofthe production line needs to be resolved. Thebottleneck is always the product DefectDetection System (DDS). So a novel DDSwas designed to cater to hard diskmanufacturing. In the current system, a laserbeam is shot onto the hard disk and defectscan be detected from the reflected beam.Instead of shooting a single beam of laser ontothe hard disk, a thousand beams of differentwavelength can be shot sequentially. Such anarray of laser beams has already beendeveloped by the fiber optictelecommunication industry to enable DenseWavelength Division Multiplexing (DWDM).The reflected laser beam can be collected

using a fiber optic cable with the insulationstripped off. That is, the reflected beam is sentto the fiber optic cable cladding. There littleliterature on this because most researchersadvance the sending of laser beams throughthe core and no one seem is advancing thelaser beam movement via the cladding of fiberoptic cables. Experiments were done to proofthat this is possible. Data in Table 1 showsthat as the angle of incidence of the laser beamto the fiber optic cable is as low as possible,the intensity of the beam at the core at the otherend is greater. There are four reasons for

Table 1: Angle at Which the Laser BeamStrikes the Fiber Optic Cable versus thePercentagebrightness Observed at the

End of the Fiber Optic Cable

3 98%

10 88%

20 78%

30 68%

40 58%

50 48%

60 38%

70 28%

80 18%

90 4%

100 5%

110 6%

120 7%

130 8%

140 9%

150 10%

160 20%

170 30%

177 35%

Angle of Beam to theFiber Optic Cable

Percentage Brightness atEnd of Fiber Optic Cable

9


signature so a proper analysis of the defectcan be made.

2. Because the source signal is a reflectedlaser beam from the hard disk, there willbe scattering so this can trigger more thanone sensor among the array of 1000sensors.

3. It is difficult to align the array of 1000 lasersensors with the reflected beams in afactory environment compared to atelecommunication center where once it isaligned no humans will contact with it and itis placed in sealed temperature controlledenvironment..

4. Using an array of 1000 laser sensors ismuch more expensive than a bare fiberoptic cable. Also changing out changing thefiber optic cable as it gets covered with dustis also cheaper.

This results in the corresponding designas shown in Figure 5 where the bare fiberoptic cable was used to capture defect

sending the reflected beam to a bare fiberoptic cable instead of an array of 1,000 laserbeam sensors:

1. The greatest advantage of using a bare fiberoptic cable to capture the reflected beamis that when the laser beam hits a goodportion of the hard disk, it will not bedispersed much but when it hits a defect, itwill be dispersed greatly. Thereby thecurrent method of using a single sensor (asmall window) to predict what defect causedthe beam to disperse is difficult. It is just likea human looking at a situation through a tinywindow, they cannot see the whole picture.Using a bare fiber optic cable will increasethe size of the window. Since the DDSdesign fires only one beam at a time, uponhitting a defect, this beam will be dispersedto a wide beam that hits the fiber opticcladding and later move into the core.Telecommunication technology already canstudy this beam sequentially; which is justlike moving a camera over the defect

Figure 5: Off-Side View of the Disk Defect Detection System

Fiber Optic Cablewith Black Plastic

Covering

1000 Laser LaserBeam Emitting Array

Angle of Incidenceof All the Beams

are the SameHard Disk

10


signatures from the laser beam reflected fromthe hard disk. This design was developedbecause placing a 1000 laser sensors todetect the reflected beam is much moreexpensive, is harder to align and also a singlereflected beam will trigger more than onesensor. With this new system there will be noneed for the tested disk to be turned for 4.5sbelow the laser beam at 12,000 RPM, oneround could be enough, thereby increasingthroughput by 900%.

With the automated buffer, a proper meansto stop the production machines andthroughput issues resolved, the next step is todesign a method for customers in any country

to order a product and this will bemanufactured immediately by robotssomewhere in the world. The justification forthis is that the cost of manufacturing a productwhich has been paid for is much lower thanthe cost of producing it with predicted demandlevels. Most factories mine data from the likesof Google, Facebook and Twitter (Bonneauet al., 2009) to predict demand levels. Butsystems are currently available, whereby assoon as customers, somewhere in the worlddebits the money from their credit cards,humans and software process this order andgives instructions to other humans in factoriesto produce the product (Srinivasan, 2004;Garofalakis et al., 2009; and Michael et al.,

Assembly of Partsfrom OEM Done in

Dell Factory inPenang

Instantly Intel Factory inPenang Starts

Manufacturing theMicroprocessor for this

Computer

Instantly WesternDigital Factory inKuching Starts

Manufacturing theExtra Hard Disk

Instantly aFactory in

Shenzhen, ChinaManufactures

the Mouse

Router in USAInstantly Gives Orders

to Factories Aroundthe World

Customer inMalaysia Orders a

Dell Computer Onlinewith a Credit Card

Within a WeekFedEX Delivers the

Computer to theCustomer in Malaysia

Figure 6: Example of an Optimal Factory System

Currently data is mined from Google. Twitter and Facebook topredict demand levels. But all these are in-between steps,because despite these forecasting models, up to 200,000 harddisks media are sometimes scrapped at the WD factory inKuching due to incorrect demand predictions. But with thisproposed optimum system, products are manufacturedimmediately after customers debits their credit cards, so thereis no need to forecast demand

11


2010). An improvement to such a system isfor customers in ordering a product with a creditcard initiating factory robots to startmanufacturing the product as depicted in

Figures 6 and 7. Security software must bedeveloped or sourced from experienced anti-virus software developers, the likes of McAfeeto couple with such a system.

JUSTIFICATION FORAUTOMATIONThere is also a need to justify advancingautomation in factories as there are quartersthat claim that such advances will take awayjobs from the workforce. But paradigm shiftsneed to occur to view these automations asinstruments that can liberate factory workersfrom the inhuman drudgery of pick and placejobs. Peering into human history will reveal thatit was never natural for humans to be doingrepetitive jobs for eight to twelve hours eachday. Currently factory workers keep performingtheir jobs for the financial benefits they gain,knowing that the money is the shackle thatkeeps them almost as slaves to their jobs(Wen, 2007). This prevents them from havingthe normal human intuition to help improve theirsurroundings. Their tight work schedule

prevents them from developing skills that canenable them to earn a living in other ways.These jobs should be delegated to machines.At the Foxconn factory based in ShenzhenChina, 14 suicides occurred in 2010, such thatthe factory actually installed nets to catchpeople falling down (A Look Inside theFoxconn Suicide Factory – Telegraph).

Industrialist and job seekers have to cometo terms with the fact that, with today’sinternational market place, customers from arich or poor country will not purchase an inferiorhuman-labor made part, while a superiorquality and cheaper, robot-made part is nextto it. People will only purchase the optimumproduct to suit their finances and needs andnot to satisfy some patriotic ideal. Thuseventually competition will cause the human-labor intensive factories to shutdown. When

Figure 7: Schematic Depicting the System for Remote Ordering of Products

LocalPC

RemotePC

Long Distance Fiber Optic Cable

Remote PC in Another Country Accesses Local PC viaDynDNS or Team Viewer

Local PC Running GE HMI

Relay

PLC

Robot orFactory

12


this happens, the only factory where a humancan get employed will be at the automatedfactories which are producing high qualityproducts at the lowest cost. Thus to increaseemployment, industrialists need to useautomated machines and processes to find aniche in the international market. By slowlybuilding up this niche with higher quality andprecision, they will increase employment.Eventually HTPPF will be populated byresearchers or other highly technical skilledworkers (Zuehike, 2010).

CONCLUSIONThe main hypothesis of this research is thatbuffers in-between production machines willenable HTPPF to innovate and thereby keepmarket share. The proof for this is theempirical observation that researchers in theWD factory in Sarawak prefer to useproduction lines to test their hypotheses. Alsothe chart Figure 1 indicates that the WD,Sarawak factory, which has manuallyoperated buffers in-between machines, hasa greater rate of research compared to theWD, Johor factory where all machines arejointed up. The other data is the empiricalobservation that HTPPF that do not have theunderstanding that buffers are important arelosing market share or have closed downespecially in Japan and Taiwan. Thesefactories automate and join up all machineswithout buffers because it is the trend in allindustries to do so. Top management reasonsout that research lines is all that is needed toperform research, but as empiricallyobserved, the research lines are frequentlyleft unused in HTPPF unless there are majorchanges to the product. But majority of thechanges are small and can best be performed

on the production line. While designingsolutions to these problems, other problemswere observed in HTPPF. First was a needfor a higher throughput defect detection systemfor which a design was developed. The secondwas a justification and feasibility study on theability for customers with credit cards initiatingfactory machines half across the world tomanufacturing the product.

Future Work: Four future works can besuggested the first being a better statisticalsoftware coupled with sensors to enablefactory engineers to more easily trace yielddetractors within HTPPF. The second is asimilar statistical software coupled withsensors to trace machine failure andtherefore change out bearings and so forthonly when they have gone out of specification.This will reduce down time of machines,increase yield and bring down overallproduction cost of the HTPPF. Work shouldalso be done to bring down prices ofultrasonic sensors, which are installed outsidepipes to detect flow rates within it. Currentpropeller and ball flow-rate sensors arebecoming too inefficient and sources ofcontamination in HTPPF. As chemicals usedin HTPPF gets purer to cater to the everincreasing capacity in smaller spaces,required in products (like hard disk),contamination from propeller flow metersbecomes unbearable. Further more, whileinstalling propellers flow sensors, pipes needto be cut which will contribute to thecontamination. Fiber optic sensors canrecently measure most physical propertiesand are non reactive to many chemicals, soresearch should be advanced in this field tobring down cost to affordable ranges.

13


REFERENCES1. “A Look inside the Foxconn Suicide

Factory – Telegraph”, Telegraph.co.uk –Telegraph Online, Daily Telegraph, SundayTelegraph – Telegraph, Web, February 01,2012, <http://www.telegraph.co.uk/finance/china-business/7773011/A-lookinside-the-Foxconn-suicide-factory.html>

2. Binding Paul (2010), Interviewed GeneralManager of Adept Technolo, Inc., Singapore,Telephone Interview, September 18.

3. Bonneau J, Anderson J and Danezis G(2009), Prying Data Out of a SocialNetwork, IEEE, July, Comput. Lab. Univ.of Cambridge, Cambridge, UK.

4. Christensen Clayton M (2010), “TheInnovator’s Dilemma: The RevolutionaryBook that will Change the Way You DoBusiness”, HarperCollins.

5. “Foxconn Reports Record High DecemberRevenues”, ResearchInChina, The VerticalPortal for China Business Intelligence,January 11, 2012, Web, February 04,2012, <http://www.researchinchina.com/Htmls/News/201201/29140.html>

6. Garofalakis Minos, Gehrke Johannes andRastogi Rajeev (2009), “Data StreamManagement: Processing High-SpeedData Streams”, Computer and InformationScience, Springer, Publisher, Mendeley.

7. Gozalvez J (2007), “WiTricity—TheWireless Power Transfer [Mobile Radio]”,IEEE Explore Digital Library, IEEEVehicular Technology Society.

8. Kalpakjian Serope and Steven Schmid(2005), Manufacturing, Engineering &Technology, ISBN 0-1314-8965-8,Prentice Hall.

9. Michael Paul, Gangi Di, Wasko Molly andHooker Robert (2010), “GettingCustomers’ Ideas to Work for You:Learning from Dell How to Succeed withOnline User Innovation Communities”,Open Journal Systems.

10. Srinivasan S (2004), “Role of Trust ine-Business Success”, Department ofComputer Information Systems, College ofBusiness and Public Administration,Information Management & ComputerSecurity, University of Louisville, USA.

11. Ting Ching Ching (2012), “InterviewedSenior Manager for Research in WesternDigital (WD) and Hitachi Global StorageTechnologies (HGST), Kuching, Malaysia”,Telephone Interview, February 2.

12. Towill Denis R (2010), “IndustrialEngineering the Toyota ProductionSystem”, Journal of Management History,Vol. 16, No. 3, pp. 327-345, LogisticsSystems Dynamics Group, CardiffBusiness School, Cardiff, UK.

13. Wen D J (2007), “Northeastern IllinoisUniversity is Located in Chicago, ‘TooMuch Growth, Too Little Development: TheReality Behind China’s EconomicMiracle’”, Palgrave Macmillan Journalso.

14. Zhang Tong, Mathew George, Zhong Haoand Icheloni Rino (2011), Modern HardDisk Drive Systems: Fundamentals andFuture Trends.

15. Zuehike Detlef (2010), “Smart Factory –Towards a Factory of Things, TechnologyInitiative V Kaiserslautern, Germany,German Research Center for ArtificialIntelligence, DFKL. Kaiserslautern”,Annual Reviews in Control, Vol. 34,No. 1, pp. 129-138.

OPTIMIZATION OF A FLEXIBLE MANUFACTURING SYSTEM (FMS… · Steven, 2005). The next manufacturing...

Documents

Transcript of OPTIMIZATION OF A FLEXIBLE MANUFACTURING SYSTEM (FMS… · Steven, 2005). The next manufacturing...