VARIANT COST ESTIMATION BASED ON INFORMATION MANAGEMENT.pdf
Transcript of VARIANT COST ESTIMATION BASED ON INFORMATION MANAGEMENT.pdf
E. ten Brinke; D. Lutters; A.H. Streppel and H.J.J. Kals
VARIANT COST ESTIMATIONBASED ON INFORMATION MANAGEMENT
E. ten Brinke; D. Lutters; A.H. Streppel and H.J.J. Kals
University of Twente; Laboratory of Design, Production and ManagementP.O. Box 217, 7500 AE Enschede, The NetherlandsTel. +31 53 4892532, Fax +31 53 4893631E-Mail: [email protected]
Key Words: variant based cost estimation, information management,product information structure, information retrieval system
ABSTRACTBased on the Manufacturing Engineering Reference Model, a method for variant based cost estimation is
proposed. The product information structure related to this reference model is suitable for variant based costestimation. It defines a product in terms of elements and their relations. The elements have properties likename, type and value. They constitute characteristics of a product and can be used to compare products.
This comparison is applied in variant based cost estimation. Its use in different engineering tasks requirescomparison criteria related to the four cost drivers geometry, material, processes and production planning.Every engineering task can associate a desired percentage of required similarity to each comparison criterionindependently, because the criteria are not equally important.
Based on product-characteristics, the similarity between a new product and a previously manufacturedproduct can be calculated for each valued set of comparison criteria. The products with a sufficiently highpercentage of similarity are used to generate the cost estimate, i.e. the average cost of the similar products.Moreover, the average cost of a single product-characteristic can be indicated. If a new product containsproduct-characteristics that have not been manufactured before, the required information has to be generated inanother way. For instance, a generative cost estimation system can be triggered to calculate a cost estimate forthat characteristic.
It is the aim to consider the cost price of a previously manufactured product in a differentiated way. Inthis way, it can be applied in comparisons for each comparison criterion on different levels of aggregation. Inaddition, methods for rating the accuracy of an estimate will become available. The proposed method hasseveral advantages over current classification and coding systems, the main advantage being a more flexibleway of defining and valuing product characteristics without the use of product classification codes.
1. INTRODUCTIONTraditionally, variant cost estimation is based on classification. Products are classified in product groups
of similar products and they are identified by similar codes. The classification code derived for a new product isused to find similar products, which are used to generate a cost estimate. Several classification and codingsystems have been introduced. They can be classified by their structure: hierarchical monocode structures,chain structures and hybrid structures [1]. Advantages of classification related to variant cost estimation are:quick retrieval of historic data; improvement and consistency in cost estimation, quotation procedures andacquisition of new machine tools; rationalisation of cost estimation [2]. However, the classification systemshave several disadvantages: insufficient retrieval chance; dependency on human interpretation; excessivecoding effort; lack of flexibility, feedback of the performance and integration with other systems; discussionsabout how to classify and who should classify; high investments [3]. Besides, classification is often used asidentification [3].
Recently, information structuring is an important research topic. It is found that a product informationstructure can easily be queried for similar products, without the use of classification and codes. In this way thedisadvantages of classification are eliminated while the advantages still apply.
E. ten Brinke; D. Lutters; A.H. Streppel and H.J.J. Kals
2. PRODUCT INFORMATION STRUCTUREConcurrent engineering has emphasised the importance of interaction and communication between the
diverse engineering tasks. The possibility of communication is based on both the availability and theaccessibility of coherent information [4]. It is desirable to have access to meaningful representations of theexisting information, reflecting the current state of affairs.The access to historical data is also of significantimportance [4]. The Manufacturing Engineering ReferenceModel distinguishes three information structures: the OrderInformation Structure, the Resource Information Structureand the Product Information Structure. This referencemodel is applicable in any manufacturing environment.
A product information structure reflects a product bymeans of elements and their relations. The elements andrelations are part of an aspect system representing aproduct. The aspect systems are referred to as domains, e.g.functional domain, physical product definition domain. Tobe able to create meaningful representations i.e. views ofthe product information, the division between objective andsubjective physical product elements is made (figure 1).
With the fundamental structure depicted in figure 2, acomplete product information structure can be constructed.The attributes represent characteristics of the elements andrelations. In practice, the attributes can be treated aselements. The elements and relations have properties asshown in figure 3. These characteristics and the attributesconstitute product characteristics.
Cost information can be added to the productinformation structure by means of the attributes. The costinformation is split up into the four cost drivers, i.e. costdriving characteristics (figure 4). This way of dealing withcost information enables a differentiated view of the costs.Due to function integration, standardisation andmodularization, costs are only allocated to geometricalelements [5]. The cost for a relation is accounted for on ahigher aggregation level e.g. the cost for connecting twocomponents is accounted for on assembly level (figure 1).The cost of each element can be calculated by means of thecost carriers, i.e. the elements, the cost drivers and the costfunctions. The total cost of a product equals the addition ofthe costs of every element. The cost carriers and costdrivers enable the construction of a cost view. This costview visualises the differentiated cost information. Costcarriers with relative high costs can be detectedimmediately and the cause of these high costs can bededuced easily.
3. VARIANT COST ESTIMATIONThe comparison of products that is required for variant cost estimation has to be based on product
characteristics directly related to costs. The proposed method has to be used by different engineering tasks andit has to be effective, independent of the amount of available information. These considerations make it obviousto relate comparison criteria to the cost drivers.
Assembly
Component
Module
Feature
Face
consistsof
has
1 1
1
1
1 1n
n
n
n
nnn
n
n
n
nn
has
has
has
hasconsistsof
consistsof
consistsof
ObjectSubject
Figure 1; The structure of the physical elements, with thedistinction between the objective and subjective view
element
relation
nn attributes
attributes
Figure 2; The fundamental product structure
Figure 3; Characteristics of elements and relations
element
relation
cost attributes
nn
attributes
prod. planning
process(es)
material
geometry
Figure 4; The cost attributes in the fundamental productstructure
E. ten Brinke; D. Lutters; A.H. Streppel and H.J.J. Kals
The comparison criteria have to be defined by the cost engineer (see table 1). First, a category has to bedefined, e.g. manufacturing feature. For every category, category items have to be defined, e.g. round hole,bending line. For every category item a determinative quantity has to be indicated, e.g. diameter, bendinglength. This determinative quantity should be a direct indication for the costs. With these categories, a ’regular’user can describe a new product. The user has to enter the desired percentage of similarity for every categoryand for every category item and the number of these items on the product has to be entered. The user can alsoadd a value for the determinative quantity.
category percentage ofdesired similarity
category item number of thecategory items
determinativequantity
value ofdeterminativequantity
manufacturing 75 % round hole 10 diameter 5 mmfeature bending line 2 bending length 100 mmmanufacturing 60 % punching 1 diameter -method bending 1 length -
Table 1; Example of product characteristic definitions (white: cost engineer, grey: user)
In case the algorithm is used in anautomatic mode, the product structure of anew product has to be searched for thecategories, category items, the number ofeach category item, the determinativequantity and the value of the determinativequantity. The percentage of desiredsimilarity has to be a default set ofsimilarity percentages, which also could bedifferent for every engineering task.
With the characteristic definitions, thedatabase of historic products can be queried.This search can be seen as reducing thehistoric database with the products that arenot similar in a sufficient way. The first reduction of the historicdatabase can be obtained by searching for products of the sameaggregation level and the same type as the new product (figure 5). Forthe remaining products, the similarity has to be calculated for everycategory (figure 6). When a historic product satisfies all categories, thehistoric product is stored. The similarity between the new product and a historic product is calculated withequation 1. The equation is applied in table 2 for only one category (manufacturing feature).
∑
⋅=
=
N
1i 2
i
1
ij d
v
d
t
N
1P
>>
=
>>
=
>>
=ijii
iijij2
ijii
iijij1
jj
j
v vif ,v
v vif , vd
t tif ,t
t tif , td
nn if,n
nnifn, N :with (1)
Pj:
n:nj:
similarity percentage between the new productand historic product j.number of category items of the new product.number of category items of historic product j.
ti:tij:vi:vij:
number of category item i of the new product.number of category item i of the historic product j.value of the determinative quantity of category i of the new product.value of the determinative quantity of category i of historic product j.
new product historic product 1: 83 % historic product 2: 44 % category itemsn = 2t1 = 1t2 = 3t3 = 0
n1 = 2t11 = 1t21 = 2t31 = 0
n2 = 3t12 = 1t22 = 1t32 = 2
t1 = bending linet2 = round holet3 = rectangular hole
Table 2; A simple example of the similarity calculation
get firsthistoricproduct
start
get firstcategory
calculateequalnesscategory
A
get nextcategoryC
storehistoricproduct
B
get nexthistoricproduct
stop
yesno
yes
yes
no
no
Figure 6; Search loop 2(A: similarity ok, B: last historic product
C: last category)
get firsthistoricproduct
Astore
historicproduct
Bget nexthistoricproduct
stop
yes
no
yes
no
start
Figure 5; Search loop 1(A: equal aggregation level andtype, B: last historic product)
E. ten Brinke; D. Lutters; A.H. Streppel and H.J.J. Kals
The cost of a new product equals the average cost of the historic products that satisfy all the categories ina sufficient manner. The number of matches can be increased by ’modifying’ the historic product. The costs areknown for every product element, so it is possible to add or delete product elements that cause dissimilarity.The average cost of a product element can be added or subtracted from the cost of the historic product. The costfor a relation between an element and the product has to be altered as well. When a product element that causesdissimilarity is not available in the historic products, it should be possible to trigger for instance a generativecost estimation system, by means of a task chain [6], to generate a cost estimate for that product element. Thenumber of found similar products and the percentages of required similarity are an indication of the accuracy ofthe cost estimate.
4. FUTURE RESEARCHThe described algorithms have been implemented in a test application, except for ’modifying’ the historic
products. Optimising the search and similarity calculation algorithms is needed in order to minimise thecomputing time, especially in the case of large databases. Other methods for the determination of the accuracywill be aimed for because the current indication is not completely unambiguous and objective. The use of thealgorithms in automatic mode can be extended and the system has to be tested in practice. It is the intention touse this variant cost estimation method in combination with a generative cost estimation system and a computeraided process planning system for sheet metal [7], both based on the Manufacturing Engineering ReferenceModel. Then the proposed combination of variant and generative cost estimation can be tested.
5. CONCLUSIONSThe proposed algorithms enable variant cost estimation without the use of classification and coding. The
disadvantages of classification and coding are eliminated while the advantages still apply. The proposed variantcost estimation method can make full use of all product data. The method is generic i.e. it can be used for everytype of product and in any kind of production environment. The integration with other system is relatively easybecause the method is based on the product information structure. Moreover, the method is highly configurableand flexible in use.
REFERENCES1. Agarwal, M.; Kamrani, A.K.; Parsaei, H.R.; “An automated coding and classification system with
supporting database for effective design of manufacturing systems”; Journal of Intelligent manufacturing,5, (1994), pp. 235-249.
2. Schuttert, M.A.; “Design of a feature-based coding and classification system”; MSc.-Thesis, University ofTwente, Enschede, The Netherlands, (1995).
3. Vliegen, H.J.W.; “Classification systems in manufacturing, managerial control of process knowledge”;Ph.D.-Thesis, University of Twente, Enschede, The Netherlands, (1993).
4. Lutters D.; Streppel, A.H.; Kals H.J.J.; "The role of Information Structures in design and engineeringprocesses"; Proc. of the third WDK workshop on product structuring, (1997).
5. Weustink, I.F.; Brinke, E. ten; Streppel, A.H.; Kals, H.J.J.; "A generic framework for cost estimation andcost control in product design"; Proc. of the 6th Int. Conf. on Sheet Metal, Vol. II, (1998), pp. 231-242.
6. Lutters, D.; Brinke, E. ten; Streppel, A.H.; Kals, H.J.J.; "Design and manufacturing processes based oninformation management"; Proc. of the 15th Int. Conf. on Production Research, (1999).
7. Streppel, A.H.; Lutters, D.; Wijnker, T.C.; Brinke, E. ten; Kals, H.J.J.; "Process planning for sheet metalparts based on information management"; Proc. of the 15th Int. Conf. on Production Research, (1999).