Cost Estimation in Mechanical Production the Cost Entity Approach Applied to Integrated Product...

ARTICLE IN PRESS

0925-5273/$ - se

doi:10.1016/j.ijp

�Correspondifax: +333 87 34

E-mail addre

Int. J. Production Economics 103 (2006) 17–35

www.elsevier.com/locate/ijpe

Cost estimation in mechanical production: The Cost Entityapproach applied to integrated product engineering

Fehmi H’midaa,�, Patrick Martinb, Franc-ois Vernadata

aLGIPM, University of Metz, Ile Du Saulcy, 57012 Metz Cedex 1, FrancebLGIPM, ENSAM CER Metz, France

Received 28 April 2003; accepted 21 February 2005

Available online 19 August 2005

Abstract

A new approach for product cost estimating in mechanical production is proposed within the framework of

integrated product engineering. The approach introduces the new concept of Cost Entity. It is made necessary due to

the current context of growth of indirect costs, especially in manufacturing. The objective, i.e. establishing a tight link

between technical variables (or manufacturing features) and economic variables (modeled as Cost Entities), requires to

model the reasoning procedure and associated knowledge related to cost estimating. To achieve this, two models, a

Product Model and a Costgrammes Model, are presented and used to represent and capitalize technical knowledge. The

cost estimating reasoning procedure, that takes into account alternative process plans of a product, is modeled and

solved by a constraint satisfaction problem (CSP). The solutions of the problem are ranked by economic satisfaction

order. The case of a Termoz part is used as an illustrative manufacturing example.

r 2005 Published by Elsevier B.V.

Keywords: Cost estimating; Integrated product engineering; Cost entity; Manufacturing features; Constraint satisfaction problems

(CSP)

1. Introduction

For most industrial companies, cost estimationmethods mostly determine the performances oftwo strategic functions: product design and pricing(or quotation).

e front matter r 2005 Published by Elsevier B.V.

e.2005.02.016

ng author. Tel: +333 87 34 69 47;

69 35.

ss: [email protected] (F. H’mida).

It is commonly admitted that product designcan engage up to 70–80% of the total product cost(Asiedu and Gu, 1998). The recent progressachieved in Integrated Engineering such as con-current engineering or integrated design opens anew field for cost estimating during the designstage. The objective of these approaches is to takeinto account manufacturing knowledge as early aspossible in the design stage (Parsaei et al., 1997;Roy et al., 1999).

www.elsevier.com/locate/ijpe

ARTICLE IN PRESS

F. H’mida et al. / Int. J. Production Economics 103 (2006) 17–3518

In a competitive market, the inability of acompany to quickly and adequately satisfy suc-cessful requests for quotation can echo severely onits capacity to survive economically (Veeramaniand Joshi, 1996). Indeed, an underestimated costwill result in losses while an overestimated cost willprevent the company from remaining competitive.

So, there is a strong need expressed by industryto have sound cost estimating solutions, both interms of design and quotation, that can improvethe performance of these two strategic functions(Wierda, 1990).

To face this need, and to replace the analytical-based methods commonly used in manufacturingprocess planning, many companies apply para-metric and analogous cost estimation methods(Duverlie and Castelain, 1999; Layer et al., 2002;Matthews, 1983; Otswald, 1992; Stewart, 1991).These methods are really fast because they areessentially synthetic, i.e. they provide the total costof the product according to some of its character-istics. In design activities, the lack of informationabout the cost structure (e.g. composition) andabout the product production processes (i.e.process plans) does not help the designer to makethe targeted modifications for cost reduction. Inquotation activities, assigning only one cost valueto the product limits the transparent negotiation ofthe cost/delay ratio with the customer.

Another major factor justifying this researchconcerns the growth of indirect activities (i.e.costs). The product cost structure includes gradu-ally a more important part of indirect costs,materialized by the support activities, than before.In addition, the causal relationships between costobjects (products or services) and the resourceconsumptions are difficult to assess. Their trace-ability (i.e. the property that makes explicit in theform of an analysis network the links of the costsfrom accounting recording to their incorporationin products or services) is more difficult to achievewith the traditional approaches for cost estimat-ing. These are based on concepts such as main andsecondary cost centers and data (e.g. hourly rates)issued from cost accounting, the limits of whichhave been analyzed by several authors (see forinstance, Johnson and Kaplan, 1987; Lorino,1991; Bouquin, 1997; Stewart, 1991).

After a detailed study of the cost estimatingproblem in mechanical engineering, we came to theconclusion that two support models are required(H’mida, 2002): a knowledge model and a reason-ing model. Knowledge modeling is carried outusing two manufacturing-oriented models: a Pro-duct Model and a Costgrammes Model. The firstone concerns the product structure on the basis ofits manufacturing features. The CostgrammesModel is based on the Cost Entity (CE) conceptintroduced in this paper. The reasoning model firstconcerns the cost estimating procedure of amanufacturing process (i.e. the sum of the manu-facturing operation costs of its manufacturingfeatures). Then, it concerns the cost estimation ofthe alternate production processes of the product,defined as a constraint satisfaction problem (CSP).

2. Cost estimating

In traditional cost accounting methods, it iscommon to classify costs from two standpoints.First, an economical classification standpointsplits costs into:

1.
Direct costs: may be directly allocated to a costobject, such as a piece of product.
2.
Indirect costs: costs that cannot be directlyallocated to a cost object.
Second, a morphological classification stand-point brakes down costs as

1.
Material costs: occur by consuming materials 2. Labor costs: occur by utilizing human labor
force
3. Overhead costs: occur by consuming cost
elements other than the above two.

In manufacturing, cost estimating is the art ofpredicting what it will cost to make a givenproduct or batch of products (Matthews, 1983;Otswald, 1992). Various techniques exist for costestimating (Stewart, 1991). The manufacturingcost of a part can be estimated using one or moreof four basic methods: intuitive, analogous, para-metric, and analytical (Duverlie and Castelain,

ARTICLE IN PRESS

Cost Entity

Input Object Resources

Driver

Activity

Output Object

Cost

Fig. 1. The Cost Entity representation.

F. H’mida et al. / Int. J. Production Economics 103 (2006) 17–35 19

1999; Layer et al., 2002; Cavalieria and Maccar-rone, 2004). The intuitive method relies on theexperience of the estimator to predict the cost. Theanalogous methods are based on cost extrapola-tion from previous estimates made for similar oranalogous parts, often retrieved using a grouptechnology code. Parametric estimating methodsclassically entails the linking of cost to technicalparameters bound in mathematical relations tobuild cost-estimating models. Finally, the analy-tical methods rely on a cost summation related tothe steps in the product production process; and assuch can only occur late in the engineering process.

Based on these methods, different softwaresystems have been developed to assess manufac-turing costs. A detailed analysis of existingcommercial software tools and major proprietarysystems has been published by DoD (1999). Fromthe academic community, noticeable contributionscan be mentioned such as TIMCES, a manufac-turing cost estimation system that integrates CADand process planning aspects in a unified system,but that only considers predefined process se-quences (Wong et al., 1992), neural network-basedapproaches (Shtub and Zimerman, 1993), a frame-work for estimating manufacturing cost fromgeometric design data (Wei and Egbelu, 2000)and, more recently, FIPER developed at NIST asa hierarchical cost estimation tool in all phases ofthe design stage (Koonce et al., 2003).

Another recent trend, related to the analyticalcost relation methods, concerns the Activity-BasedCosting (ABC) approach (Brimson, 1991; Cooperand Kaplan, 1992). In this approach, products orservices consume activities, and activities consumeresources that generate costs. The problem is toidentify cost drivers for each activity. Elements ofthese methods are more or less applicable atvarious stages of the product life cycle. The ABCmethod has been investigated for cost estimatingand management in design and manufacturing(Ben-Arieh and Qian, 2003; Ozbayrak et al., 2004).

Due to the generalized use of integrated productengineering methods and tools, the use of moresophisticated and integrated production facilities(e.g. machining centers, transfer lines, etc.) and theoperations in networked organizations (such assupply chains or virtual enterprises), the manu-

facturing industry is currently characterized by asignificant growth of indirect costs. Therefore,traditional costing methods per se cannot applyanymore. This is why we propose a novel costestimating technique, based on the CE concept.

3. The Cost Entity concept

Still adhering to the principles of the ABCmethod but augmented by the principles of themethod of Analysis Centers of cost accounting(Stewart, 1991; Garrison and Noreen, 1997), wepropose the concept of ‘‘Cost Entity’’ to profitfrom both methods. In the Analysis Centermethod, the costs of auxiliary sections (e.g.logistics) are assigned to the main sections whichdirectly benefit to the products (e.g. machiningoperations). Then the section costs are added andthe total cost is distributed between the products.To be correctly applied, both methods assumehomogeneity of resources generating costs.A CE can be graphically represented as shown

by Fig. 1. A definition follows and the funda-mental homogeneity validity condition of theconcept is explained.

Definition. A Cost Entity (CE) is a cost aggrega-tion associated with resources consumed by anactivity. The fundamental condition on a CEconcerns the homogeneity of the resources, whichpermits to associate a driver with the CE.

Homogeneous resources are stable and inter-dependent. Stable means that the imputation rateh/X (e.g. h/min, h/L) of each resource does notchange according to the product. Interdependentmeans that the resources are consumed in the sameproportion for one or the other, whatever theproduct that uses them.

ARTICLE IN PRESS


Let Ri ¼ fk : k 2 Kg represent the set of theconsumed resources for the CE (CEi) for therealization of ith activity ðAiÞ. If ½y

ikðxiÞ ¼ xiak� is

equivalent to the quantity of resource k consumedfor the realization of activity Ai of driver xi (whereak is the consumption coefficient of the resource k)and if [Ck ¼ yi

kðxiÞ. Imputation rate] correspondsto the cost of y units of k, the basic equation of themodel will be:

CostCE ¼Xk2Ri

CkðyikðX iÞÞ (1)

Example : resource cost

¼ Driver ðnbÞakðh=nbÞimputation rateðh=hÞ

¼ xiak|{z} imputation rate

¼ yikðxiÞimputation rate|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}

¼ CkðyikðxiÞÞ,

where ak is the consumption coefficient of theresource k and nb the identifier of the cost driver.

3.1. Parent Cost Entity (PCE)

In the absence of the homogeneity resourcecondition, the CE will be considered as a PCE (ormacro-entity). Its decomposition into elementaryCost Entities, in order to respect the homogeneitycondition, is mandatory. The parameters (re-sources, activity, input–output objects) and thecost of the PCE are, respectively, the union of theparameters and the sum of the costs of theelementary CE’s that make it.

CostPCE ¼Xi2N

Xk2Ri

CkðyikðX iÞÞ, (2)

where N is the number of Cost Entities making thePCE.

4. Modeling cost estimating knowledge

The necessary knowledge for the cost estimatingis represented in the form of a Product Model anda so-called Costgrammes Model. The ProductModel is built on the basis of the manufacturing

features. For design, manufacturing and costestimating, this notion presents a federative aspect(Feng et al., 1996; Wei and Egbelu, 2000; Wierda,1991). The Costgrammes Model will be based onthe CE concept.

4.1. The product model

The proposed Product Model uses three no-tions: product, manufacturing feature and opera-tion. Two levels are used (Fig. 2). The first one isabout the geometrical and specific description ofthe product in terms of manufacturing features.The second level expresses the manufacturing andthe cost estimating point of view.Fig. 2 describes the structural links between the

different components of the Product Model. Theselinks are mandatory or optional. Solid linesindicate mandatory links (Product/ManufacturingFeatures/Operations relationships) and dottedlines indicate optional ones. The last ones repre-sent all the machines able to realize a determinedoperation. Each manufacturing feature is de-scribed by internal parameters (dimensions, toler-ances, surface texture, etc.), geometrical inter-feature tolerances (perpendicularity, parallelism,etc.), and finally by topological inter-featurerelations (starts on, opens on, etc.).CostAdvantage, an expert system tool, has been

utilized to structure these data (Cognition, 2000).The representation used in CostAdvantage isbased on the notion of object classes, calledcontexts.

4.2. The costgrammes model

To each activity corresponds a CE that capita-lizes the cost estimating expertise linked to thiselement. The proposed Costgrammes Model al-lows to have a global view of all the Cost Entitiespresent in the company and concerned by theproducts being processed (Fig. 3).The first level of the Costgrammes Model

represents all the necessary operations for therealization of the manufacturing features involvedin the product. This corresponds to the directcosts. The second level presents the hierarchicalrelationship in a CE. It contains a direct/indirect

ARTICLE IN PRESS

Operation (i,j,k): Option k (machine M) of operation j generated by the manufacturing feature i

Product

Manufacturing feature N˚1

Manufacturing feature N˚i

Manufacturing feature N˚n

Operation (1,1,3)

Operation (1,3) Operation (n,1)Operation (1,2) Operation (1,1)

Operation (1,1,2)Operation (1,1,1)

Conceptual viewpoint

Operations

Alternate Machines

Manufacturing and cost

estimation viewpoint

Operation (1,2,1)

Fig. 2. The Product Model.

Tool Change

Cost Entity

Qualityplan

Cost Entity

NC-programing

Cost Entity

Scheduling Cost Entity

Logistics Cost Entity

Release Cost Entity

HandlingCost Entity

Setup Cost Entity

Process planning

Cost Entity

MachiningCost Entity

QualityControle

Cost Entity

HandlingCost Entity

Operation Cost Entity

(1,1)


(1,2)

Manufacturing preparation Production

Machine M1 Machine M2

Level 2

Level 1

…

…


(1,3)


(2,1)


(2,2)

Fig. 3. The Costgrammes Model.


and elementary/PCE relation. So, the distinctionbetween direct/indirect costs looses its usefulnesssince all the costs are directly connected to the

product. The computer tool, used to structurethese data, is again the CostAdvantage expertsystem.

ARTICLE IN PRESS


The ‘‘Operation’’ CE makes the link betweenthe Product Model and the Costgrammes Model(Figs. 2 and 3). The driver, issued from the productcharacteristics and then from the level 1 (Fig. 3),insures, respectively, the causality of thecosts engaged in level 1 (Fig. 3) and in level 2(Fig. 3).

5. Modeling the cost estimating reasoning process

The cost estimating reasoning procedure iscarried out in two subsequent phases. The firstone is relative to the generation of the costs ofthe manufacturing process associated with eachmanufacturing feature of the part or product.The second one is relative to the generation of thecosts of an alternate production process of theproduct.

5.1. Cost estimating of a manufacturing feature

For each type of manufacturing features, thereis a set t of operations potentially necessary for itsrealization. T corresponds to all the operationsthat could be used for the most varied instance ofevery manufacturing feature. The technologicalcriteria (e.g. the quality to be obtained) conditionthe decisions that lead to associate a given processas the effective manufacturing process, so formingthe subset t 2 T .

Fig. 4 shows an example of a potential opera-tion set T relative to the manufacturing feature:Threaded Hole. According to the intrinsic data of

Threaded Hole Center drilling Drilling

Manufacturingfeature Draft operations

Fig. 4. The set T of the potential operations relative

the feature (Quality, Angle b of the chamfer), thesubset t of the effective manufacturing processcannot contain the Boring and/or Counter sinking

operations.So, the expertise linked to the manufacturing of

each type of features has to be transcribed in theform of ‘‘If–Then’’ rules in the knowledge base ofthe expert system. Each operation type includesthe expertise of the associated cost estimatingprocedure. The implementation of these principles,by the means of an expert system such asCostAdvantage, allows to generate, for a definedmanufacturing feature, the estimated costs of themanufacturing process corresponding to the effec-tive manufacturing operations of the feature(H’mida, 2002).

5.2. Production cost estimating: A constraint

satisfaction problem

The cost estimation model that seems the mostsuited for explicit representation of the multiplicityof technical solutions, and to consider the varioustrue dependencies between the design, manufac-turing and production functions, appears to be aCSP. The approach consists in building thereasoning procedure from an identification of therelevant variables, knowledge and constraints.Solving a CSP consists in finding a set of eligiblevalues for all the variables such that all theconstraints are simultaneously verified (Tsang,1993). In our model, a solution takes the formof a production process with an estimated cost,noted CPP.

Boring Counter sinking Tapping

Finishing operations

to the ‘‘Threaded Hole’’ manufacturing feature.

ARTICLE IN PRESS


5.2.1. Variables

The expert system, mentioned in Section 5.1, isused to associate all possible manufacturingoperations with the product brake-down intofeatures. For each operation, the capable alter-native machines are also capitalized. This allows toparameterize every operation by the Xijk notation,i.e. machine option k realizing the operation j

generated by the manufacturing feature i.Two types of variables for the production

process can be established on the basis of thisknowledge. Using similar notations as thoseprovided by Wei and Egbelu (2000), these are:

�
Boolean variables relative to the possible pre-cedence cases between two manufacturingoperations.
X ijki0j0k0 :

1 If the option k0 of the operation j0

generated by the manufacturing feature i0

follows the option k of the operation j

generated by the manufacturing feature i

ðor X i0j0k0 follows X ijkÞ;

0 Otherwise

8>>>>>>>><>>>>>>>>:

�
Integer variables relative to the rank of eachmanufacturing operation.
Uijk :

m If the option k of the operation j

generated by the manufacturing

feature i is executed as the mth

operation in the production process:

8>>><>>>:

A third declared variable is relative to the costof a production process solution. It is
� A real variable defined over a fuzzy domain and
noted CPP.

5.2.2. Knowledge cores

The expertise on cost estimating capitalized atthe level of each operation is used to determine themanufacturing cost (ECf) corresponding to eachassociation of the operation/machine type. Theknowledge on the production system machinesand their arrangement in the workshop with thematerial handling system used allows to identifythe machine preparation cost (ECpr) and the

handling cost (ECm) relative to each variableXijki0j0k0. They are defined as follows:

� ECfijk: M
anufacturing cost of the option k of
the operation j generated by the

manufacturing feature i.

� ECpr

ijki0j0k0:P
reparation cost of the option k0 of the
operation j0 generated by the manu-

facturing feature i0 if it follows the

option k of the operation j generated

by the manufacturing feature i.

0
I f Xijk et Xi0j0k0 are
performed on

the same machine,

C
onstant [f(k)] I f they are not
performed on the

same machine.

I
nterviews with the machine
operator who knows well the existing

relations between the product and the

machine preparation time can bring

to the determination of the adequate

driver. In our case, the hypothesis is

made of assigning a constant

preparation cost according to each

machine k.

� ECmijki0j0k0 :H
andling cost of the option k0 of the
operation j0 generated by the

manufacturing feature i0 if it follows

the option k of the operation

j generated by the manufacturing

feature i.

0
I f Xijk and Xi0j0k0 are
performed on the same

machine,

E
Cmijki0j0k0 I f they are not performed
on the same machine.

5.2.3. Constraints

The cost estimating reasoning procedure of aproduct is build around four types of constraints:appropriate constraints for the model, cost con-straint, manufacturing constraints and productionconstraints. All these constraints contribute to thecost estimation of a product production processsolution (H’mida, 2002).

5.2.3.1. Inherent constraints of the model. Theseconstraints are completely independent of the

ARTICLE IN PRESS


product and of the production context. Theymainly concern the model definition.

Let us consider:

n: number of manufacturing features,ni: number of manufacturing operations of a

feature i,nij: number of options for an operation j of a

feature i,N: total number of possible manufacturing

operations.

Then the two states X 011 and X nþ1;11 are used torespectively represent the state of the product inthe raw material stock and in the finished productstock. An extract of the main inherent constraintsof the model is presented as follows:

Conjunctive mutually exclusive constraintsOnly one operation can follow a given operation:

Xnþ1i0¼1

Xni0

j0¼1

Xni0 j0

k0¼1

X ijki0j0k0 ¼ 1 for i ¼ 0; . . . ; n;

j ¼ 1; . . . ; ni; k ¼ 1; . . . ; nij. ð3Þ

Only one operation can precede a given opera-tion:

Xn

i¼0

Xni

j¼1

Xnij

k¼1

X ijki0j0k0 ¼ 1 for i0 ¼ 1; . . . ; nþ 1;

j0 ¼ 1; . . . ; ni; k0¼ 1; . . . ; ni0j0 ð4Þ

The finished part state ðX nþ111Þ is preceded byonly one operation except ðX 011Þ:

Xn

i¼1

Xni

j¼1

Xnij

k¼1

X ijk nþ1;1;1 ¼ 1 (5)

Disjunctive mutual dependence constraints:An operation can only follow or precede a given

operation:

Xn

i¼1

Xnij

k¼1

Xnþ1i0¼1

Xni0 j0

k0¼1

X ijki0j0k0

aXn

i¼1

Xnij

k¼1

Xnþ1i0¼1

Xni0 j0

k0¼1

X i0j0k0ijk

for j ¼ 1; . . . ; ni;j0 ¼ 1 or 2 or . . . ni0 . ð6Þ

Constraints relative to the enumeration ofoperation order:Operation order in the operation sequence:

Uijk �Ui0j0k0 þ X ijki0j0k0 ¼ 0 (7)

Constraints on the first and last position in theprocess (initialization):

U0;1;0 ¼ 0 Raw material stock; (8)

Unþ1;0;0 ¼ N þ 1 Finished product stock: (9)

5.2.3.2. Cost constraint. The second type ofconstraints concerns the cost constraint. Theproduct design should respect an acceptablemaximal cost, a priori fixed at the beginning ofthe estimating exercise. If this limit is not ex-ceeded, this guarantees the company’s profita-bility and assures competitiveness on the market.This constraint sanctions any production pro-cess solution far-off the economic objectivesand expresses a satisfaction degree for anysolution. An example of such constraint can takethe form given by inequality (10), if we assumethat in this example the cost is only a function ofthe set-up, manufacturing and handling costs andthat only the manufacturing cost depends on thebatch size.

Xnþ1i¼1

Xni

j¼1

Xnij

k¼1

Xn

i0¼1

Xni0

j0¼1

Xni0 j0

k0¼1

ðECpr

ijki0j0k0

þQECfi0j0k0 þ ECm

ijki0j0k0 ÞX ijki0j0k0

hAcceptable maxiCost ð10Þ

(in this case, Q corresponds to the size of themanufacturing batch and is assumed to be a prioriknown to the estimator. The determination of Q isout of the scope of this paper).The first term of constraint (10) corresponds to

the variable CPP previously defined, that is, thecost of a product production process solution.This real variable belonging to a fuzzy domainformalizes the relation existing between the knowl-edge of ECf

ijki0j0k0 , ECpr

ijkij0k0, ECm

ijki0j0k0 and theoperations Xijk. Every production process solutionhas associated with it a cost value CPP with a

ARTICLE IN PRESS


satisfaction degree.

Cpp ¼Xnþ1i¼1

Xni

j¼1

Xnij

k¼1

Xn

i0¼1

Xni0

j0¼1

Xni0 j0

k0¼1

ðECpr

ijki0j0k0

þQECfi0j0k0 þ ECm

ijki0j0k0 ÞX ijki0j0k0 . ð11Þ

Expressions (10) and (11) depend on theproblem at hand and may contain additionalterms to take into account other resourceconsumptions as categorized for instance byLakhal et al. (1999).

In Fig. 5, the function Sestimated cost (CPP)represents the satisfaction degrees associated withthe various values of the Cost CPP.

The expression type of this cost constraintallows to classify all the solutions in decreasingorder of the satisfaction degree. So, all thesolutions are included between a minimal cost(miniCost) and a maximal cost (maxiCost).

5.2.3.3. Manufacturing constraints. The manu-facturing constraints concern the geometrical

Fig. 5. A fuzzy domain associat

Fig. 6. Rules of pose constraints associa

and topological relationships between the manu-facturing features. They describe the precedenceorder to be respect between the operations of thefeatures. They have a direct influence on themachine preparation costs and on the handlingcosts.The analysis of the geometrical and topological

relationships between features is made in our casewith the CostAdvantage system. We define a rulebase that decides the possible precedence relationsto be respected between the manufacturing opera-tions. It sets the constraints to be satisfied in theprocess plan.The rules presented in Fig. 6, associated with the

‘‘parallelism’’ relationship expressed between twofeatures i and i0, impose, according to the capacityof a feature to be a support in the set-up of thepart, a precedence constraint (Sabourin andVilleneuve, 1996). According to Rule 1, theprecedence order to be respected is between thetwo finishing operations ni and ni0 belonging,respectively, to the features i and i0. Rule 2, a

ed with a cost constraint.

ted with the relation ‘‘Parallelism’’.

ARTICLE IN PRESS


constraint on the variable Xij(M)i0j0(M0), means thatthe two finishing operations ni and ni0 are realizedon the same machine (machine M or machine M0).

Fundamentally, the precedence relationshipsbetween operations are conditioned by the preci-sion of the existing geometrical and topologicalrelationships between the features. They aredecided during the detailed design stage.

5.2.3.4. Production constraints. The productionconstraints arise from the consideration of themachine availability during cost estimating (Fig. 7).The following formalization is made in theCostAdvantage expert system. First, the user selectsthe cost estimation context: Planned or Not

Planned. Under the Planned context, each machinecan take two logical textual values: Available,Unavailable. A rule base is associated with eachof these contexts. For each ‘‘Unavailable’’ valuetaken by a machine, the corresponding ruleactivates one or several production constraints inorder to satisfy any process solution.

Fig. 8. An activation rule of the production constraint r

UnavailaMachin

ExXijk112 =Xijk312 =

Production planning ormachine maintenance

Causes Constra

Fig. 7. Examples of prod

In the example of Fig. 8, the rule tests theavailability of the Cu4x machine. If the machine isunavailable, then the production constraint isapplied, any variable Xijki0j0k0 having for optionmachine k or k0 on Cu4x is forced to 0. This avoidsto have in the process plan solution this unavail-able machine and prevents an invalid product costestimation.In the case of a Not Planned context, we

consider the hypothesis of the availability ofmachines without any production constraint.

5.2.4. Problem resolution

The problem resolution consists of two phases.First, comes a filtering phase that consists ineliminating the values of the variables which haveno chance to intervene in a solution. This preventsdoing afterward useless calculations. Then, comesthe solution search phase, that is the combinationsof coherent values.To support the reasoning of cost estimating for

the product production process, we have chosen

elative to the unavailability of the machine Cu4x.

ble es

: 0 0

Alternate machines at different costs

EC112 canceled EC312 canceled

ints Effect

uction constraints.

ARTICLE IN PRESS


the Con’flex environment (Rellier, 1996). This CSPsoftware provides a language for problem expres-sion in terms of variables and constraints, and itsupplies a set of resolution algorithms. Thesealgorithms aim, on one hand, at reducing the sizeof the search space (this operation is calledfiltering) and, on the other hand, at scanning thisspace (in a treelike way) to find the solutions.Resolution refers to the procedure that implementsat the same time the filtering and the solution

search phases.

The interest of this modeling approach (con-straints modeling) is that it is able to deal withvariables of totally different natures: Boolean,numerical or fuzzy variables. The establishedconstraints can make reference to discrete domains(discrete constraints), continuous domains (con-tinuous constraints), discrete and continuousdomains altogether (mixed or hybrid constraints)as well as to fuzzy domains (flexible constraints).Constraint-based modeling accepts a tremendousheterogeneity in the expression of the constraintsand the variables.

6. Illustrative case study

First, the role of the two software packagesused is reminded: CostAdvantage and Con’flex

(Fig. 9).

Reasoning model based on CSP : • Alternative solution search,• Solution ranking by economic satisfaction o• Determination of the [maxiCost, miniCost] in

Knowledge capitalization:• Product Model,• Costgrammes Model.

Reasoning model based on «If-Then» rules: • Manufacturing operation cost generation,• Manufacturing and production constraint ac

Variables

Fig. 9. Role of the two

6.1. Specification of the problem to be solved

6.1.1. ‘‘Termoz’’ part

The part for which the cost is to be estimated iscalled the ‘‘Termoz’’ part. A layout drawing of aninstance of this part is presented in Fig. 10.From the analysis of the ‘‘Termoz’’ part

drawing, the list of its constituting manufacturingfeatures can be established (Table 1, Fig. 11).On the other hand, the analysis of the Termoz

part drawing makes it possible to establish the listof the geometrical relationships between themanufacturing features (Table 2).

6.1.2. Manufacturing workshop

It is assumed that four potential machines canbe used in the available manufacturing facility: a 4-axis milling machining center (Cu4x), a 3-axismilling machining center (Cu3x), a 3-axis turningcenter (Ct3x), and a 2-axis turning center (Ct).Each manufacturing operation realized on onecenter becomes an elementary CE ECf.The machine preparation cost corresponds to

the time interval necessary for the operator to setup the environment (assembly of the fixing devices,setting the machining parameters, etc.) for a newtype of part. The associated CE is denoted ECpr

and we assume that it is constant for everymachine type (Table 3).

rder,terval.

tivation.

CostAdvantage

(expert systemgenerator)

Con'flex

(constraints satisfaction

environment)

Constraints

software systems.

ARTICLE IN PRESS

Table 1

Example of manufacturing features information

Name Type Nominal geometrical

information

Micro-geometrical

information

Topological information

F1 Circular_Surface Ø 104 (X�Y) Flatness 0.05 A2 and R support

Ra 1.6

F2 Rectangular_Surface 52 (X)� 42 (Z) Ra 1.6 TT support

Fig. 11. The manufacturing features of the Termoz part.

Fig. 10. The Termoz part.


ARTICLE IN PRESS


In terms of material handling, it is assumed thatan electric cart with a driver and stocking contain-ers are used. The set (cart, driver) represents ahomogeneous CE, named Handling Cost Entityand denoted ECm. The handling costs, issued fromthe machine arrangement in the workshop, aregiven by Table 4.

Table 2

Example of geometrical relations between the manufacturing

features

Type Feature 1 Feature 2 Name Value

Perpendicular F1 F2 reference R? (F2,F1) 0.05

Coaxial A1 A2 reference RY(A1, A2) 0.05

Table 3

Preparation machine costs

Machine Preparation cost in Euro (indicative values)

Cu4x 7

Cu3x 6.3

Ct3x 5.5

Ct 3.8

Fig. 12. The Product Model

The choice of these types of machines andhandling systems allows to discard a large numberof scenarios frequently used in manufacturingworkshops.

6.2. Product Design—Costgrammes Models

6.2.1. Product Model

The Product Model is elaborated from all themanufacturing features in use in the company.Fig. 12 illustrates a Product Model specification inthe French version of CostAdvantage (in whichfeatures are declared as functions). With eachfeature context are associated all the potentialoperations as sub-contexts as well as a rule base

Table 4

Handling costs between two machines

Cu4x Cu3x Ct3x Ct

Cu4x 0 1.2 0.4 0.2

Cu3x 1.2 0 0.6 0.4

Ct3x 0.4 0.6 0 0.6

Ct 0.2 0.4 0.6 0

(with CostAdvantage).

ARTICLE IN PRESS


used to generate the effective manufacturingprocess according to the geometrical characteris-tics. A given manufacturing feature for a parti-cular product is an instance of the context.

The exploitation of this Product Model makes itpossible to identify and to automatically generateall the Cost Entities ECf

ijk present in a particularpart.

Reminder. ECfijk: cost of the manufacturing opera-

tion j performed on the machine k and generatedby the manufacturing feature i chosen by the user.

6.2.2. Costgrammes Model

The Costgrammes Model (Fig. 13) contains allthe Cost Entities dealing with establishing the costof the problem at hand. There are two types ofCE’s: elementary and parent Cost Entities. Theparent CE’s contain sub-contexts relative to theCE’s that make them, and which can be elemen-tary or not.

At the level of each context of a CE, thefollowing knowledge is associated:

Fig. 13. The Costgrammes Mod

�

el (

the attributes (type, resources, etc.),
� the formula of the cost estimation.
In the case of a PCE, the cost estimationformula is recorded in the form of rules that addthe cost of the elementary Cost Entities afterchecking their instantiation.

6.2.3. Generation of the manufacturing operation

cost

In Fig. 14, in addition to the specification of themanufacturing feature example (entitled ‘‘Threa-

ded_Hole’’) by the estimator operator using thefirst window, the system automatically generates,in the second window, the corresponding manu-facturing process plan and the processing cost ofeach operation by default (The values given areindicative). By double-clicking on each operation,another window pops up to indicate the operationcost for an alternative machine and the values ‘‘i, j,k’’ corresponding to the ECf

ijk.

with CostAdvantage).

ARTICLE IN PRESS

Fig. 14. Generation of the manufacturing process for a ‘‘Threaded_Hole’’.


6.2.4. Activation of the manufacturing constraints

According to the specification of some manu-facturing features, the system alerts the user of aprecedence condition among some of the opera-tions. In Fig. 15, the topological relationship‘‘Starts_on’’ among, for instance, the ‘‘Threaded_

Hole’’ and ‘‘Rectangular_Surface’’, implies theprecedence condition between the finishing surfaceoperation (X22k) and the boring and tappingoperations ðX 14k;X 14kÞ independently of their rea-lization machine k. This is translated into con-straints of the form: U22koU14k and U22koU15k

(Fig. 15).

6.2.5. Activation of the production constraints

If an Unavailable machine is selected, the systemalerts the user for impossible operation. In Fig. 16,the unavailability of the machine Cu3x cancels theoperations X 111 and X 121 concerning the manu-facturing feature ‘‘Rectangular_Surface’’; thismeans cancellation of all the variables: X111i0j0k0

and X121i0j0k0.The Costgrammes Model design provides all the

Cost Entities ECfijk of the Termoz part, as well as

the manufacturing and production constraints tobe respected. The term Xijk allows to determine all

the variables Xijki0j0k0 of the problem with theCon’flex CSP solver.

6.3. Generation of the cost of alternate production

processes

The production process of the Termoz part takesinto account the material handling costs (ECm

ijki0j0k0 ),

the machine preparation costs (ECpr

ijki0j0k0), in addi-

tion to the manufacturing operation costs (ECfijk).

We differentiate the modeling steps on threelevels:

�
The declaration of the Boolean variablesXijki0j0k0, the integer variables Uijk and the fuzzyvariable CPP, � The declaration of the model constraints, the
manufacturing constraints, the production con-straints and the cost constraint,
� The declaration of the CE (ECf
ijk, ECmiji0j0k0 and

ECpr

ijki0j0k0) knowledge.

The modeling of production cost estimating inthe form of a CSP allows to answer the followingquestions: Are there any production processes for

ARTICLE IN PRESS

Fig. 15. Example of manufacturing constraint activation.

Fig. 16. Example of production constraint activation.


ARTICLE IN PRESS


the Termoz part and the workshop at disposalwhose cost is lower than a given value and whosegeometrical specifications can be respected? Whatis the cost of each solution?

The answer to these questions provides thedesigner with an economic evaluation of his/hertechnical choices. We propose the following designscenario for the Termoz part:

The initial part design involves the F1, F2, A1 andTT1 manufacturing features. The part is produced inbatches of 100 units at a 4.9 Euros per unit cost. Anadditional function of the part requires the creationof two new manufacturing features: The Bored_Hole

A2 and the Slot R with the symmetry geometricalspecification RCðR;A2Þ ¼ 0:05 and topological spe-cification R Opens on A2. The acceptable maximalproduction cost of the part is 7.3 Euro/unit.

The calculations realized with Con’flex for thecost estimation of the production process solutionfor the manufacturing workshop at disposal showsthat:

�
For a satisfaction degree equal to at least 0.09(equivalent to the production process for whichthe part unit cost is lower than 7.084 Euro): 49solutions exist. � For a satisfaction degree equal to at least 0.190
(equivalent to the production process for whichthe part unit cost is lower than 6.844 Euros): 49solutions exist.
� For a satisfaction degree equal to at least 0.290
(equivalent to the production process for whichthe part unit cost is lower than 6.124 Euros): 0solution exists.

For the three solutions of Fig. 17, the value‘‘CoutTotal’’ represents the total batch cost (100parts) corresponding to the production processsolution. The mention ‘‘�2.500’’ means, in Con’-flex, that the solution value is an interval centered

on the shown value and the width of which is therange declared for the variable. The value‘‘Sat ¼ ’’ is the satisfaction degree of the solution.Indication ‘‘Nbr of instantiations’’ and ‘‘Nbr ofconstraints tests’’ are a measure of the path (in thetreelike search) that it was necessary to scan to findthe solutions.The solutions corresponding to the minimal

satisfaction degree (0.09) allow to determine theinterval of the part unit cost [CPP min, Cpp max] inthe fuzzy interval [4.9, 7.3] (Fig. 18). The indicativevalues of the costs put in application (machining,machine set-up and handling costs) provide theinterval [6.3, 6.7].

7. Conclusion

The objective of this research work was topropose a cost estimating method fitted to recentindustrial context evolution. It is our belief that allthe concepts and the techniques presented in thispaper sustain this objective. The CE conceptallows to unify the estimation of direct andindirect costs. Establishing the causality relationsexisting between the specification of the manufac-turing features and the engaged costs of themanufacturing operations provides support tothe economic decision at the detailed designstage. The consideration of constraints comingfrom the market, from the manufacturing shopand from production shows the integration capa-city of the cost estimation function. For a givenproduct, the generation of different cost setsrelative to the alternate production processes givesthe quotation function a tool to negotiate the cost/

delay ratio, and provides the designer with a[miniCost, maxiCost] interval to control thetechnical choices.For the future work, it seems very interesting to

investigate the possibilities offered by cost estimat-ing modeling in the form of a flexible CSP

problem. Under this formalism, the variables canhave domains with imprecise borders, the con-straints to be satisfied are more or less importantand, they can be more or less satisfied by thevarious combinations of values of variables.Finally, one could accept a solution that satisfies

ARTICLE IN PRESS

1

0.1 7.3

Cost

Satis

fact

ion

degr

ee

0.5

4.9

Fig. 18. The fuzzy subset [4.9, 7.3] of the total cost.

Fig. 17. Costs of the production process solutions for a satisfaction degree at least equal to 0.390.


only partially all the constraints. Another researchdirection could consist in searching the set ofvariables, domains and constraints that allow toevaluate the design solution by a Quality/Cost

satisfaction degree.An interesting valorization of the work that

could be investigated concerns the incorporationof the CE approach with its Product Model andCostgrammes Model in CAD or CAE work-stations. Depending on form features and partmaterial used the system could estimate a roughcost interval and warn the part designer if he

exceeds the cost limit or provide a decisionguidance in the choice of features of the part.

References

Asiedu, Y., Gu, P., 1998. Product life cycle cost analysis: State

of the art review. International Journal of Production

Research 36 (4), 883–908.

Ben-Arieh, D., Qian, L., 2003. Activity-based cost management

for design and development stages. International Journal of

Production Economics 83 (2), 169–180.

ARTICLE IN PRESS


Bouquin, H., 1997. La Comptabilite de Gestion. Collection Que

Sais-Je? Presses Universitaires de France, Paris.

Brimson, J.A., 1991. Activity Costing—An Activity-Based

Costing Approach. Wiley, New York, NY.

Cavalieria, S., Maccarrone, P., 2004. Parametric vs. neural

network models for the estimation of production costs: A

case study in the automotive industry. International Journal

of Production Economics 91 (2), 165–177.

Cognition, 2000. Cost Advantage, Modeler’s Guide, Cognition

Europe, 1 allee Jean Image, 77200 Torcy, France.

Cooper, R., Kaplan, R.S., 1992. Activity-based system:

Measuring the cost of resource usage. Accounting Horizon

September, 1–13.

DoD, 1999. Joint Industry/Government Parametric Estimating

Handbook, Spring, third ed., US Department of Defense.

Duverlie, P., Castelain, P., 1999. Cost estimation during design

step: Parametric method versus case based reasoning

method. Journal of Advanced Manufacturing Technology

15, 895–906.

Feng, C., Kusiak, A., Huang, C.C., 1996. Cost evaluation in

design with form features. Computer-Aided-Design 28 (11),

879–885.

Garrison, R.H., Noreen, E.W., 1997. Managerial Accounting,

eighth ed. Irwin, Homewood, IL.

H’mida, F., 2002. Contribution a l’estimation des couts en

production mecanique: L’approche Entite Cout appliquee

dans un contexte d’ingenierie integree. Ph.D. Thesis,

University of Metz, Metz, France (in French).

Johnson, H.T., Kaplan, R.S., 1987. Relevance lost, The rise and

fall of management accounting. Havard Business School

Press, Harvard, USA.

Koonce, D., Judd, R., Sormaz, D., Masel, D.T., 2003. A

hierarchical cost estimation tool. Computers in Industry 50,

293–302.

Lakhal, S., Martel, A., Oral, M., Montreuil, B., 1999. Network

companies and competitiveness: A framework for analysis.

European Journal of Operational Research 118 (2),

278–285.

Layer, A., Ten Brinke, E., Van Houten, F., Kals, H., Haasis, S.,

2002. Recent and future trends in cost estimation. Interna-

tional Journal of Computer Integrated Manufacturing 15

(6), 499–510.

Lorino, P., 1991. Le Controle de Gestion Strategique. Dunod,

Paris.

Matthews, L.M., 1983. Estimating Manufacturing Costs.

McGraw-Hill, NY.

Otswald, P.F., 1992. Engineering Cost Estimating. Prentice-

Hall, Englewood Cliffs, NJ.

Ozbayrak, M., Akgunb, M., Turkerc, A.K., 2004. Activity-

based cost estimation in a push/pull advanced manufactur-

ing system. International Journal of Production Economics

87 (1), 49–65.

Parsaei, H., Leep, H.R., Yang, Y., Wong, P.J., 1997. An

integrated manufacturing cost estimating system. Proceed-

ings of the First International Conference on Engineering

Design and Automation (EDA). Integrated Technology

Systems Inc., USA, pp. 116–120.

Rellier, J.P., 1996. Con’flex User’s Manual. INRA- Laboratory

BIA, Toulouse, France.

Roy, R., Bendall, D., Taylor, J.P., Jones, P., Madariaga, P.A.,

Crossland, J., Hamel, J., Taylor, M.I., 1999. Identifying and

capturing the qualitative cost drivers within a Concurrent

Engineering environment. In Advances in Concurrent

Engineering. Technomic Publishing Co. Inc., Pennsylvania,

USA, pp. 39–50.

Sabourin, L., Villeneuve, F., 1996. OMEGA an expert CAPP

system. Advances in Engineering Software 25, 51–59.

Shtub, A., Zimerman, Y., 1993. A neural-network-based

approach for estimating the cost of assembly systems.

International Journal of Production Economics 32 (2),

189–208.

Stewart, R.D., 1991. Cost Estimating. Wiley, NJ.

Tsang, E., 1993. Foundations of Constraint Satisfaction.

Academic Press, London.

Veeramani, D., Joshi, P., 1996. Methodologies for rapid and

effective response to requests for quotation (RFQs). IIE

Transactions 29, 825–838.

Wei, Y., Egbelu, P.J., 2000. A framework for estimating

manufacturing cost from geometric design data. Interna-

tional Journal of Computer Integrated Manufacturing 13

(1), 50–63.

Wierda, L.S., 1990. Design-oriented cost information: The need

and the possibilities. Journal of Engineering Design 1 (2),

147–167.

Wierda, L.S., 1991. Linking design, process planning and cost

information by feature-based modeling. Journal of Engi-

neering Design 2 (1), 3–19.

Wong, J.P., Imam, I.N., Khosravi-Kamrani, A., Parsaei, H.R.,

Tayyari, F., 1992. A Totally Integrated Manufacturing Cost

Estimating System (TIMCES). In: Parsaei, H.R., Mital, A.

(Eds.), Economics of Advanced Manufacturing Systems.

Prentice-Hall, Englewood Cliffs, NJ.

Cost Estimation in Mechanical Production the Cost Entity Approach Applied to Integrated Product...

Documents

Transcript of Cost Estimation in Mechanical Production the Cost Entity Approach Applied to Integrated Product...