Process Yield Improvement Through Optimum Design of Fixture Layouts in 3D Multistation Assembly Systems T. Phoomboplab D. Ceglarek Warwick Digital Laboratory, WMG University of Warwick, Coventry CV4 7AL, UK; Department of Industrial and Systems Engineering, University of Wisconsin–Madison, Madison, WI 53706 Fixtures control the positions and orientations of parts in an assembly process. Inaccuracies of fixture locators or nonoptimal fixture layouts can result in the deviation of a workpiece from its design nominal and lead to overall product dimensional variability and low process yield. Major challenges involving the design of a set of fixture layouts for multistation assembly system can be enumerated into three categories: (1) highdimensional design space since a large number of locators are involved in the multistation system, (2) large and complex design space for each locator since the design space represents the area of a particular part or subassembly surfaces on which a locator is placed, (here, the design space varies with a particular part design and is further expanded when parts are assembled into subassemblies), and (3) the nonlinear relations between locator nominal positions and key product characteristics. This paper presents a new approach to improve process yield by determining an optimum set of fixture layouts for a given multistation assembly system, which can satisfy (1) the part and subassembly locating stability in each fixture layout and (2) the fixture system robustness against environmental noises in order to minimize product dimensional variability. The proposed methodology is based on a two-step optimization which involves the integration of genetic algorithm and Hammersley sequence sampling. First, genetic algorithm is used for design space reduction by estimating the areas of optimal fixture locations in initial design spaces. Then, Hammersley sequence sampling uniformly samples the candidate sets of fixture layouts from those predetermined areas for the optimum. The process yield and part instability index are design objectives in evaluating candidate sets of fixture layouts. An industrial case study illustrates and validates the proposed methodology. 关DOI: 10.1115/1.2977826兴 Keywords: fixture layout, multistation assembly, process yield, instability index, genetic algorithm, Hammersley sequence sampling 1 Introduction Fixture design is one of the most important design tasks during process design for a new product development since it involves defining the locations and orientations of parts during assembly processes as well as providing physical support, which can greatly affect product dimensional variations and process yield. Generally, fixture design process can be divided into three stages which are 共1兲 fixture planning, 共2兲 fixture configuration, and 共3兲 fixture construction 关1兴. In the fixture planning stage, issues related to the number of fixtures needed, the type of fixtures, the orientation of fixture corresponding to orientation, and the joining or machining operations, which fixtures have to handle are identified. The fixture configuration stage determines the layout of a set of locators and clamps on a workpiece surface such that the workpiece is completely restrained. Finally, the fixture construction stage involves constructing fixture components and then installing them to support the workpiece. Specifically for complex assemblies such as an automotive body, a ship hull, and an aircraft fuselage, fixture layout design, which falls under the domain of the fixture planning and fixture configuration stages, is a primary concern and it Contributed by the Manufacturing Engineering Division of ASME for publication in the JOURNAL OF MANUFACTURING SCIENCE AND ENGINEERING. Manuscript received August 17, 2007; final manuscript received July 18, 2008; published online October 10, 2008. Review conducted by Shivakumar Raman. Paper presented at the 2007 International Conference on Manufacturing Science and Engineering 共MSEC2007兲, Atlanta, GA, October 15–17, 2007. involves adjusting the design nominalof locator positions in order to eliminate mean shifts and reduce variations of all key product characteristics 共KPCs兲. However, research on fixture layout design in a multistation assembly process is limited because of the lack of a methodology to predict product dimensional variations and a process yield during the product/process design phase 关2–4兴. The absence of such a methodology poses special challenges in assessing a performance of a fixture layout design and its impact on a process yield. Currently, researchers have developed a variation propagation model for multistation assembly processes using a state-space representation 关5–7兴, which allows prediction of KPC variations under known statistical characterizations of key control characteristics 共KCCs兲. The extension of this work to three-dimensional 共3D兲 rigid body assembly processes has led the development of the so-called stream-of-variation 共SOVA兲 methodology 关8,9兴. Nevertheless, the fixture layout design for multistage assembly processes continues to pose various challenges. For example, multistage assembly processes usually involve a large number of locators since these processes consist of a large number of parts assembled in many assembly stations. For instance, a typical automotive body assembly consists of 200–250 sheet metal parts assembled in 60–100 assembly stations with 1700 to 2100 locators 关4,10兴. In addition, the locating positions used in one station may be reused in the different stations. Thus, a fixture layout design methodology optimizing a fixture layout independently for each assembly station may not necessarily lead to a good design Journal of Manufacturing Science and Engineering Copyright © 2008 by ASME DECEMBER 2008, Vol. 130 / 061005-1 Downloaded 10 Oct 2008 to 137.205.175.119. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm solution because of the interdependencies between the fixture layouts in assembly systems. The challenges in designing a set of fixture layouts for multistation assembly system can be summarized into three categories as follows. 共1兲 High-dimensional design space. Multistation automotive assembly processes have a very large number of locators since the number of locators increases proportionally to the number of parts and assembly stations. 共2兲 Large and complex design space for each locator. The design space of a locator can be defined as any allowed position of the locator on a corresponding workpiece surface. Thus, the design space of each locator varies with the size and surface shape of the workpiece. In addition, the design space is further expanded since fixtures need to be designed not only for individual parts but also for all intermediate subassemblies. 共3兲 Nonlinear relations between locator nominal position and KPCs. Deviations of locators from their nominal positions have nonlinear effects on KPC variations. This paper addresses the aforementioned challenges by proposing a methodology for optimizing fixture layouts in all assembly stations simultaneously. This is to ensure that the variation propagation throughout the whole assembly line is considered to achieve the dimensional quality threshold of the final product. The SOVA model as presented in Refs. 关8,9兴 is applied to assess the performance of fixture layout design on product dimensional variation. The proposed methodology is based on the integration of the genetic algorithm 共GA兲 with Hammersley sequence sampling 共HSS兲. The instability index is also incorporated into the proposed methodology to ensure that a fixture layout design meets the locating stability requirement. The methodology is conducted in two steps. First, the genetic algorithm is used for design space reduction by estimating the areas of optimal fixture locations in initial design spaces. Then, Hammersley sequence sampling uniformly samples the candidate sets of fixture layouts from those predetermined areas for the optimum. GA is selected in the first step since it can handle the nonlinearity between the KPC and locator positions. The shortcoming of GA as a time-consuming heuristic optimization technique is alleviated by incorporating the instability index and discretizing the continuous design spaces of all locators to improve convergence to an optimal solution. The instability index helps to expedite the search capability by eliminating those fixture layouts that do not meet the locating stability requirement. Discretization of continuous design spaces reduces the initial candidate design space of locator positions, which GA can select. HSS is conducted in the second step to compensate for any potentially missed optimum solutions by uniformly sampling the candidate locator positions in the areas defined by conducting GA search in the first step. The uniform sampling of HSS evenly selects the representatives of fixture layouts in all assembly stations, which increase the probability to select near-optimal locator positions. The rest of this paper is organized as follows. The stateof-the-art in fixture layout design is reviewed in Sec. 2. The problem formulation, as well as the descriptions of a process yield calculation based on the SOVA model and instability index, is described in Sec. 3. The proposed multifixture layout design methodology is presented in Sec. 4. A case study illustrating the application of the proposed methodology on an automotive underbody assembly and a comparison of optimization algorithms are presented in Sec. 5. Finally, conclusions are drawn in Sec. 6. 2 Literature Review In general, the fixture layout design has to satisfy four functional requirements, which are 共i兲 locating stability; 共ii兲 deterministic workpiece location; 共iii兲 clamping stability; and 共iv兲 total restraint 关1,11–13兴. Locating stability is related to the design of a fixture layout that can provide static equilibrium of a workpiece when it is placed on fixtures. Second, the fixture should provide 061005-2 / Vol. 130, DECEMBER 2008 the deterministic location for the workpiece to ensure position accuracy during operation. Third, clamping stability involves determining the sequence of clamping and its layout that does not disturb the stability and position accuracy of a workpiece established by locators in the previous two functional requirements. Last, clamps should completely restrain the workpiece to withstand any forces and couples to maintain the workpiece in an accurate position. Past research in fixture layout design has proposed approaches to address these functional requirements 关14–24兴. Locating stability is one of the most important requirements in fixture design since a workpiece has to satisfy this requirement before achieving other functional requirements. Locating stability is mainly concerned with static equilibrium under the given fixturing condition in the presence of manufacturing forces. In addition, the fixture layout design has to ensure that all locators maintain contact with the workpiece throughout the manufacturing operation. Issues involving locating stability begin when the workpiece is placed on locators as these locators provide a support against gravity forces until the workpiece is processed. Thus, locating stability also involves fixture force and kinematic analysis to estimate the necessary clamping forces to maintain a workpiece in equilibrium. Roy and Liao 关14兴 presented a quantitative evaluation of the part location stability based on screw theory. The method proposed by Roy and Liao 关14兴 can help in designing a fixture configuration in automated fixture design environment. Deterministic workpiece location involves designing the locator positions or a fixture layout to provide a unique and accurate position and orientation of a workpiece with respect to its fixture reference frame 关15兴. Common challenges involving the fixture layout design that will meet this functional requirement include the positioning accuracy, which is subject to a random manufacturing error of fixture elements, geometric variability of the workpiece, and workpiece positioning errors induced by fixture position. In general, the position variability of the workpiece can be predicted from the statistical characterization of the dimensioning and tolerancing scheme assigned to the fixtures and their contact points on the workpiece. Thus, determining the fixture layout, which is not sensitive to these variation sources, can minimize the workpiece positional variability. Researchers have responded to this challenge by proposing various methods and sensitivity indices in order to determine the optimal locator positions in a fixture layout. For example, Cai et al. 关16兴 proposed a variational method to design a robust fixture layout using the Euclidean norm of the fixture sensitivity index. Wang 关15兴 and Wang and Pelinescu 关17兴 determined the optimal fixture layout by maximizing the determinant of the information matrix 共D-optimality兲. Carlson 关18兴 assessed the fixture locating scheme in terms of workpiece position errors by using the quadratic sensitivity equation. However, all aforementioned methodologies are limited to fixture layout design involving a single workpiece. Clamping stability and total restraint are functional requirements that are related to determining the clamping positions and forces, which do not affect the part locating stability and position accuracy of workpiece provided by the locators. Clamps apply forces on the workpiece against any external force to ensure total restraint 关11兴. The challenge in designing clamping locations is to minimize the workpiece deformations under clamping and external forces. Many studies have addressed these challenges by adopting the finite element method 共FEM兲 to design locator and clamp layouts. For instance, Menassa and DeVries 关19兴 proposed the integration of Broyden–Fletcher–Goldfarb–Shanno optimization algorithm and FEM simulation to determine the fixture layout that can minimize deflection of the workpiece. Cai et al. 关20兴 proposed the fixture layout optimization for deformable sheet metal parts based on nonlinear programming and FEM analysis. In a similar vein, Krishnakumar and Melkote 关21兴 employed GA to optimize a fixture layout that can minimize the deformation of the machined surface due to clamping and machining forces. In Transactions of the ASME Downloaded 10 Oct 2008 to 137.205.175.119. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm Table 1 State-of-the-art in fixture layout design Fixture design functional requirements Fixture layout design complexity Single workpiece Multistation assembly processes 3D 2D 3D Locating stability Deterministic workpiece location Roy and Liao 关14兴 Wang 关15兴 Cai et al. 关16兴 Wang and Pelinescu 关17兴 Carlson 关18兴 Clamping stability Total restraint Menassa and DeVries 关19兴 Cai et al. 关20兴 Krishnakumar and Melkote 关21兴 Li and Melkote 关22兴 DeMeter 关23兴 Marin and Ferreira 关24兴 Camelio et al. 关25兴 Kim and Ding 关26兴 Izquierdo et al. 关27兴 Proposed in this paper addition, there are also several approaches proposed to determine the clamping design. For instance, Li and Melkote 关22兴 proposed an approach to minimize the maximum positional errors by considering the workpiece dynamics during machining. DeMeter 关23兴 proposed a technique to determine the optimal locator layout based on min-max loading criteria. Marin and Ferreira 关24兴 presented the method to optimize the 3-2-1 fixture layout based on screw theory. Nevertheless, these studies mainly focus on design of fixture layout for a single workpiece. Currently, most of the fixture layout design methodologies are limited to a single workpiece fixture layout design. Table 1 summarizes the current stateof-the-art research in the area of fixture layout design and presented in the context of fixture functional requirements. Current research related to fixture layout design for multistation assembly processes is limited because of the challenges in developing a variation propagation model and computational complexity. Recently, Camelio et al. 关25兴 presented the fixture layout design for compliant part assembly by considering the effects of part variation, tooling variation, and assembly springback. Kim and Ding 关26兴 proposed a methodology to design multifixture layouts in multistation assembly based on a station-indexed state-space model 关5–7,28兴. The extension of Kim and Ding 关26兴 in designing fixture layouts for a product family can be found in the work of Izquierdo et al. 关27兴. Kim and Ding 关26兴 involved determining a set of fixture layouts for all assembly stations that are insensitive to the variations of random manufacturing errors of fixture elements, geometric variability of the workpiece, and workpiece positioning errors induced by fixturing position. The methodology developed by Kim and Ding 关26兴 uses E-optimality to minimize the eigenvalue of the information matrix and exchange algorithm 共EA兲, first proposed by Cook and Nachtsheim 关29兴, to determine the optimal set of fixture layouts. The design objective of the methodology proposed by Kim and Ding 关26兴 is similar to the design objective of this paper in designing a set of multiple fixture layouts, which is robust to environmental noise. However, there are fundamental differences between both methodologies in terms of two-dimensional 共2D兲 versus 3D problem formulation as well as in specifics of the developed methodologies as elaborated below. The design problem addressed by Kim and Ding 关26兴 is limited to 2D assembly processes while this paper focuses on 3D assembly processes. Designing fixture layout in 2D leads to three simplifications, as follows: 共i兲 the locating stability functional requirement is not taken into consideration, 共ii兲 the variations caused by mating joints between the two parts are not included in the 2D model, and 共iii兲 the design space dimensionality, as well as nonlinearity between the KCC locator positions and KPCs in 2D fixture layout design is significantly less than those in the 3D fixture layout design problems since no out-of-plane variation is included in the 2D model. These three simplifications significantly limit the industrial application of 2D fixture layout design methodology Journal of Manufacturing Science and Engineering since 共i兲 the locating stability consideration is mandatory and must be achieved before meeting other functional requirements and 共ii兲 there is a variety of part-to-part joints in the assembly processes such as lap, butt, and T-joints, which have different impacts on product dimensional variations. This paper considers the fixture layout design for 3D assembly processes, which addresses the aforementioned challenges in 2D problems. Furthermore, the methodology proposed in this paper also differs from Kim and Ding’s 关26兴 approach with respect to the following specifics. 共1兲 An evaluation index to assess fixture layout design. Kim and Ding 关26兴 minimized the sensitivity index of information matrix in their approach while the approach proposed in this paper integrates the following two objectives: 共i兲 maximize the percentage of KPCs conforming to specifications and 共ii兲 minimize part locating instability. Evaluation indices used in both methodologies have different advantages and disadvantages. The sensitivity index used by Kim and Ding 关26兴 is less computationally intensive than the calculation ratio of KPCs conforming to design specifications by using simulation techniques such as the Monte Carlo approach. However, sensitivity index is difficult to interpret and to provide explicit relations between the index and product quality while the percentage of KPCs conforming to design specifications can explicitly indicate product dimensional quality. Moreover, part locating stability is used in this paper to ensure that fixture layout design satisfies necessary fixturing functional requirement, which is not considered by Kim and Ding 关26兴. 共2兲 Searching algorithm to select optimal locator positions. Kim and Ding 关26兴 used EA with enhanced computational capabilities done by increasing the exchange rate in each iteration and reducing the candidate fixture locations by using an experience-based approach. Since 2D fixture layout design optimizes only the layout of two locators per part, four-way and two-way locating pins, the reduction of design space based solely on experience is feasible. On the other hand, the methodology proposed in this paper is analytically based and conducts two-step optimization: 共i兲 initial reduction of design space by using GA approach and 共ii兲 uniform sampling for the optimum multiple fixture layouts by using HSS approach. Instead of relying on the designers’ experience, in the first step, GA is employed to reduce the size of all locator design spaces, especially important for the design spaces of NC blocks, which have highly nonlinear relations with KPCs. The first step integrates GA with the part instability index, which further enhances search performance by eliminating the candidate fixture layouts that do not meet the location stability requirement. Overall, GA performs an initial search to deterDECEMBER 2008, Vol. 130 / 061005-3 Downloaded 10 Oct 2008 to 137.205.175.119. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm Table 2 Comparative analysis of Kim and Ding †26‡ and the proposed methodology Kim and Ding 关26兴 Design problem Proposed in this paper 2D fixture layout design • Functional requirement: Deterministic workpiece location • Fixture elements: only four-way and twoway locating pins • Mating joint: only 2D lap-joint design between parts • Design space: 2D design space with moderate nonlinearity between KCCs and KPCs Multifixture layout design methodologies Exchange algorithm • Evaluation index: Minimizing sensitivity index based on Eoptimality approach • Enhanced capability in searching optimal locator positions: Increase number of exchanges per iteration and reduce the candidate space by experience-based approach mine areas that have a higher probability of containing optimal positions for each locator. Then, HSS conducts uniform sampling in the areas around the locator position identified by GA for optimal locator positions. Two-step optimization, which integrates GA and HSS, is very beneficial in a large design space problem with nonlinear relations between locator nominal positions and KPCs, especially when determining the locations of NC blocks in 3D fixture layout design. A comparative analysis of the methodologies proposed in this paper with the approach proposed by Kim and Ding 关26兴 is summarized in Table 2. 3D fixture layout design • Functional requirement: Locating stability and deterministic workpiece location • Fixture elements: four-way locating pins, two-way locating pins, and NC blocks • Mating joint: lap joints, butt joints, Tjoints, and mixed joints in 3D assembly • Design space: 3D design space and high nonlinearity between KCCs and KPCs Two-step optimization which integrates GA and HSS • Evaluation indices: 共i兲 Maximizing percentages of KPCs conforming to specifications and 共ii兲 minimizing the instability index • Enhanced capability in searching optimal locator positions: Using GA and instability index as strategic design space reduction and using HSS to conduct local search The above relations of locator positions, fixture layouts, and a set of fixture layouts are illustrated in Fig. 1. In a single assembly station, all locator positions are required to provide the locating stability for workpieces. Additionally, the set of fixture layouts in a multistage assembly process has to be robust to environment noise, which results in minimum dimensional variations of the final product. In this paper, a process yield and an instability index are used as criteria for determining an optimal fixture system. The design parameters are the locator positions,rk, in a set of fixture layout ⌳. Therefore, the optimization scheme is expressed as maximize⌳ 3 Problem Description subject to Let us denote that there are N locators in a given multistation assembly process. A locator position in the assembly system is described as rk = 兵x,y,z其k, k = 1,2, . . . ,N 共1兲 where 兵x , y , z其k represents the Cartesian coordinates of a locator rk In this paper, a binary assembly process is taken into consideration where two parts are assembled at each station. A fixture layout L in an assembly station includes information about locator positions, which are used for locating two parts: a root part 共S兲 and a mating part 共M兲. Thus, a fixture layout L is represented as a collection of two distinct sets. L = 兵S,M其 S = 兵ri其, M = 兵r j其, 共2兲 i = 1, . . . ,n 共3兲 j = 1, . . . ,n − m 共4兲 where n = 6, if the corresponding part is a rigid body, n ⬎ 6, if a corresponding part is a compliant part, and m is a number of degree-of-freedom 共DOF兲, which the part-to-part joint constrains a mating part. Let ⌳ represents a set of fixture layouts for a given multistation assembly process with p assembly stations. A set of fixture layouts, ⌳, is expressed as ⌳ = 兵L1,L2, . . . ,L p其 061005-4 / Vol. 130, DECEMBER 2008 共5兲 yield共⌳兲 共⌳兲 ⬍ 0 共6兲 where yield 共·兲 is a function for calculating process yield, and 共·兲 is a function for assessing the instability index for a set of fixture layouts in an assembly system. The remaining part of Sec. 3 is organized as follows. The review of design evaluation indices and method to calculate process yield based on SOVA model are presented in Sec. 3.1. To assess and compare the location stability between the two sets of fixture layouts, this paper adopts the instability index based on screw theory proposed by Roy and Liao 关14兴, which is presented in Sec. 3.2. 3.1 Design Evaluation Indices. A fundamental aspect of fixture layout performance is its robustness against environmental Fig. 1 Fixture layout representation for a multistage assembly process Transactions of the ASME Downloaded 10 Oct 2008 to 137.205.175.119. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm noise to minimize product dimensional variability, which is characterized by KPC variations. In assembly processes, the variations of KPCs are caused by two main factors, which are 共1兲 variability of locator positions and 共2兲 part-to-part interactions 关9兴. The position variability of a locator depends on the dimensioning and tolerancing scheme assigned to locators and geometrical shape errors of a workpiece. Product dimensional variability caused by part-to-part interactions stem from errors related to part-to-part mating features, which characterize variation propagation between parts 关10兴. The SOVA model is used in this paper as an assembly response function to assess the robustness of a designed set of fixture layouts against the aforementioned variation sources. The SOVA model allows evaluating KPC variations since it defines relations between KPC and KCC vectors, as shown in Eq. 共7兲. The details in formulating the SOVA model are discussed in Refs. 关8,9兴. 冤 冥 冤 KPC1 KPC2 = ] KPCm m⫻1 SOVA matrix a11 a12 . . . a1,n a21 a22 . . . a2,n ] ] am,1 am,2 ] . . . am,n 冥冤 冥 KCC1 KCC2 ] m⫻n KCCn n⫻1 or KPC = A共⌳兲 · KCC 共7兲 where A共⌳兲 is the SOVA matrix, which is formulated based on the dimensional relationships among fixture layout design, partto-part mating features, and the assembly sequence. In practical, the relations between KCC variations, including locator position variations, and KPC variations are non-linear. Huang et al. 关8,9兴 proposed the technique to approximate the relations into linear model, as shown in Eq. 共7兲. Each element in the SOVA matrix, ai,j, represents as sensitivity of KCC j on KPCi, which consists information regarding to nominal positions of locators and partto-part joint. Locator position adjustment while conducting fixture layout optimization results in different structure of SOVA matrix, which then can be used to compare the performance of each set of fixture layouts. If the variations of KPCs illustrated by variance analysis or simulation are reduced after a fixture layout is adjusted, it indicates that the fixture layout increases in its robustness. We use the SOVA model as the assembly response function,A, to evaluate the impact of KCC variations, u, on the KPC variations, y. The relationship in Eq. 共7兲 can be expressed as y = Au 共8兲 In general, KPC variations can be reduced by 共i兲 tightening the tolerances of KCCs, u or 共ii兲 optimizing the assembly response function represented as a SOVA matrix, A, to be robust to KCC variations. Tightening KCC tolerances is the most straightforward approach to reducing KPC variations. However, its trade-off involves increasing tooling cost in order to produce tooling at higher precision. Optimizing the assembly process is more appealing in practice since KCC tolerance ranges can be increased and the assembly process is still able to achieve the same KPC variation levels. Relaxing KCC tolerances usually leads to lower production cost. Fixture layout design is one approach that can increase assembly process robustness. In a fixture layout design, we aim to determine the locator nominal positions, which can minimize the KPC variations while the tolerances that control variations of KCCs, u, are constant. Different locator positions contribute to the alteration of elements in the SOVA matrix, A. The robustness of the SOVA matrix resulting from altering the locator positions can be assessed by two approaches: 共i兲 a loss function based on sensitivity indices and 共ii兲 an estimated percentage of nonconforming items. In the sensitivity index approach, the product dimensional quality is measured by the variations of yTy = uTATAu. To minimize the variations of yTy, the robustness of the SOVA matrix, A, has to be improved in order to be insensitive to the KCC variation Journal of Manufacturing Science and Engineering inputs, u. The sensitivity index can be defined as the variations of output signals to input noise 关26兴, which can be expressed as S= yTy uTATAu = u Tu u Tu 共9兲 The sensitivity index, S, has to be minimized such that the significant variations of uTu contribute to minor variations of yTy. If the KCC variations of vector u are constant, the KPC variations depend on the assembly response function A. The challenge is to select the design index to assess ATA. Several measures are proposed based on optimality criteria in experimental design. Kim and Ding 关26兴 provided the analysis of the three optimality criteria in fixture layout design which are 共i兲 D-optimality 共min det共ATA兲兲, 共ii兲 A-optimality 共min tr共ATA兲兲, and 共iii兲 E-optimality 共min max共ATA兲; max is the extreme eigenvalue兲. The advantages and disadvantages of these three optimality criteria for fixture layout design are discussed below: D-optimality is to minimize the determinant of a matrix ATA, 共min det共ATA兲兲. The advantage of D-optimality in fixture layout design is that it minimizes both the variances and the covarinces of matrix ATA. It is equivalent to minimizing the overall process m variations; min det 共ATA兲 = min 兿i=1 i, where i is an eigenvalue. D-optimality is very effective to evaluating the design problems, which inherent highly nonlinear relationships such as fixture layout design. However, the singularity of matrix ATA is a major obstacle to the use of D-optimality in multistage fixture layout design. A-optimality is to minimize the trace of matrix ATA, min tr共ATA兲, which is the summation of sensitivities of all KCC-KPC pairs in the assembly processes. Nevertheless, A-optimality does not consider the dimensional variation impact from covariances within matrix ATA. Thus, A-optimality does not imply that the percentage of nonconforming items will be reduced since the covariances among the locator nominal positions on KPC variations are high. E-optimality is to minimize the extreme eigenvalue of matrix ATA, min max共ATA兲. E-optimality is similar to D-optimality, which considers both variances and covariances of all pairs of KCC-KPC, but E-optimality considers only max共ATA兲. Thus, E-optimality can avoid the singularity of matrix ATA during computation, and it is aligned with the Pareto principle in quality engineering 关26兴. However, minimizing only the maximum eigenm value, max共ATA兲 cannot guarantee that overall variations, 兿i=1 i, of the new set of fixture layouts design are decreased. It leaves the possibility that several principle components dominate the overall variations of matrix ATA, and the summations of these eigenvalues can contribute to larger variations even though its extreme eigenvalue is lower than the previous fixture layout design. Therefore, it is difficult to decide that process increases its robustness by assessing only the extreme eigenvalue. On the other hand, the percentage of nonconformance items can be used to evaluate the performance among fixture layout designs by maintaining constant u. In general, process performance is measured by process capability indices, C p or C pk, where C p can be defined as 共USL− LSL兲 / 6; USL and LSL are the upper and lower specification limits, respectively, and is the standard variation of a single KPC variable. In multivariate cases, the KPC tolerance/specification region in multivariate m space is the volume of the hyper-rectangular cube 关30兴, which can be defined as m 兿 共USL − LSL 兲 i i 共10兲 i=1 The KPC variations of a multivariate process can be assessed by using chi-square distance defined as 20 = 共y − 兲⬘ 兺 −1 共y − 兲 共11兲 DECEMBER 2008, Vol. 130 / 061005-5 Downloaded 10 Oct 2008 to 137.205.175.119. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm Fig. 2 KPC variations compared with KPC tolerance region: „a… before optimizing fixture layouts „b… after optimizing fixture layouts However, C p in evaluating multivariate normality KPC variations cannot be obtained directly by dividing the volume of the KPC hyper-rectangular cube specification shown in Eq. 共10兲 with the actual process chi-square distance expressed in Eq. 共11兲 because KPC tolerances/specifications are a hypercube while the chi-square distance has elliptical probability region. Thus, to determine C p, it is necessary to estimate the KPC tolerance region into an ellipsoid shape. As a result, when the process is centered at the target and C p = 1, this implies that 99.73% of the process variations are inside the estimated KPC tolerance ellipsoid. Taam et al. 关30兴 proposed an approach to calculate C p by approximating the KPC tolerance hypercube with the largest ellipsoid that can lie inside the KPC tolerance hypercube. However, to estimate the largest ellipsoid volume is difficult in the case where m ⬎ 3. The process yield proposed in this paper is similar to C p in a multivariate process. However, instead of focusing on estimating the KPC tolerance ellipsoid shape, the process yield defines the probability that the KPC variation vector, y, lies in KPC tolerance hypercube, as illustrated in Fig. 2共b兲. The process yield provides an understandable design criterion for design engineers to evaluate their process design. Moreover, the process yield does not depend on the multivariate normality distribution assumption. To benchmark the robustness of any two processes in the case that 100% probability of KPC variation vectors, y, lies in the KPC tolerance hypercube can also be performed by integrating concept of sensitivity indices to process yield. A-optimality and E-optimality can be used to analyze the variances or principle components of interest. In this paper, the process yield is used as the quality index in assessing the performance of a set of fixture layouts. Yield is defined as a function of KPCs, which represents the probability of all KPCs simultaneously being within their respective specification ranges as shown 再 m yield 共⌳兲 = Pr 艚 LSLi 艋 KPCi 艋 USLi i=1 冎 共12兲 where LSLi and USLi are the lower and upper specification limits for KPCi, respectively. Yield can be estimated by using Monte Carlo technique by simulating k KCC vectors, KCC1, KCC2 , . . . , KCCk, where KCCi = 关KCC1 . . . KCCn兴Ti . A variation of each KCC expressed in Cartesian coordinate, 共␦x , ␦y , ␦z兲, is randomly generated based on its statistical characterizations. Then, the KCCi ; i = 1 , . . . , k, is substituted into Eq. 共7兲 to obtain a vector of, KPC variations, KPC1 , . . . , KPCk, where KPC = 关KPC1 . . . KPCm兴T. ⌽共KPCi兲 is a function to provide a response whether all KPCs are in-specification windows. If all KPC variations are within specification windows; LSL 艋 KPCi 艋 USL, then ⌽共KPCi兲 = 1; otherwise ⌽共KPCi兲 = 0. Thus, yield can be expressed as: 3.2 Instability Index. The instability index, , is adopted from Roy and Liao 关14兴 to compare the locating stability between two fixture layouts. In this paper, a binary assembly process is taken into consideration where at each station a root part is located on the fixture layout S, and then a mating part located by fixture layout M, and a part-to-part joint is assembled to the root part. Thus, the root part has to be in static equilibrium and is fully constrained by fixture layout S before being assembled to a mating part. The static equilibrium of a root part located by fixture layout S, as shown in Fig. 3, can be expressed in matrix form as 冤 0 0 0 −1 −1 0 0 0 0 0 0 −1 1 1 1 0 0 0 r1y r2y r3y 0 0 r6z − r1x − r2x − r3x − r4z − r5z 0 0 0 r4y r5y 0 − r6x 冥冤 冥 冤 冥 0 F1 F2 0 F3 − Wg F4 + − Wgrgy F5 Wgrgx F6 0 =0 共14兲 where Fi, i = 1 , 2 , . . . , 6 represent supporting and locating forces; r1x , r1y , . . . , r6z represent the x , y , z coordinates of six locators in the fixture layout; rgx , rgy , rgz represent the x , y , z coordinates of the center of gravity of the workpiece; Wg represents the weight of the workpiece. If there are only supporting forces from three NC blocks, F1, F2, and F3, against the weight of a root part, Wg, in order to maintain static equilibrium, Eq. 共14兲 can be simplified to 冤 1 1 1 r1y r2y r3y − r1x − r2x − r3x 冥冤 冥 冤 冥 F1 − Wg F2 + − Wgrgy = 0 F3 Wgrgx 共15兲 To calculate the instability index, the following information is required: 共1兲 a new wrench, wd, of external forces and couples to rebalance the static equilibrium in an adjusted fixture layout and 共2兲 twist caused by root part weight, tmg. The wrench and twist can be obtained as described below. 共1兲 A new wrench, wd, of external forces and couples. When k 兺 ⌽共KPC 兲 i yield 共⌳兲 = i=1 k ⫻ 100% 061005-6 / Vol. 130, DECEMBER 2008 共13兲 Fig. 3 3-2-1 fixture layout for prismatic workpiece Transactions of the ASME Downloaded 10 Oct 2008 to 137.205.175.119. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm Step 1: Fixture Planning 1.1 Define the number and types of locators required in the assembly process Output: Λ = L1, L 2 , K, L p { } 1.2 Define the design spaces for N locators Output: ℜ system = {ℜ1 ℜ 2 K ℜ N } 1.3 Discretize design space, ℜi , into t nodes Outputs: ℜi → Oi = {O1 O2 K Ot }i { O system = O 1 , O 2 , K , O N Fig. 4 External wrench, wd, and twist caused by gravity force, tmg Step 2: Design space reduction by Genetic Algorithm (GA) 2.1 Design space reduction by GA maximizeΛ Yield(Λ) the position of any NC block used to locate a root part is adjusted while conducting fixture layout optimization, the previous static equilibrium condition, as shown in Eq. 共15兲, is altered and equilibrium needs to be determined again. For example, if the position of NC block No.1 is changed from 共x1 , y 1兲 to 共x⬘1 , y 1兲, the new equilibrium equation with a new external force and moment vector 共we兲 to balance the root part locating stability can be expressed as in Eq. 共16兲 where the force vector, 关F1 F2 F3兴T, acting at the NC block is unchanged. 冤 subject to 1 1 ry1 ry2 ry3 − r⬘x1 − rx2 − rx3 fz { } 2.1 Define design spaces of interest, A i , around GA candidate nodes Output: ℜnew = {A1 A 2 K A N } Step 3: Local search by Hammersley Sequence Sampling (HSS) 3.1 Projection of design space of interest, A i , into 2D plane 冥冤 冥 冤 冥 冤 冥 冤 冥 − Wg F1 ϖ ( Λ) < 0 1 2 N , O GA , K , O GA Output: O GA = O GA we 1 } D = {B1 B 2 K B N } Output: ℜ 2new 3.2 Uniform sampling sets of fixture layouts by HSS Output: Λ1, Λ 2 ,K, Λ n 0 F2 + − Wgrgy + M x = 0 0 F3 Wgrgx My 共16兲 Step 4: Evaluation of sampled sets of fixture layouts 4.1 Evaluation of sets of fixture layout generated by HSS against design objective maximizeΛ Yield(Λ ) Both the force and moments of we can be obtained from Eq. 共16兲, and they can be presented in a wrench form as wd = 关0 0 fz Mx 0兴T My 共17兲 共2兲 Twist caused by root part weight, tmg: To obtain the twist tmg of a root part caused by its weight, let us assume that the workpiece undergoes an infinitesimal movement caused by gravity force Wg, as shown in Fig. 4. This movement can be expressed as a twist about the origin of coordinates: tmg = x y z vx vy v z subject to ϖ ( Λ ) < 0 Output: an optimum set of fixture layouts Λ optimum Fig. 5 The methodology 1/rgx 0 0 0 − 1兴 共20兲 The instability index defined by using virtual work can be interpreted as follows. 共1兲 ⬎ 0 represents the positive work done by wrench wd in accomplishing twist tmg. This positive virtual work implies that the adjustment of a locator position in a new fixture layout reduces the workpiece stability. Therefore, in the proposed methodology in this paper, it is concluded that the adjusted fixture layout is worse than the pre-adjusted one. 共2兲 = 0 represents that no work is done by wrench wd in accomplishing twist tmg. The virtual work = 0 can be interpreted that there is no improvement in locating stability condition after the adjustment of a locator position. 共3兲 ⬍ 0 represents the negative work done by wrench wd in Journal of Manufacturing Science and Engineering of the proposed accomplishing twist tmg. The negative virtual work implies that the adjustment of locator position into a new fixture layout increases the workpiece stability. Therefore, in the proposed methodology in this paper, it is concluded that the adjusted fixture layout is better than the pre-adjusted one. 共19兲 The instability index represents the virtual work done by the wrench wd against the twist tmg. The instability index, , can be obtained by = f xv x + f y v y + f zv z + M x x + M y y + M z z procedure 共18兲 where x , y , z are the components of part angular displacement, and vx , vy , vz are components the of part translational displacement. For example, the twist of gravity force, Wg, about the origin of coordinates, as shown in Fig. 4, is tmg = 关− 1/rgy optimization 4 Methodology In keeping with the previous discussion, multiple fixture layouts of all assembly stations are required to be designed simultaneously results in a high-dimensional optimization problem. Moreover, the design space of each locator is large, and the locator nominal positions have nonlinear relationships with KPCs. In order to address these challenges, this paper proposes a two-step optimization approach based on genetic algorithm for design space reduction and Hammerley sequence sampling for direct search optimization scheme in a design space predetermined by GA. Yield and instability index are incorporated to the proposed design approaches as design objectives. The procedure of the proposed methodology in designing multiple fixture layouts is shown in Fig. 5. The proposed methodology is based on the following assumptions. 1. All parts are rigid body. 2. The Locator-part constraint is characterized by frictionless point contact. DECEMBER 2008, Vol. 130 / 061005-7 Downloaded 10 Oct 2008 to 137.205.175.119. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm Fig. 6 Mating part degrees of freedom allocation between part-to-part joint and locators 3. The part-to-part joints always maintain full contact to each other. 4. Locators are not only to determine the positions and orientations of the part but also function as clamps in constraining parts. 5. Only fixture and part-to-part joint errors are considered, which are considerably small compared with part and assembly dimensions. Step 1. Fixture planning. Step 1.1. Define the number and types of locators required in the assembly process. The number and types of locators required to constrain root, S, and mating, M, parts in each assembly station 共L = 兵S , M其兲 are defined in this step. It also can be observed that the number of locators in each assembly station depends on the part-to-part joint design and assembly sequence. To illustrate this, let us assume an assembly process of two rigid parts, each with 6 DOFs. The first part positioned in the assembly station called a root part is fully constrained by fixture locators. The second part positioned in the assembly station is called a mating part and has its 6 DOFs constrained by part-to-part joint and fixture locators. The potential allocation of mating part 6 DOFs to be constrained by fixture locators 共M兲 and by part-to-part joint is shown in Fig. 6. Furthermore, the assembly sequence in selecting of root and mating part in each station directly affects the fixture layout design. An example of fixture layouts for S and M parts is illustrated in Fig. 7. The root part in Fig. 7共a兲 is located by a typical 3-2-1 fixture layout. The locators for the root part S consist of two locating pins, four-way pin 共P4way兲 and two-way pin 共P2way兲, and three net contact blocks 共NC1–3兲. Two locating pins constrain 3 DOFs in the X-Y plane, where P4way controls the part translation in the X and Y axes, and P2way controls the rotation of the root part about Z axis. Three NC blocks constrain the remaining 3 DOFs, which are translation in Z axis and rotations about the X and Y axes. Thus, the locators required for the root part in 3-2-1 fixture locating scheme can be defined as r r , P2way ,NCr1,NCr2,NCr3其 S = 兵P4way 共21兲 Then, the mating part is located by part-to-part joint and the remaining of DOFs are constrained by fixture locators, M. The DOFs of a part-to-part joint, which constrain the mating part, affect the fixture planning, as shown in Fig. 6. The part-to-part joint shown in Fig. 7共b兲 constrains the 3 DOFs of the mating part: translation in Z axis and rotations about X and Y axes. Therefore, the remaining DOFs of the mating part are constrained by P4way and P2way fixture locators. Fig. 7 3-2-1 fixture layout for a single station assembly 061005-8 / Vol. 130, DECEMBER 2008 Transactions of the ASME Downloaded 10 Oct 2008 to 137.205.175.119. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm Rsystem = 兵R1 R2 ... R N其 共27兲 Step 1.3 Discretize design space, Ri, into nodes. To reduce the computational effort of GA, the continuous design space, Ri, for each locator is discretized into nodes. These nodes represent the candidate design space to be considered in the next step. The design space of locator ri discretized into t nodes, as shown in Fig. 9, can be expressed as Oi = 兵O1 O2 ... O t其 i where O j = 共x j,y j,z j兲 苸 Ri 共28兲 and j = 1, . . . ,t Fig. 8 Example of part-to-part joints used in automotive body assembly In an automotive body assembly, the following three types of part-to-part joints are widely used: 共1兲 lap joint, 共2兲 butt joint, and 共3兲 T-joint 关4,10兴, as shown in Fig. 8. In this paper, a part-to-part joint is assumed to constrain 3 DOFs of a mating part. A set of fixture locators required for the mating part M where part-to-parts are lap joint, butt joint, and T-joint can be defined as follows, respectively: m m , P2way 其 Mlap = 兵P4way 共22兲 m m ,NCm Mbutt = 兵P2way 1 ,NC2 其 共23兲 m m ,NCm MT = 兵P2way 1 ,NC2 其 共24兲 Thus, the fixture layout for the single assembly station with a predetermined part-to-part joint 共Mjoint兲 can be defined as r r , P2way ,NCr1,NCr2,NCr3,Mjoint其 L = 兵S,M其 = 兵P4way 共25兲 Step 1.2. Define the design space for each locator. The design space, Ri, of each locator, ri, is defined as an area on a workpiece that the locator can be placed. Design engineers justify the locator design space by considering other design constraints in the subsequent fixture construction stage such as allowable maximum deformation of workpiece or the potential difficulty in locator installation and calibration. The continuous design space of a locator ri can be defined as Ri = 兵共x,y,z兲其, x 苸 关lx,ux兴, y 苸 ly,uy z = f i共x,y兲 共26兲 where 共x , y , z兲 represents the Cartesian coordinate of a locator ri; lx,y and ux,y are the lower and upper boundaries in the x and y axes, respectively; and z = f i共x , y兲 is the workpiece surface shape function. For example, in Fig. 9 the design space Ri covers the whole part. Let us assume that a set of fixture layouts ⌳ consists of N locators in a given assembly system. Then, the design space of the assembly process can be expressed as The continuous design spaces 共R1 , . . . , RN兲 of N locators in a given assembly system after the discretization into nodes can be expressed as Osystem = 兵O1,O2, . . . ,ON其 共29兲 Step 2. Design space reduction by genetic algorithm (GA). Step 2.1: Design space reduction by GA. Since the design space for each locator can be relatively large and its position has the nonlinear relations with multiple KPCs, it is necessary to reduce the design space to the area that potentially contains the optimal locator position. The design space reduction is conducted by using the GA approach. In each iteration, GA selects one node from each candidate space Oi; i = 1 , . . . , N, to formulate a candidate set of fixture layouts, ⌳ = 兵O1j , . . . , ONj 其. The GA optimization scheme is expressed as determining a set of fixture layout ⌳, which maximizes the process yield subjected to a constraint of instability index,共⌳兲, as shown below: maximize⌳ subject to yield共⌳兲 共⌳兲 ⬍ 0 共30兲 In general, the genetic algorithm adopted in this paper involves four major steps, as illustrated in Fig. 10. In the first step, the Cartesian coordinates of locator positions, which are aimed to be optimized, are modeled into a chromosome vector. In the second step, the initial population size in each generation is defined. Initial populations in this paper are selected randomly with uniform distribution function. Then, the process yield of individual population is evaluated subjected to locating stability requirement, as shown in Eq. 共30兲. In the last step, the chromosome of population is improved by selecting the best individual to reproduce in the next generations. The reproduction process to improve the chromosome involves two functions: 共i兲 crossover function and 共ii兲 mutation function. Crossover function is to combine two individuals, or parent, to produce a new individual. In this paper, crossover function randomly selects chromosomes from parents by generating binary vector, which have a length equal to a number of chromosome in a population. If an element of a binary vector is 1, a chromosome is selected from the first parent. On the other hand, a chromosome is selected from the second parent if an element of Fig. 9 Illustration of the proposed methodology Steps 1.2–3.1 Journal of Manufacturing Science and Engineering DECEMBER 2008, Vol. 130 / 061005-9 Downloaded 10 Oct 2008 to 137.205.175.119. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm Fig. 11 Sets of fixture layouts generation by HSS 2D Rnew = 兵B1 Fig. 10 Step 2 design space reduction by genetic algorithm binary vector is 0. Mutation function is to create small random changes in individual population, which help to prevent GA trapped in local optima. GA is terminated when there is no improvement in a process yield or the number of iterations reaches the maximum number of generations designated in the GA procedure. Denote the ⌬ as the improvement in the instability index, i , which is i.e., ⌬ = 共⌳i兲 − 共⌳i−1兲. Figure 9 shows a node OGA the optimal position of locator ri selected by GA. The optimal positions of all locators in the system selected by GA can be expressed as 1 2 N OGA = 兵OGA ,OGA , . . . ,OGA 其 Rnew = 兵A1 Ai = 兵共x,y,z兲其, A2 x 苸 关l⬘x ,u⬘x 兴, A N其 共32兲 y 苸 关l⬘y ,u⬘y 兴, z = f i共x,y兲 共33兲 ... where l⬘x,y and u⬘x,y are the lower and upper boundaries of x and y coordinates in design space of interest Ai, respectively; z = f i共x , y兲 is the workpiece surface shape function. Step 3. Local search by Hammersley sequence sampling (HSS). Step 3.1. Projection of design space of interest, Ai, into 2D plane. The direct sampling of potential locator positions on part surface is very complex and time consuming since geometrical information of part surface has to be included in the HSS algorithm. Therefore, the sampling procedure can be simplified by selecting the locator positions in 2D plane. The 2D plane is obtained by projecting design space of interest Ai into a given plane. Design space of interest in 3D space, Ai, is projected into 2D space, Bi, as shown in Fig. 9. The 2D space, Bi, can be expressed as follows: Bi = 兵共x,y兲其, x 苸 关l⬘x ,u⬘x 兴 and y 苸 关l⬘y ,u⬘y 兴 共34兲 where Bi is the projection area in 2D plane of Ai, which consists of a set of points within the lower and upper boundaries of x and y coordinates. Design spaces of interest for all locators in 2D plane can be expressed as 061005-10 / Vol. 130, DECEMBER 2008 ... B N其 Bi:共x j,y j兲 → Ai:共x j,y j,z j = f共x j,y j兲兲 = ri, ⌳ = 兵r1, . . . ,rN其 共31兲 Step 2.2. Define design space of interest around GA candidate nodes. The area around each node predetermined by GA in Eq. 共31兲 is defined as the new design space so-called design space of interest, Ai. The design space of interest covers the area that GA did not consider, which might contain the optimal locator position. The size of design space of interest is related to the grid size defined in Step 1. It can be seen in Fig. 9 that design space of interest of locator ri is significantly smaller than an initial design space, Ri. The new design space of the assembly system can be expressed as B2 共35兲 Step 3.2. Uniform sampling sets of fixture layouts by HSS. The optimal positions of all locators are searched by using HSS. To sample the locator positions, first the transformed 2D design spaces 兵B1 B2 . . . BN其 are formulated into N-dimensional hypercube design space. The example hypercube with three 2D plane design spaces 共B1–3兲 is shown in Fig. 11共a兲. The samples are selected uniformly in the hypercube shown in Fig. 11共a兲. Then, the sample points are projected onto each facet of the hypercube, which is a 2D plane design space, as shown in Fig. 11共b兲. Finally, the sampling locator positions in the 3D design space 兵A1 A2 . . . AN其 can be obtained by substituting the coordinates of samples in the 2D design spaces into corresponding part surface functions; z = f i共x , y兲 ; i = 1 , . . . , N, as shown in Fig. 11共c兲. The transformation of locator position sampled in 2D plane, Bi, into 3D design space, Ai, is mathematically expressed as i = 1, . . . ,N 共36兲 共37兲 Step 4. Evaluation of sampled sets of fixture layouts. The n sets of fixture layouts; ⌳1 , ⌳2 , . . . , ⌳n, generated by using HSS are evaluated by 共i兲 formulating SOVA models, A共⌳兲, for each set of fixture layouts; and 共ii兲 conducting Monte Carlo simulation to obtain a process yield. The optimization problem is formulated as follows: maximize⌳ subject to 5 yield共⌳兲 共⌳兲 ⬍ 0 共38兲 Case Study: Floor Pan Assembly The developed methodology is illustrated and validated by applying it to automotive underbody assembly process. Floor Pan subassembly consists of four parts; floor pan left 共FPL兲 and right 共FPR兲, and bracket left 共BrktL兲 and right 共BrktR兲, assembled in three stations, as shown in Fig. 12. The dimensional quality of the floor pan assembly is evaluated by 12 KPCs and reported as process yield. The manufacturer aims to design the fixture layouts in three assembly stations to improve the robustness of the floor pan assembly process by increasing the process yield from a current level of 85% without tightening any KCC tolerances, i.e., without increasing tooling cost. The process yield is assessed by Monte Carlo simulation in which the variations of fixture locators and part-to-part joints are set according to their tolerances. Tolerances of locators are assumed to be ⫾0.3 mm. while part-to-part joint tolerances are assumed to be ⫾0.5 mm. for linear variation and ⫾0.25 degree for angular variation. The proposed multifixture layout optimization methodology is applied to this case study. The comparative study in terms of optimization performances among the proposed methodology and other fixture layout optimization algorithms are also shown in this section. Step 1: Fixture planning. Step 1.1. Define the number and types of fixtures required in the assembly process. Transactions of the ASME Downloaded 10 Oct 2008 to 137.205.175.119. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm ⌳ = 兵L1,L2,L3其 共40兲 Step 1.2. Define the design space for each locator. The design spaces of all locators in all assembly stations are defined in Tables 4–6. The design space for each locator is shown in Figs. 13–18. The coordinates of center of gravity and weight of each root part in each assembly station are shown in Table 7. Step 1.3. Discretize design space, Ri, into nodes. The design spaces of all four-way and two-way locating pins are discretized into grid of 50⫻ 50 mm2. Similarly, the design spaces of NC blocks are discretized into grid of 100⫻ 110 mm2. Figure 19 shows one example of dicretization of design space R3 representing NC block NC1s 1 into 60 nodes. All 21 discretized design spaces in the system can be expressed as Osystem = 兵O1,O2, . . . ,O21其 Fig. 12 Floor pan assembly The assembly sequence, part-to-part joints, and required locators in each assembly station are shown in Table 3. The fixture layouts for three assembly stations and a set of fixture layouts for the system can be formulated as follows L1 = 兵S1,M1其, L2 = 兵S2,M2其 and L3 = 兵S3,M3其 共39兲 共41兲 Step 2. Design space reduction by GA. Step 2.1. Design space reduction by GA. In each iteration, GA selects one node from each of 21 candidate design spaces in Eq. 共41兲 to formulate a set of fixture layouts ⌳, which is then evaluated by using a process yield and an instability index, . The optimization scheme is expressed, as shown in Eq. 共30兲. The genetic algorithm configurations are selected as follows: 共1兲 chromosomes are formulated by Cartesian coordinates of 21 locators; 共2兲 population size has 50 individuals in each generation; 共3兲 selection of parents for reproduction is based on ranking for the most fitness and stochastic uniform selection with Table 3 Required locators in each assembly station Station No. Parts Joints Required locators FPL 共root part兲 FPR 共mating part兲 Lap joint 1s 1s 1s 1s , P2way , NC1s S1 = 兵P4way 1 , NC2 , NC3 其 1m 1m M1 = 兵P4way , P2way其 Station 1 Station 2 Subassembly 共FPL+ FPR兲 Bracket left 共mating part兲 Lap joint 2s 2s 2s 2s S2 = 兵P4way , P2way , NC2s 1 , NC2 , NC3 其 2m 2m M2 = 兵P4way , P2way 其 Station 3 Subassembly 共FPL+ FPR+ BrktL兲 Bracket right 共mating part兲 Lap joint 3s 3s 3s 3s S3 = 兵P4way , P2way , NC3s 1 , NC2 , NC3 其 3m 3m , P2way 其 M3 = 兵P4way Table 4 Design spaces of locators in Station 1 Root part Station 1 Locators Design spaces Mating Part 1s P4way 1s P2way NC1s 1 NC1s 2 NC1s 3 1m P4way 1m P2way R1 R2 R3 R4 R5 R6 R7 Table 5 Design spaces of locators in Station 2 Root part Station 2 Locators Design spaces Mating Part 2s P4way 2s P2way NC2s 1 NC2s 2 NC2s 3 2m P4way 2m P2way R8 R9 R10 R11 R12 R13 R14 Table 6 Design spaces of locators in Station 3 Root part Station 3 Locators Design spaces Mating part 3s P4way 3s P2way NC3s 1 NC3s 2 NC3s 3 3m P4way 3m P2way R15 R16 R17 R18 R19 R20 R21 Journal of Manufacturing Science and Engineering DECEMBER 2008, Vol. 130 / 061005-11 Downloaded 10 Oct 2008 to 137.205.175.119. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm Fig. 16 Locator design spaces of bracket left „BrktL, mating part… in Station 2 Fig. 13 Locator design spaces on floor pan left „FPL, root part… in Station 1 a crossover fraction of 0.8; 共4兲 mutation function is based on the Gaussian distribution; 共5兲 crossover function is based on using a binary vector as criterion; and 共6兲 the GA is terminated at 5000 generations or there is no improvement in a process yield within 50 consecutive generations. The locator positions selected by GA are shown in Tables 8–10 and Figs. 20–22. Step 2.2. Define design space of interest around GA candidate nodes. The area around each node predetermined by GA is defined to be the design space of interest, Rnew = 兵A1 A2 . . . A21 其. However, in this case study, some locators are required to be defined with additional constraints. First, both bracket left and right have limited areas to place four-way and two-way locating pins. Thus, 2m the positions of four-way and two-way pins for both parts 共P4way , 2m 3m 3m P2way, P4way, and P2way兲 are assigned to be the same as the locations selected by GA. Second, to minimize the need for locating holes on the parts, some locators are reused and have the same position in different assembly stations. These locators are 共1兲 1s 2s 3s 1s 2s 3s P4way , P4way , and P4way ; 共2兲 P2way , P2way , and P2way ; 共3兲 NC1s 3 , 3s 1s 2s 3s 2s NC2s , and NC ; 共4兲 NC , NC , and NC ; and 共5兲 NC and 1 1 1 2 2 3 NC3s 3 . Therefore, the design spaces of interest in each group of these locators are equivalent to the integration of areas around each node within the group. The design spaces of interest for all Fig. 17 Locator design spaces of the root part in Station 3 „the coordinates of the boundaries are the same, as shown in Fig. 15… Fig. 18 Locator design spaces of bracket right „BrktR, mating part… in Station 3 Table 7 Center of gravity coordinates of a root part in three assembly stations Fig. 14 Locator design spaces on floor pan right „FPR, mating part… in Station 1 Fig. 15 Locator design spaces of the root part „FPL+ FPR… in Station 2 061005-12 / Vol. 130, DECEMBER 2008 Center of gravity coordinates Station 1 共FPL兲 Station 2 Subassembly 共FPL+ FPR兲 Station 3 Subassembly 共FPL+ FPR+ BrktL兲 X Y Z Weight 共kg兲 1250 −290 0 60 1250 0 0 120 1340 −110 0 150 Fig. 19 An example of design space discretization Transactions of the ASME Downloaded 10 Oct 2008 to 137.205.175.119. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm Table 8 Locator positions selected by GA in Station 1 Root part Mating part Station 1 Coordinates 1s P4way 1s P2way NC1s 1 NC1s 2 NC1s 3 1m P4way 1m P2way X Y Z 900.00 −250.00 0.00 1700.00 −250.00 0.00 1650.00 −180.00 0.55 750.00 −400.00 0.00 1250.00 −510.00 8.94 1700.00 231.35 0.00 800.00 231.35 0.00 Table 9 Locator positions selected by GA in Station 2 Root part Mating part Station 2 Coordinates 2s P4way 2s P2way NC2s 1 NC2s 2 NC2s 3 2m P4way 2m P2way X Y Z 750.00 −300.00 0.00 1700.00 −300.00 0.00 1350.00 −620.00 25.34 1550.00 −290.00 0.00 1050.00 191.35 1.39 1556.00 −154.00 90.00 1556.00 −551.00 90.00 Table 10 Locator positions selected by GA in Station 3 Root part Mating part Station 3 Coordinates 3s P4way 3s P2way NC3s 1 NC3s 2 NC3s 3 3m P4way 3m P2way X Y Z 750.00 −300.00 0.00 1700.00 −300.00 0.00 1450.00 −620.00 25.34 1550.00 −290.00 0.00 1150.00 301.35 1.39 1556.00 135.00 90.00 1556.00 532.00 90.00 locators are shown in Fig. 23. This case study uses 2 ⫻ 2 grid size around the nodes to define design space of interest. Step 3. Local search by HSS. Step 3.1. Projection of design space of interest, Ai, into 2D plane. The boundaries of the 2D planes, B, are shown in Tables 11–13. Step 3.2. Uniform sampling sets of fixture layouts by HSS. One thousand sets of fixture layouts, ⌳i ; i = 1 , . . . , 1000, are sampled from the hypercube of 21 2D plane design spaces, B1–21 by using HSS. The sampled locator positions in 2D plane design spaces, B1–21, are projected onto part surfaces to obtain the actual positions in 3D design space, A1–21 by using CAD software to acquire the coordinates on workpiece. Step 4. Evaluation of sampled sets of fixture layouts. One thousand sets of fixture layouts are evaluated for their process yield and stability. The set of fixture layouts that has the highest process yield and satisfy stability requirement is considered to be the optimum set of fixture layouts. The locator positions of the optimum set of fixture layouts are shown in Tables 14–16 and Figs. 24 and 25. The process yield of the optimum set of fixture layouts is 96.16% while the process yield of the initial industrial design is around 85%. The performance of the proposed optimization algorithm is Fig. 20 Locator positions selected by GA in Station 1 Fig. 22 Locator positions selected by GA in Station 3 Fig. 21 Locator positions selected by GA in Station 2 Fig. 23 Locator design spaces of interest Journal of Manufacturing Science and Engineering DECEMBER 2008, Vol. 130 / 061005-13 Downloaded 10 Oct 2008 to 137.205.175.119. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm Table 11 The upper and lower boundary coordinates of the projected 2D plane design space in Station 1 Locator 1s P4way 1s P2way NC1s 1 NC1s 2 NC1s 3 1m P4way 1m P2way Design space of interest Projected 2D plane design space Upper boundary of 2D plane 共x,y兲 Lower boundary of 2D plane 共x,y兲 A1 A2 A3 A4 A5 A6 A7 B1 B2 B3 B4 B5 B6 B7 共950, −250兲 共1750, −250兲 共1750, −180兲 共850, −300兲 共1550, −400兲 共1750,281兲 共850,281兲 共750, −350兲 共1650, −350兲 共1450, −400兲 共750, −500兲 共1150, −620兲 共1650,231兲 共750,231兲 benchmarked with other optimization algorithms in terms of 共i兲 methodology ffectiveness measured by closeness of its solution to the global optimum and 共ii兲 computational efficiency measured by time to converge to optimum solutions. However, it is considerably difficult to determine the global optimum in nonlinear optimization problems by all available optimization algorithms without an exhaustive search 关26兴. Instead of searching for the global optimum, this optimization algorithm proposed in this paper provides an optimum set of fixture layouts, which contributes to a significant improvement in the process yield and workpiece sta- bility with a minimal computational effort. The methodology effectiveness and computational efficiency of the proposed optimization algorithm are compared with those of sequential quadratic programming 共SQP兲, simplex search, genetic algorithm, and EA used by Kim and Ding 关26兴. The performances of optimization algorithms are summarized in Table 17. A gradient-based search such as sequential quadratic programming is used in determining an optimal fixture layout design 关16,19兴. The disadvantage of a gradient-based search is that the design solutions are easily entrapped in a local optimum since its Table 12 The upper and lower boundary coordinates of the projected 2D plane design space in Station 2 Locator 2s P4way 2s P2way NC2s 1 NC2s 2 NC2s 3 2m P4way 2m P2way Design space of interest Projected 2D plane design space Upper boundary of 2D plane 共x,y兲 Lower boundary of 2D plane 共x,y兲 A8共=A1兲 B8 A9共=A2兲 B9 A10共=A5兲 B10 A11共=A3兲 B11 A12 B12 A13 B13 A14 B14 共950, −250兲 共1750, −250兲 共1550, −400兲 共1750, −180兲 共1250,191兲 共1556, −154兲 共1556, −551兲 共750, −350兲 共1650, −350兲 共1150, −620兲 共1450, −400兲 共950,401兲 共1556, −154兲 共1556, −551兲 Table 13 The upper and lower boundary coordinates of the projected 2D plane design space in Station 3 Locator 3s P4way 3s P2way NC3s 1 NC3s 2 NC3s 3 3m P4way 3m P2way Design space of interest Projected 2D plane design space Upper boundary of 2D plane 共x,y兲 Lower boundary of 2D plane 共x,y兲 A15共=A1兲 A16共=A2兲 A17共=A5兲 A18共=A3兲 A19共=A12兲 A20 A21 B15 B16 B17 B18 B19 B20 B21 共950, −250兲 共1750, −250兲 共1550, −400兲 共1750, −180兲 共1250,191兲 共1556,135兲 共1556,532兲 共750, −350兲 共1650, −350兲 共1150, −620兲 共1450, −400兲 共950,401兲 共1556,135兲 共1556,532兲 Table 14 The optimal locator positions in Station 1 Root part Station 1 Coordinates X Y Z Mating part 1s P4way 1s P2way NC1s 1 NC1s 2 NC1s 3 1m P4way 1m P2way 800.00 −270.00 0.00 1700.00 −270.00 0.00 1743.00 −205.00 1.42 777.00 −275.00 0.00 1308.00 −611.00 29.62 1700.00 240.00 0.00 800.00 240.00 0.00 Table 15 The optimal locator positions in Station 2 Root part Station 2 Coordinates X Y Z Mating part 2s P4way 2s P2way NC2s 1 NC2s 2 NC2s 3 2m P4way 2m P2way 800.00 −270.00 0.00 1700.00 −270.00 0.00 1308.00 −611.00 29.62 1743.00 −205.00 1.42 1099.00 299.00 0.00 1556.00 −154.00 90.00 1556.00 −551.00 90.00 061005-14 / Vol. 130, DECEMBER 2008 Transactions of the ASME Downloaded 10 Oct 2008 to 137.205.175.119. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm Table 16 The optimal locator positions in Station 3 Root part Station 3 Coordinates X Y Z 3s P4way 3s P2way NC3s 1 NC3s 2 NC3s 3 3m P4way 3m P2way 800.00 −270.00 0.00 1700.00 −270.00 0.00 1308.00 −611.00 29.62 1743.00 −205.00 1.42 1099.00 299.00 0.00 1556.00 135.00 90.00 1556.00 532.00 90.00 searching algorithm is based on the steepest ascent/descent direction. Moreover, it is difficult for the gradient-based search to obtain the derivative information where a process yield and an instability index are the quality measures. To illustrate this, the process yield is multivariate probability density function of random KPC variables, which depend on locator position variations. Therefore, to formulate the explicit model showing relationships between the process yield and nominal locator positions, which allow obtaining derivative information, is infeasible in the highdimensional problems. In addition, the gradient-based search does not consider the workpiece stability because of the difficulty in formulating the model that can represent instability index and yield measure simultaneously. A simplex search can be used to determine the optimal fixture layouts. Although a simplex search is a direct search method, which does not require gradient derivative information, the design solutions obtained from a simplex search also easily converges to local optimum. A simplex search was applied on the floor pan subassembly case study by using fminsearch function available in MATLAB. The design spaces are defined as continuous and the Fig. 24 The optimal locator positions selected by HSS Fig. 25 The optimal locator positions selected by HSS on „a… bracket left and „b… bracket right Table 17 Comparison of optimization methods Methodology effectiveness Optimization methodologies Gradient-based search Simplex search Genetic algorithm Exchange algorithm Proposed methodology Mating part Computational efficiency Process yield Workpiece stability consideration 共Yes/No兲 Computational time 共s兲 — 88.24% 95.40% — 96.16% No Yes Yes No Yes — 3600 8103 — 2062 Journal of Manufacturing Science and Engineering searching operation is terminated in 1 h. The sets of fixture layouts sampled by the simplex search are validated for workpiece stability first, and then the process yields are calculated. The result shows that the process yield of the optimal set of fixture layouts increases only 2.1% from an initial design and 7.9% lower than process yield of the optimum fixture layouts obtained from the proposed method in this paper. The performance of genetic algorithm in determining the optimal set of fixture layouts was also conducted. Although genetic algorithm can avoid the design solutions converging to local optima, it takes considerably computational time in continuous design spaces. For example, genetic algorithm was applied on the floor pan subassembly case study to determine the optimum fixture layouts. It takes 8103 s in computational time to obtain a set of fixture layouts that can meet 95% yield and satisfy workpiece stability requirement, while the proposed methodology can achieve the same result within 2062 s. The performances of the exchange algorithm proposed by Kim and Ding 关26兴 is also studied and compared with that of the proposed methodology. In terms of algorithm effectiveness, both exchange algorithm and the proposed methodology can provide optimal sets of fixture layouts, which meet the dimensional quality requirement. However, the optimal result obtained from EA is sometimes practically infeasible since the workpiece stability is not incorporated in its design criteria, sensitivity index. Therefore, comparison between the proposed methodology and EA in terms of the computational efficiency is difficult because both methods use the different assessment indices. In addition, the proposed multi-fixture layout optimization methodology can be used to eliminate the limitation of KCC tolerance optimization in improving the assembly process robustness. In some design problems, a tolerance optimization methodology might not be able to identify a set of KCC tolerances which can achieve the process yield requirement 关33兴. The framework which can integrate multiple design synthesis tasks 共e.g., multifixture layout optimization and KCC tolerance optimization兲 in optimizing assembly process design can be found in Phoomboplab and Ceglarek 关34兴. Additionally, the proposed methodology can be extended into fixture workspace synthesis for reconfigurable assembly 关35兴. 6 Conclusions This paper presents a methodology to improve a process yield by optimizing the locator positions in a multistation assembly system. The performance of fixture layouts is assessed by a process yield, which represents the robustness of fixturing system in terms of a final product dimensional quality. In addition, fixture locating stability is taken into consideration to ensure that the design of fixture layouts is feasible in practical. The variation sources in real industrial assembly processes, which are locator positions and part-to-part joint variations are also taken into consideration. The proposed methodology is based on two-step optimization, which integrates heuristic algorithm 共GA兲 with a low-discrepancy sampling technique 共HSS兲. The application of the proposed methodology is illustrated through a case study using an automotive underbody assembly where process yield greatly increases from 85% to 96% after optimizing the locator positions with no increase of tooling cost. DECEMBER 2008, Vol. 130 / 061005-15 Downloaded 10 Oct 2008 to 137.205.175.119. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm Acknowledgment n ⬅ nm · nm−1 . . . n2 · n1 · n0 = n0 + n1R + n2R2 + ¯ + nmRm The authors gratefully acknowledge the financial support of the UK EPSRC Star Award EP/E044506/1 and US NSF-CAREER Award DMII-0239244. The authors also appreciate the fruitful discussions with Professor W. Huang, University of Massachusetts, Professor James Kong, Oklahoma State University and Dr. Ying Zhou from Dimensional Control Systems, Inc. where R1 , R2 , . . . , Rk−1 are the first k − 1 prime numbers and m = 关logR n兴 = 关ln n / ln R兴 共the square brackets denote the integer part兲. A unique fraction between 0 and 1 called the inverse radix number can be constructed by reversing the order of the digits of p around the decimal point as follows: Nomenclature The Hammersley points on a k-dimensional cube are given by the following sequence: KPC KCC ri Li Si Mi ⌳i ⫽ ⫽ ⫽ ⫽ ⫽ ⫽ ⫽ s P4way ⫽ s P2way ⫽ NCsi ⫽ m P4way ⫽ m P2way ⫽ NCm i ⫽ Fi Rx,y,z M x,y,z Wg wd f m l⬘x,y and u⬘x,y ⫽ ⫽ ⫽ ⫽ ⫽ ⫽ ⫽ ⫽ tmg v 关A共⌳兲兴 LSL, USL Ri Oi ⫽ ⫽ ⫽ ⫽ ⫽ ⫽ ⫽ ⫽ Oi OGA z = f i共x , y兲 Ai Bi ⫽ ⫽ ⫽ ⫽ ⫽ key product characteristic key control characteristic vector defining the location of a locator fixture layout for a single station fixture layout for a root part fixture layout for a mating part a set of fixture layouts for a particular assembly system four-way pin in a fixture layout for a root part 共Si兲 two-way pin in a fixture layout for a root part 共Si兲 NC block in a fixture layout for a root part 共Si兲 four-way pin in a fixture layout for a mating part 共Mi兲 2-way pin in a fixture layout for a mating part 共Mi兲 NC block in a fixture layout for a mating part 共Mi兲 supporting and locating force of a locator the resultant force moment of resultant force weight of a part external wrench to balance locating stability force in external wrench wd moment in external wrench wd lower and upper boundary of x and y coordinated in design space of interest Ai twist of a workpiece due to weight angular displacement in twist tmg translation displacement in twist tmg instability index SOVA matrix lower and upper specification limits design space for a locator a set of nodes discretized from design space Ri a Cartesian coordinate of a discretized node a node selected by GA surface function of a workpiece design space of interest 2D projection of design space of interest 共Ai兲 Appendix: Hammersley Sequence Sampling HSS is used in Step 3 of the proposed methodology. The challenge in the local search step is high-dimensional design space, which usually requires a large number of iterations or samples. HSS is a sampling technique that selects samples uniformly in a hypercube design space, which requires fewer samples to converge to the solution within desired variance compared with other sampling technique or space filling technique such as Number– Theoretical Net 关31兴 or Latin hypercube 关32兴. Kalagnanam and Diwekar 关32兴 provided a procedure for selecting N Hammersley points in k-dimensional hypercube. Any integer n共n 苸 兵1 , 2 , . . . , N其兲 can be written in radix-R notation 共R is a prime number兲 as follows: 061005-16 / Vol. 130, DECEMBER 2008 R共n兲 = n0n1n2 ¯ nm = n0R−1 + n1R−2 + ¯ + nqR−m−1 zk共n兲 = 冉 冊 n , R1共n兲, R2共n兲, . . . , Rk−1共n兲 , N n = 1,2, . . . ,N The Hammersley points generated in a unit hypercube are xk共n兲 = 1 − zk共n兲 References 关1兴 Chou, Y.-C., Chandru, V., and Barash, M. M., 1989, “A Mathematical Approach to Automatic Configuration of Machining Fixtures: Analysis and Synthesis,” ASME J. Eng. Ind., 111共2兲, pp. 299–306. 关2兴 Shalon, D., Gossard, D., Ulrich, K., and Fitzpatrick, D., 1992, “Representing Geometric Variations in Complex Structural Assemblies on CAD Systems,” ASME Paper No. DE-44/2, pp. 121–132. 关3兴 Ceglarek, D., Huang, W., Zhou, S., Ding, Y., Kumar, R., Zhou, Y., 2004, “Time-based Competition in Manufacturing: Stream-of-Variation Analysis 共SOVA兲 Methodology-Review,” International Journal of Flexible Manufacturing Systems, 16共1兲, pp. 11–44. 关4兴 Ceglarek, D., and Shi, J., 1995, “Dimensional Variation Reduction for Automotive Body Assembly,” Manuf. Rev., 8共2兲, pp. 139–154. 关5兴 Jin, J., and Shi, J., 1999, “State Space Modeling of Sheet Metal Assembly for Dimensional Control,” ASME J. Manuf. Sci. Eng., 212, pp. 756–762. 关6兴 Ding, Y., Ceglarek, D., and Shi, J., 2000, “Modeling and Diagnosis of Multistage Manufacturing Process: Part I State Space Model,” Proceedings of 2000 Japan-USA Symposium of Flexible Automation, July 23–26, Ann Arbor, MI, Paper No. 2000JUSFA-13146. 关7兴 Camelio, J., Hu, S. J., and Ceglarek, D., 2003, “Modeling Variation Propagation of Multi-station Assembly Systems with Compliant Parts,” ASME J. Mech. Des., 125共4兲, pp. 673–681. 关8兴 Huang, W., Lin, J., Bezdecny, M. R., and Ceglarek, D., 2007, “Stream of Variation in 3D Assembly I: Generalized Rigid Body Model for Single Station,” ASME J. Manuf. Sci. Eng., 129共4兲, pp. 821–831. 关9兴 Huang, W., Lin, J., Zhou, S., and Ceglarek, 2007, “Stream-of-Variation 共SOVA兲 Modeling II: A Generic 3D Variation Model for Rigid Body Assembly in Multi Station Assembly Processes,” ASME J. Manuf. Sci. Eng., 129共4兲, pp. 832–842. 关10兴 Shiu, B. W., Ceglarek, D., and Shi, J., 1996, “Multi-stations Sheet Metal Assembly Modeling and Diagnostics,” Trans. NAMRI/SME, 24, pp. 199–204. 关11兴 DeMeter, E. C., 1994, “Restraint Analysis of Fixtures Which Rely on SurfaceContact,” ASME J. Eng. Ind., 116共2兲, pp. 207–215. 关12兴 Rong, Y., and Zhu, Y., 1999, Computer-Aided Fixture Design, Dekker, Dordrecht. 关13兴 Rong, Y., and Bai, Y., 1996, “Machining Accuracy Analysis for ComputerAided Fixture Design Verification,” ASME J. Manuf. Sci. Eng., 118共3兲, pp. 289–300. 关14兴 Roy, U., and Liao, J., 2002, “Fixturing Analysis for Stability Consideration in an Automated Fixture Design System,” ASME J. Manuf. Sci. Eng., 124, pp. 98–104. 关15兴 Wang, M. Y., 2000, “An Optimum Design for 3-D Fixture Synthesis in a Point Set Domain,” IEEE Trans. Rob. Autom., 16共6兲, pp. 839–846. 关16兴 Cai, W., Hu, S. J., and Yuan, J. X., 1997, “A Variational Method of Robust Fixture Configuration Design for 3-D Workpieces,” ASME J. Manuf. Sci. Eng., 119, pp. 593–602. 关17兴 Wang, M. Y., and Pelinescu, D. M., 2001, “Optimizing Fixture Layout in a Point-Set Domain,” IEEE Trans. Rob. Autom., 17共3兲, pp. 312–323. 关18兴 Carlson, J. S., 2001, “Quadratic Sensitivity Analysis of Fixtures and Locating Schemes for Rigid Parts,” ASME J. Manuf. Sci. Eng., 123, pp. 462–472. 关19兴 Menassa, R. J., and DeVries, W. R., 1991, “Optimization Methods Applied to Selecting Support Positions in Fixture Design,” ASME J. Eng. Ind., 113, pp. 412–418. 关20兴 Cai, W., Hu, S. J., and Yuan, J. X., 1996, “Deformable Sheet Metal Fixturing: Principles, Algorithms, and Simulations,” ASME J. Manuf. Sci. Eng., 118共3兲, pp. 318–324. 关21兴 Krishnakumar, K., and Melkote, S. N., 2000, “Machining Fixture Layout Optimization using the Genetic Algorithm,” Int. J. Mach. Tools Manuf., 40共4兲, pp. 579–598. 关22兴 Li, B., and Melkote, S. N., 2001, “Optimal Fixture Design Accounting for the Effect of Workpiece Dynamics,” Int. J. Adv. Manuf. Technol., 18, pp. 701– 707. 关23兴 De Meter, E. C., 1995, “Min-Max Load Model for Optimizing Machining Transactions of the ASME Downloaded 10 Oct 2008 to 137.205.175.119. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm Fixture Performance,” ASME J. Eng. Ind., 117, pp. 186–193. 关24兴 Marin, R. A., and Ferreira, P. M., 2001, “Kinematic Analysis and Synthesis of Deterministic 3-2-1 Locator Schemes for Machining Fixtures,” ASME J. Manuf. Sci. Eng., 123, pp. 708–719. 关25兴 Camelio, J. A., Hu, S. J., and Ceglarek, D., 2004, “Impact of Fixture Design on Sheet Metal Assembly Variation,” J. Manuf. Syst., 23共3兲, pp. 182–193. 关26兴 Kim, P., and Ding, Y., 2004, “Optimal Design of Fixture Layout in Multistation Assembly Processes,” IEEE. Trans. Autom. Sci. Eng., 1共2兲, pp. 133–145. 关27兴 Izquierdo, E. L., Du, H., Hu, S. J., Jin, R., Shi, R. J., and Jee, H., 2006, “Robust Fixture Layout Design for a Product Family Assembled in a Multistage Reconfigurable Line,” Proceedings of 2006 ASME International Conference on Manufacturing Sciences and Engineering, October 8–11, Ypsilanti, MI. 关28兴 Zhou, S., Huang, Q., and Shi, J., 2003, “State-Space Modeling for Dimensional Monitoring of Multistage Machining Process using Differential Motion Vector,” IEEE Trans. Rob. Autom., 19, pp. 296–308. 关29兴 Cook, R. D., and Nachtscheim, C. J., 1980, “A Comparison for Constructing Journal of Manufacturing Science and Engineering Exact D-Optimal Designs,” Technometrics, 22共3兲, pp. 315–324. 关30兴 Taam, W., Subbaiah, P., and Liddy, J. W., 1993, “A Note on Multivariate Capability Indices,” Journal of Applied Statistics, 20共3兲, pp. 339–351. 关31兴 Huang, W., Ceglarek, D., and Zhou, Z., 2004, “Tolerance Analysis for Design of Multistage Manufacturing Processes Using Number-Theoretical Net Method 共NT-net兲,” International Journal of Flexible Manufacturing Systems, 16, pp. 65–90. 关32兴 Kalagnanam, J. R., and Diwekar, U. M., 1997, “An Efficient Sampling Technique for Off-Line Quality Control,” Technometrics, 3共3兲, pp. 308–319. 关33兴 Huang, W., Phoomboplab, T., and Ceglarek, D., 2008, “Process Capability Surrogate Model Based Tolerance Synthesis of Multistation Manufacturing System,” IIE Trans., 40共6兲. 关34兴 Phoomboplab, T., and Ceglarek, D., 2007, “Design Synthesis Framework for Dimensional Management in Multistage Assembly System,” CIRP Ann., 56共1兲, pp. 153–158. 关35兴 Kong, Z., and Ceglarek, D., 2006, “Fixture Workshop Synthesis for Reconfigurable Assembly Systems,” J. Manuf. Syst., 25共1兲, pp. 25–38. DECEMBER 2008, Vol. 130 / 061005-17 Downloaded 10 Oct 2008 to 137.205.175.119. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm