Abstract
Fixture design is an important issue in the process of manufacturing. As a critical design activity process, automation in fixture design plays an integral role in linking computer aided designs and computer aided manufacturing. This paper carries out a literature review of computer aided fixture design (CAFD) developments using intelligent methods that have been commonly utilized in automation in the last two decades. The first part of this review considers the steps of fixture design along with the significant researches and requirements of fixtures over time. After that, the paper presents important and relevant research carried out in the field of CAFD using intelligent approaches and the working principles surrounding this issue. The following section concentrates on the details of case based reasoning (CBR) approach, the most successful approach in CAFD. The examination of this approach is carried out based on applications, stages of CBR based systems fixture design, working principles, and pertinent proposed approaches. Lastly, the present drawbacks of the current methods and shortfalls in research are identified for future studies.
1. Introduction
Fixtures are a type of vital tooling which are used during manufacturing, assembly, inspection, and other operations; they set and secure the orientation and position required by workpieces during operations based on the specifications of the design [1]. An accurate fixture design is critical for quality of product based on accuracy, precision, and product of the designed parts. Studies show that about 40% of parts that are rejected are mainly because of dimensioning errors due to poor fixture designs [2]. The designs are influenced by the workpiece, machining techniques, and performances of material, which are known as the fixture requirements. The designer has to be very experienced in order to significantly reduce the design costs and cycle time. A good design means lesser need for rework of parts which in turn also helps to control cost and time in the process of manufacturing a product [3]. In addition, costs related fixturing represents 10–20% of the total costs of a system in manufacturing [4].
Computerization and automation of fixture designs are required to lower the cost and lead time of product development in the field of fixture design. Thus, CAFD have been introduced and utilized as an integration of computer aided designs and manufacture (CAD/CAM) [5]. In addition, several artificial intelligence methods have also been introduced which enhances the extent of automation in the CAFD techniques. Several techniques have been utilized in the design of fixtures, namely, CBR, genetic algorithm (GA), rule based expert system, multiagent approach, geometric analysis, machine learning, and others [6]. However, fixture design is still an issue of concern in the present manufacturing field [2]. Thus, it is essential to identify and analyze all the intelligent methods as the main techniques in these areas in terms of their applications, strengths, and working principles based on the criteria of the fixture design. This would create a better understanding of the intelligent approaches as they can be easily compared and evaluated, which would assist in overcoming the current challenges found in CAFD.
Fixture designs are normally dependent on the knowledge and experience of the engineers working on the fixture design and this is one of the main factors for fixture designs being an unresolved matter in this field. On the contrary, CBR utilizes knowledge and experience to overcome new challenges. Furthermore, a lot of good design cases and abundance of technical knowledge are available readily in manufacturing companies which can be utilized as knowledge and experience in using the CBR based systems for fixture design. This is why CBR is one of the most effective techniques in this sector. Similarity-based matching is performed by CBR to search for the most similar cases in database when facing the new challenge. Accordingly, CBR can perform the automated fixture design process properly and efficiently. On the other hand, the industrial demand for intelligent and automated technical methods for fixture design is quite high. This is the key reason why the process of CBR based systems and review of CBR methods in fixture design are clarified in this study.
This paper is organized as follows: in first part of the paper (Section 2), the efforts that have been done in CAFD by focusing on steps of fixture design based on design requirements and intelligent techniques that have been used are discussed. The second part of paper (Section 3) is devoted to CBR technique in fixture design by considering description of CBR and applications, existing CBR base fixture design methods, and steps of CBR based fixture design systems. The final section contains some conclusions about gaps as well as some future trends in fixture design field and specifically in CBR based methods. The value of this paper is that it provides an in-depth critique review for easy understanding based on current intelligent approaches and also the focus on steps fixture design over time in CAFD. Another contribution is that it critically studies all existing significant CBR based methods and also how CBR based fixture design methods can support the design process in details.
2. Computer Aided Fixture Design
Computer aided fixture design (CAFD) is the use of computers to help aid in the design of fixtures [7]. Computer aided fixture designs (CAFD) have grown tremendously since the 1980s, and a lot of work has been done to enhance the process of fixture designs [8].
Computers have dramatically reduced the design process time. By using a computer, designers are able to design in a virtual atmosphere. This helps the designers identify potential problems and undertake different ideas without actually physically creating the fixture. These programs have the added benefit of keeping a designer from missing steps while designing, and by avoiding mistakes time and costs can be kept low [7].
During fixture design process, several design requirements should be considered at the same time for each step of fixture design. The generic requirements are such as physical requirements (PR), collision prevention (CP), design constraining (DC), usability requirements (UR), affordability requirements (AR), and tolerance requirements (TR). PR is that the fixture must be physically capable of accommodating the workpiece geometry and weight. CP is to avoid colliding tool path-fixture, workpiece-fixture, and fixture-fixture. DC is such as workpiece stability, fixture stiffness, and workpiece stiffness. For UR the fixture must meet some requirements such as weight, being as tool guidance, facilitating chip shedding, and avoiding surface damage. In case of AR, fixture must be designed considering cost, assembly, and operation time limitations. TR is such that the fixture locating tolerances should be sufficient to satisfy part design tolerances.
The conflicting nature of these requirements is problematic. For example, a heavy fixture can be advantageous in terms of stability but can adversely affect cost (due to increased material costs) and usability (because the increased weight may hinder manual handling) [9]. Such conflicts add to the complexity of fixture design and contribute to the need for intelligent techniques in CAFD. Moreover, considering fixture design requirements in each step of the design could help clarify fixture design problems and accordingly facilitate solving them.
2.1. Steps of Fixture Design
Fixtures are generally composed of four components, that is, locators, clamp, support, and base plate. Commonly, fixture design is how to place the components onto a workpiece in order to achieve immobility and stability of the workpiece during machining or other operations [10].
Normally, there are four steps in the design process of fixture design: setup planning, fixture planning, unit design, and verification, which are shown in Figure 1 [11]. Setup planning could be defined as the identification of machining setup by not moving the parts in each setup. Fixture planning is definition of a fixture requirements physical, collision prevention, and layout plan. Unit design is the conceptual and detailed definition of the locating and clamping units of a fixture composed with the base plate. Verification ensures that developed fixture designs satisfy the fixturing requirements. Moreover, the design has to meet other design concerns which may contain fixture cost, fixture weight, assembly time, and so on [12].
The steps of fixture design.
These steps can be generalized as analyze the part, define suitable locating and clamping points, identify tooling and environmental requirements, and create a fixture to satisfy criteria [7]. For instance, during the fixture planning step, fixturing requirements are typically generated which can be grouped into six classes as mentioned above [13]. Fixture planning focuses on determining precise location and clamping of workpieces according to a part design and process requirements. This step is a largely experience driven activity and the output of this step is an automatically configured fixture layout that guarantees part stability [14].
Unit design step is to achieve location accuracy and stiffness and avoid fixture components deformation. As an example, the tolerances of locating units should be in a way whereby the accuracy of the location can be reached, and deformation as a result of the clamping forces and machining should be avoided. The specification of clamping and locating units’ design must meet these essential requirements [10].
Verification of a fixture design is the method of verifying the current fixture design through means of analyzing the ability of the geometric constraints, tolerance achieved, deformation, and fixture-workpiece system stability. It also involves providing relevant suggestions for the design improvement [11, 13, 15]. Verifying a fixture design must be done due to these arguments: (1) the design process involves many factors and thus it is not easy to implement proper models of analysis of the process; (2) the constraints of design are individually considered; some constraints that are contradictory could appear as a result of being regarded together; and (3) the design of the fixture is closely linked to other activities in a manufacturing system such as CAM and computer aided process planning (CAPP) [2].
In machining fixtures, minimizing workpiece deformation due to clamping and cutting forces is essential to maintain the machining accuracy [16]. Thus, the CAFD research strengths are mainly found in the methods of verification which concentrated on examining the stability of the workpiece and deformation in the process of machining and the fixture planning methods that aim to minimize the deformation of the workpiece as a result of forces during machining [9]. In Table 1 significant researches that have been done in the past two decades are considered which proves that the focus of CAFD has been on fixture planning and verification steps. The table is constructed based on the date of each work, fixture design step, and fixture design requirements. Finally the percentage of each step in each period to clarify the flow of fixture design by the time is calculated.
Significant current fixture design approaches.
Note: fixture design steps (SP: setup planning, FP: fixture planning, UD: unit design, and VR: verification); fixture design requirement: FR (PR: physical requirements, CP: collision prevention, DC: design constraining, UR: usability requirements, AR: affordability requirements, and TR: tolerance requirements).
2.2. Intelligent Approaches to Fixture Design Systems
Intelligent methods also known as artificial intelligence methods simulate the processes that a human undergoes when reasoning through a problem [7]. On the other hand, the fixture design process is a highly subjective process, that is, based on heuristic knowledge of tool designers [17]. Moreover, many design requirements are required to be taken into account at the same time. Therefore, researches have been focused on intelligent methods to improve performance of computer aided fixture design systems.
The efforts over past decades in fixture design field have resulted in numerous computer aided fixture design (CAFD) applications using various intelligent methods such as expert systems, case based reasoning (CBR), genetic algorithm (GA), and artificial neural network (ANN). In Table 2, the important intelligent techniques in fixture design with their procedures, applications, assumptions, and weaknesses are considered and in the next sections some typical fixture design approaches, which are based on these techniques, are studied. Finally, CBR as the most efficient technique would be explained in detail at the last section.
The important intelligent techniques in CAFD.
2.2.1. GA, ACA, and FEM
GA and ACA are evolutionary algorithms, which are often used to solve optimization problems, with GA as the more popular technique. Goldberg and Holland [18], Subramaniam et al. [19], and Kumar et al. [20] have already proved that GA is a powerful technique in solving optimization problems in engineering. Applying GAs in support of fixture planning, layout plan, and fixture configuration design solutions have been successfully used.
In case of fixture layout optimization, GA is typically used to determine the optimal clamping condition for an elastic workpiece by Krishnakumar and Melkote [21]. A set of points of contacts are encoded and tested. The GA is utilized to create the new contact point sets till an optimum is derived which minimizes deformation of the workpiece as a result of clamping forces and machining [21]. In contrast, for a rigid fixture-workpiece, Wu and Chan [22] optimized the fixture layout model using GA. They ignored elastic deformation of the workpiece due to clamping and machining forces. Yeung and Chen [23] also optimized the fixture layout for a 3D component using GA. In case of deformation, Krishnakumar and Melkote [21] developed a GA-based fixture layout optimization technique to find the fixture layout that minimizes the deformation of the machined surface due to static machining forces. In this work, the fixture layout is optimized for a single node system, which may not be sufficient to find the minimum elastic deformation of the workpiece.
To overcome finding the minimum elastic deformation, GA has been interfaced with FEM for the fixture layout optimization problems [16]. Finite element analysis (FEA) is becoming a common platform for the modeling and simulation of various manufacturing processes [24]. Typically deformation testing is employed using a FEA in which a workpiece is discretized to create a series of nodes that represent potential locating and clamping contact points [25]. As an example, Kulankara et al. [26] regarded the fixturing elements as rigid and the workpiece to be elastic and then utilized the FEM to simulate the machining operation. The forces of machining are regarded as point forces acting over the tool path. Static analysis is also considered to determine the workpiece deformation. To develop that method, Chen et al. [27] established a multiobjective optimization model to minimize the deformation and improve the uniform distribution of deformation by considering the friction and chip removal effects. The optimization process was performed through the integration of GA and FEM. FEM was used to calculate the machining deformation for various clamping force and cutting force under a specific fixture layout. GA was to optimize the fixture design used, the fixture layout, and the clamping force as design variables. Prabhaharan et al. [28] also optimized the fixture layout to minimize the dimensional and form errors. FEM was used to predict the workpiece deformation.
The analysis of deformation analysis has primarily focused on examining the deformation of workpieces using the finite element analysis [24, 25, 29]. In general, this includes the act of discretizing the workpiece into elements that form a mesh, choosing the kind of analytical elements to represent the mesh in the analysis, and defining the boundary conditions that are present on the workpiece such as at the interface of the workpiece/fixture [30].
In some certain fields of fixture design, FEM and GA are utilized such as in the generation of an optimal fixture configuration layout [31]. The layouts identify optimum positions at the point where the fixture should have contact with the workpiece, that is, undergoing machining. These optimal positions are achieved by performing these three steps. Firstly, the usage of the process analysis method of machining to estimate the machining force, that is, used on the workpiece, is considered [2]. Secondly, a deformation analysis of the system of fixture-workpiece is carried out by using the predetermined load cases. Deformation analysis typically utilizes the FEM approach. The last step involves employing a process of optimization (GA is commonly utilized) to look for a possible solution space and decide on the fixtures locations within an acceptable candidate region with minimized deformation to the workpiece [32]. Many relevant works assume some conditions such as regarding treat fixture-workpiece contact as point contact; the workpiece is deformable while the tools and fixtures are rigid. Thus, a limitation of these approaches is that they normally offer a list of coordinates identifying the location of the fixture, without offering the fixture unit's actual physical form [21, 26, 33].
GA and ACA were adopted separately for the optimization method. The performances of GA and ACA were tested and compared based on different node systems as the workpiece deformation varies according to the node system. They defined three different numbers of node systems on the same geometry to evaluate the performance of GA and ACA. The minimum workpiece deformation for the possible layout, that is, optimal solution, was predicted for all the three node systems using both GA and ACA. A comparison was made between both the algorithms on the basis of minimum objective function value, and it was found that ACA solutions are better than those of GA [28]. By applying ACA, Padmanaban et al. [34] recently optimized the machining fixture layout to minimize the workpiece elastic deformation discrete-based and continuous-based optimization methods. The dynamic response of the workpiece with respect to machining and clamping forces was determined using FEM. From the results of the case study, it is concluded that the ACA is suitable for problems where no direct relation exists between the objective function values and the constraints [35].
2.2.2. ANN, Machine Learning, and GA
Neural networks represent simple elements of interconnected networks, where the interconnections are studied from a range of sample data. After the learning, these networks can create solutions for new challenges that are fed into the networks. Ming and Mak [36] used a neural network method to prove the applicability of the NN in the setup planning, where tool approach direction, feature precedence, and tolerance relations are fed into a self-organizing Kohonen neural network to gather the operations for singular features into the setup planning. In case of fixture layout design, recently a NN-based algorithm with design of experiments is also proposed by Selvakumar et al. The approach is to design an optimum fixture layout to lower the workpiece maximum elastic deformation created while machining by the machining and clamping forces action on the workpiece [16].
NN methods have also been employed for the purpose of conceptual unit design support. In a study by Kumar et al. [20], they utilized a combination of GA/NN methods where the neural network is coached with a choice of past design issues and the solutions. The GA creates potential solutions, which are assessed utilizing the NN and after that these assessed solutions are used to guide the GA.
Kumar et al. [37] applied machine learning to achieve automatic fixture concept design; however, the result was not very remarkable. They implemented fixture design knowledge with machine learning technology such that fixture design decision tree and fixture classification rules are built. The composition fixture unit model is also classified with machine learning algorithm, and design knowledge description is formed. Lastly, decision-making capacity of fixture design is elevated by taking advantage of genetic algorithm and fuzzy logic [37].
2.2.3. RBR Based Systems
Since artificial intelligent (AI) has obtained great acceptance as a technology for manufacturing, expert systems can provide a logical tool for automating the fixture design by using the advantages of management of experience-based knowledge [17]. RBR is one of the methods of the experts system and utilizes induction rules to decide if a new challenge should be further inspected [8].
When CAFD was still new in the field, Nnaji and Alladin [38] and Nnaji et al. [39] introduced an expert system that was rule based for fixturing on a CAD system by using a flexible fixture, based on the understanding of the geometry of workpiece, machine tool, and machining operations. The users interactively supply the machining and workpiece information, and the fixturing rules in the system database are assessed for potential execution at each point. System outputs are the clamping and locating positions, a list of the components to be utilized for every fixture unit, and verification of stability of the workpiece. Nee and Kumar [40] also developed an automated rule based fixture design. Besides the functionality provided by Nnaji, Nee and Kumar carried out a limited check on the likely displacements at each locating point due to the machining forces and also executed a simplified justification module that used heuristic rules to decide if a dedicated or modular fixture design should be created.
Besides using heuristic rules for creating the conceptual designs, Kumar et al. [37] utilized an inductive reasoning approach to develop decision trees from where the fixturing rules can be retrieved by assessing of each path of the decision tree. Ma et al. [41] and Hunter et al. [42] developed a set of fixture automatic layout design system that is able to automatically reason out positioning and workpiece-clamping position based on workpiece geometry and machining information.
2.2.4. Geometry Based Approaches and Information Representation Technologies
Geometry is an important aspect in the individual units design where the aim is to choose, assemble the defined elements of the unit, and specially offer the suitable unit of acting height [43, 44]. Geometric approaches have been mainly used in attempts to design a complete fixture unit [43] in which the fundamental concept is to recognize the critical dimensions of a specific fixture unit. The critical dimension is usually the units’ height, which then relates to all the other dimensions of the component via a preexisting mathematical relationship. As an example, a geometry-based system is developed where the dimensions of the individual component are created in association with the element's primary dimension (normally the needed height) via parametric dimension relations [45]. This is augmented with an association relationship knowledge base of how various elements could be configured to develop a single unit. A recent study by Peng et al. [46] utilized the reasoning of geometric constraint to facilitate the assembly of elements selected by users to develop individual units for an approach that was more interactive.
2.2.5. Summary
There are several areas of concern in the present status of the CAFD research based on the review, namely, the following.
According to the importance of fixture planning (layout plan and fixture configuration) step, much attention has been paid to search and optimization of those stages but research on knowledge-driven in this field of fixture design still needs deepening.
FEM can be embedded with optimization techniques along with other intelligent methods in order to improve applicability of future proposed method in case of verification.
More focus is needed to support the unit design, specifically relating to deciding the unit stiffness and the related requirements for unit stiffness to the unit structure.
Most of the CAFD techniques have been examined for simplified workpieces that are not reflective of those experienced in industry; hence, the efficacy of the developed methods cannot be represented confidently.
3. Case Based Reasoning
CBR is an artificial intelligence technique which is a general problem solving method using past experiences to solve novel problem [47]. CBR allows a user to learn from previous mistakes by keeping them stored and easily available. Users will also be more accepting of solutions proposed by CBR systems because the proof is completely visible in the previous case. According to Aamodt and Plaza's [48] definition, formal model of the problem solving cycle in CBR has four main modules. The descriptions of each step are as follows.
Retrieve: by using a problem, retrieve relevant cases from database for solving it.
Reuse: draw the answer from the earlier case to the faced problem. It may include adjusting the answer as required toward fitting the new condition.
Revise: making plan for the earlier case to the object condition, exam the created answer in the valid space (like a simulation) and then, if needed, make revision.
Retain: when the answer is well adapted to the faced problem, store the practicing problem as a new case in database.
3.1. CBR Based Systems and Applications
CBR [49, 50] consists of indexing, representing, and organizing the previous design cases in a case library so that they can be recalled, modified, and reused for future design scenarios. Fundamentally, when encountering a new design problem, the CBR technique identifies a prior case, which best matches the present requirement of design. Then, it modifies the case to offer a solution that is satisfactory to the new design situation [10]. Key considerations to design a CBR technique are stated as follows [51]:
design case representation;
index and retrieve the similar cases from the database;
adapt the retrieved cases for the present design.
In general, CBR has been proven beneficial in many applications. Reviewing the type of problem solving used in CBR shows that similar solutions are provided for similar problems. It has been proven to work in simple scenarios and is validated empirically in many domains in the real world [52–54]. Typically, Cardone et al. [55] proposed a CBR based method for assessing the cost of machining for newly design prismatic parts. The method was similarity-based so that the similar part in database to the target part is retrieved to predict the cost of machining for the newly design part. The similarity in that method is measured by using the machining features on those parts. In addition, by using this method for layout planning, a workpiece layout plan is gained by retrieving the plan utilized for similar workpieces from the case library consisting of knowledge from prior workpieces and their layout plans [6, 56]. In a study by Marefat and Britanik [57], they presented an object-oriented model for 3D prismatic parts in case based process planning. The model contained the feature-based parts information and the process plan information model and knowledge of tools knowledge dealing with the process plan.
The design of fixture is often heavily dependent on the knowledge and experience of engineers working on the fixture design, which can cause changes in the fixture design's quality. Meanwhile, the knowledge of the fixture design and resources are required to be suitably stored for reuse in the future [1]. For these reasons, CBR approach is one of the most effective methods, which can organize previous experiences as cases to solve new problems applied to this field.
3.2. Steps of CBR in Fixture Design Systems
A CBR based computer aided fixture design system, which acts as the evaluator and advisor, has been introduced to assist the fixture designer in finding and reusing appropriate cases that have been successful. The general CBR framework in fixture design proposed by Peng et al. [8] is modified here and shown in Figure 2. This framework still makes use of the general four steps as outlined earlier. Firstly, it emphasizes on structuralizing cases in the fixture design and focuses on some important data, which will be available while calculating the similarities. After this, the final solution will be produced by modifying the most similar problem by comparing the cases [2]. The stages of CBR in fixture design are addressed as follows [6].
Indexing is used to identify the aspects of prior solutions that are related to the present problem. Hence, indexing the challenge involves defining a comprehensive approach to index the related design information based on the fixture design criteria [8].
Retrieval is used to identify cases that have some or all of the needed attributes; a main process of CBR is the case retrieval. The quality of retrieved case from the case base has an effect on the general CBR based application's performance.
Adaptation identifies the differences that exist between the new problem and the selected design that requires a design solution. After the required changes are noted, they are used to make the changes.
One of main concerns in CBR is the representation technique of the fixture design. Representation is often critical to the success and feasibility of an intelligent method that is proposed.
The modified framework of a CBR based system.
3.2.1. Fixture Indexing and Case Library Module
Indexing is used to identify the features of prior solutions that are related to the present problem. Hence, indexing the challenge involves defining a comprehensive approach to index the related design information based on the fixture design cases [6].
The task of identifying appropriate indexes in CBR is a difficulty. A particular difficulty with indexing relates to identifying the concepts that can be used to describe an item. In fact, an important issue in indexing of CBR based systems is the inability of the method to differentiate between two cases whereby two cases may share similar values for all the attributes of indexing, but it is not likely that both design cases will be suited equally to the present requirements of the design [58, 59]. A CBR system itself may be unable to help the user make a choice because as far as the system is concerned the designs are identical. This condition, known as “inseparability,” is a result of inadequate indexing [10].
Inseparability is an important issue in indexing [58]. Inseparability occurs due to having too few attributes in indexing or by choosing poor attribute choices such as many designs that may have similar attributes and values. A commonly used method to index is defining the attributes linked to the design problems. Nee et al. [60], for example, indexed design cases by using attributes which were used to describe the workpiece in which a fixture will be designed, such as interfeature relations and machining features.
The usual methods on case indexing are based on an attribute set. Chen and Liu [61], for example, utilized a case template to structure the fixture design cases, while Liqing and Kumar [51] classified the data for the fixture design into three groups of attributes, namely, part, setup, and fixture representation. However, there are few ways of defining and selecting the suitable attributes. Boyle [10, 62] introduced a methodology to divide the information of fixture design into two case libraries: a fixture design solution base that retains detailed solutions for fixture design and a fixture unit depository that has the individual component (units) and maintains an internal relation with the design base. Furthermore, Wang and Rong [6] reorganized fixture design information into the 3-level case library, such as the workpiece information, fixturing plan, and fixture unit, which is used so that detailed data/information of a design solution is organized as a relationship model of abstract objects with a hierarchical structure in one case base. These objects are stored independently in the case base while having close internal mapping relations among one another.
In order to reuse and design a new fixture using CBR systems, two indexing methods can be used, similarity-based retrieval and adaptability-based, which affect highly on indexing the cases in the case library. Adaptability-based retrieval is an expensive method computationally due to two factors. First, it needs the CBR system to decide what changes are required to fix a case and then to determine how to achieve this change as there may be various ways of affecting this change and to examine how the change will affect the other aspects of the design. This is rather complicated compared to simply examining the attribute similarity, as it needs the predictive ability to decide the efficacy of the decisions for the design. Secondly, the method may need to navigate through a very large search space, which may lead to the problems under control. Therefore, some approaches that are able to initially constrain the search space must be determined. Therefore, it is essential to maintain the library small in the adaptability-based retrieval, which is in contrary to similarity-based retrieval where a large quantity of similar design cases is preferred to increase the chances of reaching an appropriate solution [62]. Nevertheless, an important limitation in the similarity-based method is that there is mostly an assumed linear relationship between the worth of the performance level of a feature and the feature's performance level to the designer.
With regard to CBR, indexing of design cases is a problematical issue. No formal method for defining case indexes exists and it is generally left to the experience of the designer to determine what indexes should be used. A poor understanding or definition of the design requirement is directly related to inadequate indexing. If the requirements of a fixture design problem can be adequately formalized, then it seems likely that these requirements can be used, if not directly then at the least as a guide, to determine the indexing attributes. Indexing must adequately represent the information contained within a case and must be easy to use and search for retrieval processes.
3.2.2. Case Similarity Measurement for Retrieval Process
Retrieval begins when the cases that have some or all of the features required are identified and ranked based on similarity. The closest approach, the knowledge guide approach, and template search approach are significant and commonly used searching algorithms in the retrieval process. These approaches can be utilized on their own or combined together as hybrid retrieval strategies [63]. The main consideration in the retrieval process of the past cases in the case base is the similarity. The similarity measure is a function that assesses the similarity between cases in the case base and the given query. It calculates the differences of each dimension or attribute between an existing case and the new query [60, 64].
The CBR method by Nee et al. [60] assumed that similar workpieces could be used for similar workpieces. Therefore, the workpieces were grouped into part families and a suitable fixture design was attached to each workpiece. The fixture structures similarity relies on the similarity of the parts structure and the process that fixtures are used for [1]. Now, this automatically appears to be a sensible and valid approach.
In CBR method various retrieval methods have been developed. These changes depend on the knowledge accessed into fixture design, which leads to efficient access to existing design case [3]. However, there have been difficulties to measure similarity between cases in the case base because of many attributes that need to be considered [6, 60]. Accordingly, workpiece similarity is typically characterized through indexing their part family classification, workpiece geometry, workpiece dimensions, machining features, tolerances, and so on [65]. Traditionally, cases are mostly indexed by means of attribute value pairs and then retrieving process was like a query in database [60]. Currently, some distance metric methods have been employed such as cosine formula [6, 7] and chebyshev [8] and also adaptability-based method [62] which uses utility curves for each fixture unit to navigate the proper case in database.
3.2.3. Adaption and Modification Step
Case adaptation is another bottleneck of CBR. Too much effort would be required to calculate the feasibility of each fixture and then decide on the required modifications for the infeasible fixtures, particularly for the complex similarity measure algorithms and large amount of cases [60]. That goes beyond the simple verification functions that are able to judge a design's quality. The case adaptation should be decided by the users as suggested by most experts in this field [8, 60, 66].
Nevertheless, many efforts have been done to link other parts of the CBR process to the adaptation more closely [67, 68]. For instance, Smyth and Keane [67] method was directly based retrieval on the likelihood of adaptability. They decided on the required adaptation strategies needed to fix a particular case and retrieved the most preferable adaptation strategy for the design. Wang and Rong [6] used an adaption strategy in their CBR based system. The system is able to determine differences between workpiece and fixture components dimensions and then suggest the designer to modify the design. Zhou et al. applied a deductive reasoning method by use of rule matching to facilitate case adaptation [1].
In reality, measuring the efficacy of a design is rather complicated as this type of problem mainly depended on the experience of the designer. However, several methods of verification have been developed to assist the designers to assess the performance of the design. These verification methods involve checking of stability and stiffness, analysis of the degree of freedom, analysis of the loading/unloading of fixtures, and so forth [6].
3.2.4. Case Representation
Case representation is generally regarded as an important issue and is crucial to success of case based reasoning system. Case representation deals with two research areas in the distributed system of fixture design. Firstly, as a basic aspect of the CBR paradigm, case is described as a problem, solution, and also the outcome in the domain of fixture design and is represented in a way, which enables an efficient retrieval, easy maintenance, and transmission over a network. Secondly, a case representation can be designed to be open standard in the domain of fixture design which allows information exchange with other systems in computer aided manufacturing [51].
Essentially, developing a methodology for fixture design requires the clarification of critical issues, namely, how fixture design knowledge could be represented in a computer and how the problem solving process can be implemented [2]. Wang and Rong [6, 51] provide an indexing scheme using XML formatting. It is a language that is able to be sent over the internet easily and is very popular when creating online web applications [7]. XML schema is adopted to represent cases mainly because it can assist in case retrieval, case storage, and network transmission over the Internet. It is also due to the fact that it can be viewed as an open standard to exchange information on manufacturing with other CAD/CAM systems [51].
In case of case storage and representation, there are some methods to construct database such as feature-based technique, parametric based modeling, geometric based modeling, and assembly based modeling. Feature-based modeling refers to the construction of geometries as a combination of form features. There are dual properties in features that explain the function and geometry of the design [1]. Due to these properties, features have become popular in the use of integrating geometry with semantics of engineering. Parametric modeling refers to a solid model that has constructed features and sketches by a set of mathematical equations to describe a workpiece [69]. Kumar et al. [70] applied a parametric modeling technique to construct the database. In that method, the input data is mostly manual and the relation between dimensions and features should be constructed by equations; however, changes in dimensions can be made easily. In contrast, Zhou et al. [1] used a model with feature-based information to explain the part general and geometric information and manufacturing process information. The model of feature-based part information involves the feature information in general and information of the machining features. He could integrate the fixture design experience and knowledge with part geometry and enhance the level of automation of fixture design by utilizing this technique.
3.3. CBR Based Fixture Design Approaches
In the past decade at the initial time of applying CBR in fixture design field, Nee et al. [60] proposed the significant CBR based approach which is still used as the basis of CBR based fixture design systems. They proposed a framework of a case based fixture design system so that past design cases could be indexed by using the attributes that described the workpiece for which a fixture was to be designed (this approach would be explained in detail in the next sections). Then, Sun and Chen [71] proposed another method, saying that fixture design cases are stored in the database, from which similar cases meeting the design requirements are retrieved according to functional features; thus, a new fixture design is obtained by modifying the existing case. The weakness of this approach is when growing the number of fixture in database, and then the retrieving process would take long time because of volume of computation. In the next proposed method, Lin and Huang [72] combined a NN with approach in which fixturing problems are coded in terms of their geometrical structure and the NN used to find similar workpieces and their unit designs. Similarly, Subramaniam et al. [73] also proposed a CBR system that uses a GA for searching the indexed database. From the search results the system identifies the best result and rates the choices in order for the designer to select which features they would like to reproduce.
According to the request for facilitating usage of new computer based methods in flexible manufacturing and also advances in computer technologies, some researchers at the early of current decade applied representation technologies especially in CBR based fixture design systems. The representation methods such as extensible markup language (XML) and virtual reality (VR) are to support fixture knowledge representation [8, 46]. The XML technology is utilized as a representation tool for fixture knowledge, even though the virtual reality is used in the systemization of intelligent methods in the design of the fixture [6, 51]. Typically, Chen and Liu [61] utilized a case template to structure the fixture design cases, and Liqing and Kumar [51] introduced the XML-based CBR technique which divides the data of the fixture design into three groups with attributes of parts, setup, and fixture representation. Nevertheless, there are few ways of defining and selecting the suitable attributes. This mainly depends on the designer's experience and knowledge. The designer chooses several significant attributes, which he considers essential to the present design case and uses these attributes to compute and match the stored cases. Therefore, this gives rise to the demand for systemizing the domain knowledge of the fixture design to clarify the CAFD design requirements.
Accordingly, Boyle et al. [10, 62] reported a CBR methodology called CAFixD based on systemizing fixture design knowledge. The CAFixD adopts a demanding approach to define indexing attributes based upon axiomatic design and functional requirement decomposition. A design requirement is decomposed in terms of functional requirements, and the design is then reconstituted to generate a complete fixture design. This system uses the normal CBR technique to retrieve units based on the requirements of the fixturing functions, which are then refined and/or modified during the unit design in detail.
After that, Kang et al. [74] combined a CRR with knowledge based reasoning (KBR) to develop the process of fixture design in CBR based system so that CBR is employed as its core and KBR as the assistant. The designer can achieve a satisfactory subset of potential configuration designs of a fixture from the KBR process by utilizing the design information. After that, the designer, by utilizing this subset and other relevant information, can achieve a similar fixture design from the retrieved case and utilize the adaptation process of CBR to finally achieve a satisfactory fixture design.
To systemize the knowledge information and also visualize and facilitate the design process, Wang and Rong [6] presented an XML based with multilevel CBR methodology which was based upon the procedure of human designing. They proposed an expression of XML based fixture design information to systematically manage fixture resources. They suggested a three-level case library and three-stage retrieval system. In this research, an indexing system that represents the workpiece and fixture information is created and integrates the designer's feedback in each step of the process. Three levels of case library can be navigated by the result of previous retrieval step, designer feedback, and also combination of them.
Recently some researchers in CBR based fixture design systems have got the idea that a single CBR system using attribute similarity often has not a good performance on accurate results. Typically, Peng et al. [8, 75] accomplished fixture design in VR by combining RBR and CBR. They deduced positioning and clamping features with RBR and fuzzy evaluation and obtained feasible fixture design based on workpiece similarity with CBR. Zhou et al. [1] also proposed a feature-oriented fixture design method CBR based combined with RBR for aircraft structural parts. They also adopt model based design technology to define structure performances and machining features of 3D workpiece model and then workpiece feature is identified with feature recognition technology, and previous fixture design cases and rules are obtained by matching it with fixture knowledge models. In Table 3, the significant CBR based fixture design methods regarding design process, retrieval, and indexing process are considered.
The CBR based fixture design methods.
3.4. Summary
As mentioned above, many efforts have been done on CBR systems as the most appropriate and effective method in fixture design task to obtain powerful and automated CAFD systems. There are still gaps between the industrial demands and academic research in the CBR based research despite recent achievements in this sector [6]. Research in some scopes can overcome the current shortcoming, namely, combining CBR with other intelligent methods to come out with effective and more comprehensive fixture design systems.
Furthermore, more research in case adaption and modification is still needed to propose some useful strategies in order to reduce reliance on designer experience and increase degree of automation in fixture design systems as well. Accordingly, developments on information representation and integration of CBR into computer aided fixture design could lead to a revolutionary progress in traditional fixture design.
A formalized methodology is still required for deciding on the case indexes that explain the design requirements, which can further enhance the retrieval step of the CBRs. More importantly, the retrieval methods of CBR based fixture design systems as the main part of the methodology are required to be more efficient. More research is required on discovering a better method to measure similarity. The present computation techniques, which are mainly linear equations, constrain the performance of the CBR method from becoming better.
4. Conclusion
Even though several artificial intelligence techniques have been utilized to facilitate the designing process in the four steps of fixture design (i.e., setup planning, fixture planning, unit design, and verification), very few techniques carry out all the four steps of fixture design. Therefore, the integration of all these separated systems needs to be done to achieve an efficient method. Future researches could concentrate on the importance of better and efficient fixture systems integration with other manufacturing systems. And also, the fixture design task must be put into an overall process of manufacturing to achieve the best solution for fixture design.
In terms of fixture design steps, there are still some needs in these substeps especially for fixture design optimization. There are still very few methods that provide more than just basic assistance in optimizing or verifying fixture information. Moreover, there is a need for integrating the different modeling and control techniques for the optimization and verification of fixture performance. The current survey suggests that there is a lack of concentration on determining the designs of individual clamping and locating units.
In case of CBR based fixture design systems, some fields remain necessary and crucial to be considered. Accordingly, there needs to be an efficient technique to refine, model, and use the domain knowledge of fixture design (fixture design cases). In addition, there is a need for an effectual technical system that can assist the designer of the fixture to simplify the design process and generate design ideas. Moreover, research on methods of automated case adaptation is required as it can allow the automated fixture design configuration. Finally, there is still a strong need for totally effective specific method to determine the best case in database in fixture design field.
As future trends two areas that can be interesting to investigate are
fixture design in nano- and micromachining as nanometric machining has quite different physics compared to conventional machining as in the material and physical properties and the manufacturing process;
to position a network of multisensors into the workpiece fixture system and utilizing online intelligent control methods to adapt the fixturing contacts and forces in the process of machining adaptively.
Conflict of Interests
The authors declare that there is no conflict of interests regarding the publication of this paper.
Footnotes
Acknowledgments
Financial support from the Ministry of Education, Malaysia, and the Universiti Teknologi Malaysia through Research University Grant no. 05H27 is gratefully acknowledged.