The computer-assisted configuration of progressive die is a computationally mixed problem. This paper describes the progressive die design problem and reviews some existing research related to the configuration of progressive dies. It explains the model-based reasoning approach used to configure the intermediate and final solutions to the design problem. In particular, the various shape representation and spatial reasoning techniques are presented. Finally, a computer software architecture for the die design system is presented.

In configuration design parts are selected and connected to meet customer specifications and engineering and physical constraints. Specifications include preferences (e.g., “prefer lower cost to higher performance, all things being equal”), bounds on various resources (e.g., “the computer should have four PCI slots”), and other information to customize a configuration. Constraints typically arise from exogenous concerns, such as the available parts, the way parts can interact, and the manufacturing plant.

An integrated design and manufacturing approach allows economic decisions to be made that reflect an entire system design as a whole. To achieve this objective, we have developed and utilized integrated cost and engineering models within a focused design perspective. A framework for the integrated design of an aircraft system with a combined performance and economic perspective is described in this article. This framework is based on the concept of Design Justification using a Design-for-Economics approach. We have developed a knowledge-based system that can be used to evaluate aircraft structural concept material and process selections. The framework consists of the knowledge-based system, integrated with numerical analysis tools including an aircraft performance/sizing code and a life-cycle cost analysis code. Production cost estimates are applied for evaluation of process trades at the subcomponent level of design. Life-cycle cost estimates are used for evaluation of process trades at the system level. Results of a case study are presented for several advanced wing structural concepts for a future supersonic commercial transport aircraft. Cost versus performance studies indicate that a high-speed civil transport aircraft with a hybrid wing structural concept, though more expensive to manufacture than some homogeneous concepts, can have lower direct operating costs due to a lower take-off gross weight and less mission fuel required.

This issue of AIEDAM is based on a workshop on Machine Learning in Design held at the 1996 Conference on Artificial Intelligence in Design, AID'96 (Gero & Sudweeks, 1996), the third of such workshops, with the previous two being held at AID'92 (Gero, 1992) and at AID'94 (Gero & Sudweeks, 1994). The first two workshops also resulted in special issues of AIEDAM (Maher et al., 1994; Duffy et al., 1996).

The collection of papers that follow this Editorial (beginning on page 43) come from the symposium “Computing Futures in Engineering Design” that was held on the campus of Harvey Mudd College in Claremont, California, on May 2–3, 1997. The symposium focused on future roles of computing in doing design and engineering and in the teaching of design and engineering. The intention of the symposium organizer, Clive Dym, was to provide useful insight, advice, and information to educators about how they might think about the future of design in engineering education.

In this paper we intend to motivate various research areas in configuration, based on our experience in developing commercial configuration solutions. Informal definitions are given for the configuration task and for configuration specification and description languages. We also offer an abstract characterization of the issue addressed in configuration models, most often represented as constraints or rules. With these definitions and examples of configuration problems from various domains, we motivate research in (1) a common ontology for configuration, (2) function representation and functional reasoning in configuration, and (3) scaling configuration to large problems.

This paper addresses an important application of machine learning (ML) in design. One of the major bottlenecks in the process of engineering analysis by using the finite-element method—a design of the finite-element mesh—was a subject of improvement. Defining an appropriate geometric mesh model that ensures low approximation errors and avoids unnecessary computational overhead is a very difficult and time-consuming task based mainly on the user's experience. A knowledge base for finite-element mesh design has been constructed using the ML techniques. Ten mesh models have been used as a source of training examples. The mesh dataset was probably the first real-world relational dataset and became one of the most widely used training set for experimenting with inductive logic programming (ILP) systems. After several experiments with different ML systems in the last few years, the ILP system CLAUDIEN was chosen to construct the rules for determining the appropriate mesh resolution values. The ILP has been found to be an effective approach to the problem of mesh design. An evaluation of the resulting knowledge base shows that the mesh design patterns are captured well by the induced rules and represent a solid basis for practical application. The aim of this paper is not only to present the real-life ML application to design, but also to describe and discuss a relation of the work being done to the topic of this special issue: the proposed “dimensions” of ML in design.

This paper proposes a fuzzy neural network (FNN) based approach to construct an individual-oriented car-following system. The feature of this system is firstly to incorporate a personal risk-taking factor in addition to other mechanical factors as the input parameters. Through the learning capability of artificial network, the complex membership functions between the input factors and the output (i.e., the appropriate car-following headway) can be efficiently established, and then the fuzzy logic rules can be properly constructed. The performance of the FNN system is finally assessed against the field data. The results are inspiring that the system is proven capable of providing highly accurate predictions of the required car-following headways from person to person at various speeds. The success of this study provides some clues of utilizing FNN techniques in exploring some individual-oriented machines.

Precision machines are necessary to manufacture parts and subassemblies that require tight tolerances. During the design of precision machines, like any design, it is critical that the best concept is chosen in the early stages of the design process because 80% of the final cost and quality of a product are designed in at this phase. In addition, changes and optimization late in the design process have limited impact on cost and quality. Typically, during the design of precision machines, engineers and skilled machinists develop several machine concepts and down select based on heuristics and past design experience rather than quantitative measures. This paper describes a computation tool, Precision Machine Design Assistant (PMDA), which automates basic machine error simulation and concept evaluation. The tool uses a combination of machine error motion modelling and constraint-based design methods. By combining these methods in a computational environment, multiple machine concepts may be rapidly modeled, analyzed, and compared. The goal of the program is to assist the designer in the selection of a superior concept for detail design. The PMDA methods and implementation are demonstrated in an example.

One of the biggest problems in pneumatic system design is over design. Thus, the results are excessive in costs of the initial investment and it requires too much energy. This article describes the development of expert system based pneumatic design system, PNEUDES (PNEUmatic Design Expert System), prototype that enables the user the optimal design of pneumatic system. Once the design requirements such as cylinder type and work load, etc. are input to the system, optimal cylinder specifications with standardized order-based size, valves types, and necessary accessories are all determined. Also the configuration information such as the connectivity among components and cylinder image data are supplied to the user. It can also help the novice of pneumatic design. The rule-based reasoning approach is used as a reasoning strategy with Intelligent Rule Element shell.

This article contains a description of a constraint solver that determines complete solution spaces. These spaces are defined by sets of constraints in continuous variables. Interactive design is supported through improvements to existing algorithms that have increased system performance. In addition, heuristics for activating the best set of preferred constraints for a design task are presented, and two ways are proposed for interactively exploring design alternatives. IDIOM is an application framework that has successfully tested these algorithms for interactive apartment layout design.

Learning is an intrinsic product of case-based reasoning. Acquiring new cases is one possible way of learning in a case-based system. These cases comprise mainly success knowledge. The successful cases are essentially used to generate new design solutions. But a case-based system also can make use of failure knowledge. In this paper we present how a case-based system can acquire failure cases for verification of the solution created by success cases. We describe IM-RECIDE, a system that uses case-based reasoning for solving design problems that are imperfectly described and explained. The learning aspect is focused and some of the machine learning dimensions in design are criticized. Experimental results in the domain of room configuration also are presented.

The paper describes and evaluates three methods of modelling earthquake accelerograms using artificial neural networks and linear predictors. These are a radial-Gaussian neural networking system, a linear predictor, and a hybrid of these two. Two methods of using these models to predict ground acceleration are adopted. The first is based on a direct prediction of ground acceleration at some point in time in the future from a series of ground accelerations that occurred earlier in the earthquake event. The second is a recursive approach whereby a model predicts a sequence of future ground accelerations by feeding back its predictions to its inputs. The performances of the models are tested using the 1985 Mexico earthquake and aftershock. The linear predictor and hybrid models perform best at direct prediction while none of the models perform particularly well at recursive prediction. The paper concludes with an outline of some areas for future work.

Evolutionary and Adaptive strategies (ES & AS) for diverse multilevel search across a preliminary, whole-system design hierarchy defined by discrete and continuous variable parameters are described. Such strategies provide high-level decision support when integrated with preliminary design software describing the major elements of an engineering system. Initial work involving a Structured Genetic Algorithm (stGA) with appropriate mutation regimes to encourage search diversity is described and preliminary results are presented. The shortcomings of the stGA approach are identified and alternative strategies are introduced. A dual agent strategy (GAANT) involving elements of an ant colony search and an evolutionary search concurrently manipulating the discrete and continuous variable parameter sets is presented. Appropriate communication between the two search agents results in a more efficient search across the hierarchy than that achieved by the stGA, while also simplifying the chromosomal representation. This simplification allows the further development of the preliminary design hierarchy in terms of complexity. The technique therefore represents a significant contribution to configuration design where multilevel, mixed discrete/continuous parameter design problems can be prevalent.

This paper is intended to serve as part of the context in which the other papers in this special issue should be read. Its main goal is to revisit the basic definition of the configuration task, on which many people depend, to show some of its flaws, and to point out how it shapes thinking about the problem. We are concerned about characterizing the reasoning processes used to produce a configuration.

Many of the design systems developed in recent years incorporate some machine learning. The number of such systems already available, and the multitude of design learning opportunities that are slowly being revealed, suggest that the time is ripe to attempt to put these developments into a systematic framework. Consequently, in this paper we present a set of dimensions for machine learning in design research. We hope that it can be used as a guide for comparing existing work, and that it may suggest new directions for future exploration in this area.

The paper proposes and evaluates a binary network architecture and complementary training algorithm designed for pattern classification, with applications in a variety of engineering problems. The advantages of the system are that it can always converge on zero error for a set of unambiguous training patterns (if required), converges rapidly, and circumvents the issue of how many hidden neurons to incorporate in a network. The main benefit of the proposed system over Hamming networks, its main counterpart, is that it can group patterns into different classes along any boundary. The system is shown to outperform 1) the Hamming network for a character recognition problem where the images are subject to both position change and noise; and 2) a radial-Gaussian network in a truck-type classification problem. A variant of the system, where hidden neuron thresholds are set to zero, is shown to further improve performance if a comprehensive set of noise-free input patterns are available for training.

This paper describes the technical principles and representation behind the constraint-based, automated configurator COCOS. Traditionally, representation methods for technical configuration have focused either on reasoning about structure of systems or quantity of components, which is not satisfactory in many target areas that need both. Starting from general requirements on configuration systems, we have developed an extension of the standard CSP model. The constraint-based approach allows a simple system architecture, and a declarative description of the different types of configuration knowledge. Knowledge bases are described in terms of a component-centered knowledge base written in an object-oriented representation language with semantics directly based on the underlying constraint model. The approach combines a simple, declarative representation with the ability to configure large-scale systems and is in use for actual production applications.

As manufacturing systems become more sophisticated and complicated, effective managers know how they play a crucial role in managing an enterprise and managers know how to deal with their dynamics and uncertainty. In this article, a formalism based on the computer-integrated manufacturing open-system architecture (CIMOSA) reference model is presented to specify the business processes and enterprise activities at the knowledge level. The formalism uses an integration of multiple types of knowledge, including precise, muddy, and random symbolic and numerical knowledge to systematically represent enterprise behavior and functionality. To support the modelling process, a prototype is developed and an example for a maintenance activity is presented to demonstrate the effectiveness of the proposed method.

Existing product models do not adequately support the modelling of weld products during the early design stages. A newly developed aggregate product model for welding is introduced, which forms part of CAPABLE, a concurrent engineering toolkit for machining and welding. This product model supports the representation of the design at multiple levels of detail and allows the rapid modelling of alternative designs where solid model representation is inappropriate. The potential applications of the aggregate product model in weld product design, such as the integration of joint design and weldability assessment, are discussed using simple examples. The design of weld products must consider process-dependent properties including geometry, materials combinations, and weld joint characteristics to arrive at a near-optimal configuration. The object-oriented and feature-based model structure of CAPABLE enhances the design process by the automatic generation of compatible options and allows rapid design evaluation. The aggregate product model provides the input data necessary for tasks like process selection, optimization, process sequencing, fixture design and welding orientation selection. The main advantage of this product model over traditional CAD environments is that planning of the welding and fabrication operations can start during the conceptual and embodiment stages of the design process, thus ensuring the concurrency of design and planning tasks and resulting in product and process improvement and cost reduction.