Adaptation of design cases is usually the most challenging part in building any case-based reasoning design system. The success of the adaptation process in finding a solution for a new design problem determines the success of the entire case-based reasoning (CBR) system. The techniques used for generating design solutions have many common aspects among the various engineering design classes that make them amenable to be captured in a generic framework for an acceptable level of abstraction. This paper proposes a design-plan-oriented methodology for adapting design cases to produce a solution to a new design problem in the domain of engineering design. The proposed methodology uses multicase adaptation and case built-in adaptation knowledge to produce a design plan for a new design problem. We first define the model of case representation to work with the proposed methodology. We then define the overall structure of the procedural framework of this methodology and its subprocesses. The system is then demonstrated through an application from the structural engineering domain.

A new methodology for solving resource leveling problems is introduced using Artificial Neural Networks (ANN). This paper describes a new efficient and robust approach which is unique to those utilized by traditional heuristic and optimization resource leveling techniques. The Resource Leveling Artificial Neural Network (RLANN) exploits advantages of both Hopfield networks and competition-based artificial neural networks. The universal scheme of the RLANN is applicable to construction project networks produced with Critical Path Method (CPM), in forms of either arrow or precedence diagrams. The scheme is comprised of two layers, an input and a competition layer, of artificial node matrices fully connected by links. Solving mechanisms inside the RLANN are based on an equation of motion and a competition strategy that control the level of daily resource usage. While the equation of motion governs activities to be shifted within schedule constraints, the competition process finds the best positions for the activities to achieve optimum results. The approach is simple and can be implemented on either a personal computer or a parallel processing device. The solutions produced are comparable to, or better than, those generated by other heuristic or optimization techniques. This paper describes the development of the RLANN, its solving mechanisms, and its uses in construction resource leveling problems. The comparison of the result of the RLANN to those of other traditional techniques is also included. The conclusions highlight the applicability of this model to other civil engineering problems.

Configuration assistants are tools for guiding the final user in simple configuration tasks, such as product assembling and customization or study plans generation. For their wide availability, web-based configuration assistants are valuable in fields such as electronic commerce and information services. We describe a general approach for building web-based configuration assistants: from a high-level description of the configuration constraints and of the basic items, given in a declarative language, the hypertext files for user guidance and the Java code for constraint checking are generated. We claim that the general approach of process-oriented configuration, where the user is guided through the configuration process by an explanatory hypertext, as opposed to product-oriented configuration, where one starts from a high-level description of the product of the configuration, is better suited for many application domains.

This paper is intended to be the opening salvo of the workshop, Computing Futures in Engineering Design (Dym, 1997). Thus, I want to take this privileged moment to ask you to think with me about the role of design in engineering. In particular, I want to reflect upon how design is articulated and how design is taught; about the role of design in engineering education and in the practice of engineering; and about the role that could be played locally and, perhaps, nationally by a center devoted to design education. Because I teach here at Harvey Mudd College (HMC), and because most of you are visitors, I will place my remarks in our context by telling you about what we do here and how that doing has shaped my thinking.

The research and development of a simulation-based decision support system (SB-DSS) capable of assisting early collaborative design processes is presented. The requirements for such a system are included. Existing collaborative DSSs are shown to lack the capability to manipulate complex simulation-based relationships. On the other hand, advances within the machine learning in design community are shown to have a potential for providing, but have not yet addressed, simulation-based support for collaborative design processes. The developed SB-DSS is described in terms of its four principal components. First, the behavior-evaluation (BE) model is used to both structure individual, domain-specific decision models and organize these models into a collaborative decision model. Second, a probabilistic framework for the BE model enables management of the uncertainty inherent in learning and using simulation-based knowledge. Significantly, this framework provides a constraint satisfaction environment in which simulation-based knowledge is used. Third, a statistical neural network approach is used to capture simulation-based knowledge and build the probabilistic behavior models based on this knowledge. Fourth, since a probability distribution theory does not exist for the nonlinear neural network approaches, Monte Carlo simulation is introduced as a method to sample the trained neural networks and approximate the likelihoods of design variable values. Consequently, constraint satisfaction problem-solving capability is obtained. In addition, a mapping of the SB-DSS architecture onto a collaborative design agent framework is provided. Experimental evaluation of a prototype SB-DSS system is summarized, and performance of the SB-DSS with respect to search and usability metrics is documented. Initial results in developing the simulation-based support for collaborative design are encouraging. Lastly, a categorization of the machine learning approach and a critique of the proposed categorization scheme is presented.

The important questions for instructors to address concern what skills the student is to learn and how the student is to be motivated to acquire those skills. Questions about simulations, graphics tools, and the like are unimportant until the first two questions have been answered adequately. We discuss the role of explanation by students and describe a mechanism for motivating students to learn.

Abstraction and generalization of layout design cases generate new knowledge that is more widely applicable to use than specific design cases. The abstraction and generalization of design cases into hierarchical levels of abstractions provide the designer with the flexibility to apply any level of abstract and generalized knowledge for a new layout design problem. Existing case-based layout learning (CBLL) systems abstract and generalize cases into single levels of abstractions, but not into a hierarchy. In this paper, we propose a new approach, termed customized viewpoint—spatial (CV–S), which supports the generalization and abstraction of spatial layouts into hierarchies along with a supporting system, SPIDA (SPatial Intelligent Design Assistant).

Representing objects and their interactions can be quite challenging when an application requires many complicated, interconnected objects that are restricted in how they can be instantiated. In this paper, we present our approach to conceptual modelling. We have used this approach with success in a number of applications, the largest of which is the PROSE family of configurators. PROSE was first deployed in 1990 and has been used to configure over 4 billion dollars worth of AT&T and Lucent telecommunications equipment. We will discuss our approach to conceptual modelling, which is based on knowledge representation, show how it meets our representation and reasoning needs, and then discuss the relative merits of the approach.2

While there are many project management tools and software packages available, these are not widely used in the design curriculum at colleges and universities. This may reflect some of the differences between conventional projects and design activities. In particular, the open-ended nature of design activities and the need to clarify the client's intention may lead to the conclusion that conventional project management tools are only useful for the most routine activities in the design process. It is suggested that there is a market for a new set of tools for teaching the management of design. These tools should incorporate the most useful of the current management tools and integrate them with some of the requirements of effective design, including support for clarification of objectives, functional analysis, and generation and evaluation of alternatives.

Configuration is a complex task generally involving varying measures of constraint satisfaction, optimization, and the management of soft constraints. Although many successful systems have been developed, these are often difficult to maintain and to generalize in rapidly changing domains. In this paper, we consider building intelligent knowledge-based systems with maintainability well to the fore in our requirements for such systems. We introduce two case studies: the initial proof of concept, which was in the domain of computer configuration, and a further field-tested study, the configuration of compressors. Central to our approach is the use of the proof planning technique, and the clean separation of different kinds of knowledge: factual, heuristic, and strategic.

The application of machine learning (ML) to solve practical problems is complex. Only recently, due to the increased promise of ML in solving real problems and the experienced difficulty of their use, has this issue started to attract attention. This difficulty arises from the complexity of learning problems and the large variety of available techniques. In order to understand this complexity and begin to overcome it, it is important to construct a characterization of learning situations. Building on previous work that dealt with the practical use of ML, a set of dimensions is developed, contrasted with another recent proposal, and illustrated with a project on the development of a decision-support system for marine propeller design. The general research opportunities that emerge from the development of the dimensions are discussed. Leading toward working systems, a simple model is presented for setting priorities in research and in selecting learning tasks within large projects. Central to the development of the concepts discussed in this paper is their use in future projects and the recording of their successes, limitations, and failures.

This article presents a generalized ontology of product configuration as a step towards a general ontology of configuration, which is needed to reuse and share configuration knowledge. The ontology presented consists of a set of concepts for representing the knowledge on a configuration and the restrictions on possible configurations. The ontology is based on a synthesis of the main approaches to configuration. Earlier approaches are extended with new concepts arising from our practical experience on configurable products. The concepts include components, attributes, resources, ports, contexts, functions, constraints, and relations between these. The main contributions of this work are in the detailed conceptualization of knowledge on product structures and in extending the resource concept with contexts for limiting the availability and use of resources. In addition, constraint sets representing different views on the product are introduced. The ontology is compared with the previous work on configuration. It covers all the principal approaches, that is, connection-based, structure-based, resource-based, and function-based approaches to configuration. The dependencies between the concepts arising from different conceptualizations are briefly analyzed. Several ways in which the ontology could be extended are pointed out.

Gradient-based numerical optimization of complex engineering designs offers the promise of rapidly producing better designs. However, such methods generally assume that the objective function and constraint functions are continuous, smooth, and defined everywhere. Unfortunately, realistic simulators tend to violate these assumptions, making optimization unreliable. Several decisions that need to be made in setting up an optimization, such as the choice of a starting prototype and the choice of a formulation of the search space, can make a difference in the reliability of the optimization. Machine learning can improve gradient-based methods by making these choices based on the results of previous optimizations. This paper demonstrates this idea by using machine learning for four parts of the optimization setup problem: selecting a starting prototype from a database of prototypes, synthesizing a new starting prototype, predicting which design goals are achievable, and selecting a formulation of the search space. We use standard tree-induction algorithms (C4.5 and CART). We present results in two realistic engineering domains: racing yachts and supersonic aircraft. Our experimental results show that using inductive learning to make setup decisions improves both the speed and the reliability of design optimization.

This article addresses the issue of educating undergraduate engineering students in the appropriate use of computer simulation in the design process. The premise that poorly designed assignments involving simulation can actually impair understanding is addressed. A set of goals for simulation-based exercises is suggested, and some tactics for meeting these goals are introduced. Finally, a specific example of a half-term assignment that is used to meet these goals is provided for illustration.

A fixed set of components with a fixed set of properties is often regarded as a defining characteristic of configuration design problems. In configuration design for larger and more complex systems (e.g., street cars), however, this often is not really true. The reason is that sometimes the component library offers no component that meets the actual requirements, so that a new component has to be added to the library. In general, configuration design for larger products must be interactive: interactive in the sense that the user controls the configuration process and in the sense that the phases of knowledge acquisition and configuration are partly mixed. This paper describes a method and a corresponding software tool (SyDeR, System Design for Reusability) that support the interactive configuration design of complex products, especially in the tendering phase. It combines three different technologies:

[bull] structural modeling of technical systems;

[bull] a library for technical solutions, which is based upon taxonomies and allows the reuse of the technical solutions;

[bull] constraint techniques to propagate design decisions and check designs for consistency to support interactive configuration design.

This paper gives an overview of the functionality of the SyDeR tool and describes the main ideas behind the modeling language, which is especially oriented towards the requirements of system design problems. It also explains how we integrated structural modeling, taxonomies, and advanced constraint reasoning techniques into real-world applications.

This paper presents a formalism for considering the issues of learning in design. A foundation for machine learning in design (MLinD) is defined so as to provide answers to basic questions on learning in design, such as, “What types of knowledge can be learnt?”, “How does learning occur?”, and “When does learning occur?”. Five main elements of MLinD are presented as the input knowledge, knowledge transformers, output knowledge, goals/reasons for learning, and learning triggers. Using this foundation, published systems in MLinD were reviewed. The systematic review presents a basis for validating the presented foundation. The paper concludes that there is considerable work to be carried out in order to fully formalize the foundation of MLinD.

The thesis presented here is that the result of engineering is the design, construction, or operation of systems or their subsystems and components and that the teaching of systems must be central to engineering education. It is maintained that current undergraduate engineering curricula do not give the student adequate appreciation of this major intellectual element of their profession. Five proposals for approaches to correct this deficiency are offered: opportunities for clinical practice throughout all the undergraduate years; the use of distributed interactive simulation technology in semester-long projects; courses or course material on the phenomenology and behavior of systems; use of project management tools in engineering clinics; and encouraging engineering faculty to spend some part of their sabbaticals engaged in system design or operation. Issues of implementation are addressed, including the scaling of these ideas to universities that must meet the needs of large numbers of students.

Within the last 20 years, much research and development has been conducted to deliver computer tools to assist engineers in performing their tasks. We are now experiencing the power of worldwide networked computing environments and the ability to easily share large quantities of information over geographically dispersed environments. However, are the computing environments available today supporting engineers in doing their job or defining how they must do their job? We need to develop systems that are more transparent and understandable to the users and that are more responsive to the individual needs and idiosyncrasies of the persons using these assistants.

One of the main difficulties with configuration problem solving lies in the representation of the domain knowledge because many different aspects, such as taxonomy, topology, constraints, resource balancing, component generation, etc., have to be captured in a single model. This model must be expressive, declarative, and structured enough to be easy to maintain and to be easily used by many different kind of reasoning algorithms. This paper presents a new framework where a configuration problem is considered both as a classification problem and as a constraint satisfaction problem (CSP). Our approach deeply blends concepts from the CSP and object-oriented paradigms to adopt the strengths of both. We expose how we have integrated taxonomic reasoning in the constraint programming schema. We also introduce new constrained variables with nonfinite domains to deal with the fact that the set of components is previously unknown and is constructed during the search for solution. Our work strongly focuses on the representation and the structuring of the domain knowledge, because the most common drawback of previous works is the difficulty to maintain the knowledge base that is due to a lack of structure and expressiveness of the knowledge representation model. The main contribution of our work is to provide an object-oriented model completely integrated in the CSP schema, with inheritance and classification mechanisms, and with specific arc consistency algorithms.

This position paper proposes the P5BL initiative and vision, that is, Problem-, Project-, Product-, Process-, People-Based Learning. A definition is proposed for the P5BL teaching and learning approach. The discussion identifies key pedagogical issues and innovative roles of computing for each of the five critical Ps in a multisite, cross-disciplinary, project-centered, and team-oriented Architecture/Engineering/Construction environment.