Evaluation of Design Feasibility under Uncertainty However, the evaluation of design feasibility under uncertainty is often a computationally intensive process. The use of simplified approaches in existing applications may lead to either over-conservative or infeasible design solutions. We investigate several feasibility-modeling techniques for robust design optimization (Du and Chen, 2000). These methods were classified into two categories: methods that require probability and statistical analysis (the probabilistic feasibility formulation and the moment matching method) and methods do not require probability and statistical analysis (the worst case analysis, the corner space evaluation, and the variation pattern method). The effectiveness of each method is compared in terms of its efficiency and accuracy. Constructive recommendations are made to employ different techniques under different circumstances.

A Probabilistic-Based Approach to Simulation-Based Design under Uncertainties Our objective in this research is to develop a probabilistic based approach under the decision-based design framework for the quantification, propagation, and mitigation of a variety of uncertainties in simulation-based complex systems design. The development of the proposed methodology is characterized by research activities in three important areas: (1) Classification and quantification of a variety of uncertainties in simulation-based design, such as parameter uncertainties and the inaccuracy associated with models, (2) Development of computationally efficient techniques for assessing the global impact of uncertainty sources on confidence in design in terms of probabilistic distributions of the system attributes (or total design utility) and design confidence regions, and (3) Development of a probabilistic-based decision making model that can assist designers to make reliable design decisions which are tolerant to the expected system variations. Revolutionary approaches to uncertainty propagation are being developed to overcome the limitations of existing approaches (Du and Chen, 2000). These methods are expected to be the most useful for highly coupled, closed-form, and "black box" type of complex design problems.

Propagation and Management of Uncertainties in Simulation-Based Collaborative Systems Design (sponsored by U.S. Army Tank–Automative and Armaments Command and National Automotive Center): Simulation-based design has become an inherent part of complex systems design. Complications arise when the simulation programs may have input parameters with deviations (external uncertainties) as well as internal uncertainties due to the inaccuracies of the simulation tools or system models. These uncertainties will have a great influence on the design negotiations between various disciplines and may force designers to make conservative decisions. We propose an integrated methodology for managing (mitigating) the effect of uncertainties (Du and Chen, 1999). Two approaches, namely, the extreme condition approach and the statistical approach, are developed to propagate the effect of uncertainties. An uncertainty mitigation strategy based on the principles of robust design is developed.

Exploring Technology and Cost Tradeoffs in Multidisciplinary Based Microelectronics Systems Design (Sponsored by Motorola Advanced Technology Center): Our objective in this project is to develop methods and computer-based tools at Motorola for modeling the interactions, communication, and cooperation among different disciplines under a multidisciplinary design environment. The proposed methods and developed tools will help to predict the effect of design decisions made by individual disciplines on the designs of other disciplines and eventually on the final form of the product at the system level. The proposed project will also facilitate the identification of key interactions between disciplines, and support decision making and exploration of the tradeoffs when multiple technological and economical attributes exist. Motivated by the need for searching efficient and robust optimization algorithms that are practically useful for problems with a wide range of complexity, UIC and Motorola have been collaborating on the comparative studies optimization algorithms. The goal is to identify scaling issues associated with large-scale optimization involving both continuous and discrete variables.

An Open Workshop on Decision-Based Design (Sponsored by the National Science Foundation): In the engineering research community, there is a growing recognition that decisions are the fundamental construct in engineering design. The goal in this web-based workshop (http://dbd.eng.buffalo.edu/) is to engage design theory researchers in scholarly and collegial dialogue to establish a rigorous and common foundation for decision-based design. The website is in its fourth year operation and has attracted more than 300 on-line (registered) participants over the world from industries, national laboratories, and over 60 academic institutions. Three face-to-face meetings are organized annually to supplement, plan, and direct the open workshop on line. . Dr. Wei Chen is one of the three organizers of this workshop. A special journal edition on "Decision-Based Design: Status & Promise", co-edited by the organizers, is being published through the Journal of Engineering Valuation & Cost Analysis (Vol3, Issues 1 & 2).

The Decision-Based Design Approach to Product Design To implement the decision-based design approach, it is critical to correctly derive the design utility by incorporating the needs of the customers as well as the various considerations of a company. It is also critical to employ a utility forming procedure that could avoid the paradox due to aggregations of multiple rankings based on multiple attributes. The decision-based design framework we are investigating consists of product concepts, a demand model, a producer model, and the design utility model. The demand model expresses the customers' preference by deriving the product demand as a function of product price and the multiple key attributes that are of concern to the customers. The producer model contains the lifecycle cost of the product as the function of the design variables. The design utility function measures the worth of the alternative with respect to the corporate preference related to profit, return on investment, market share, corporate image etc., etc. The design utility is derived based on both the demand model and the producer model, and therefore it brings the needs of the customer and the producer together. In our approach, the way in which the engineering attributes are treated in deriving the design utility is fundamentally different from most of the conventional approach where the engineering attributes are directly used as a part of a multiattribute utility function.

Inverse Concurrent Design in Extrusion Technology (collaborating with Alcoa)There is an urgent need for the metal industry to accelerate the entire product development cycle through concurrent and systems engineering based on quality engineering concepts so that new and better products can be produced economically. This research is focused on incorporating material considerations under the integrated product and process development scheme. Of a particular interest is the emerging technologies in the extrusion industry to (1) design ideal aluminum extrusion dies, and (2) to design thin-walled structures and their forming processes.

Implemented Projects

The RCEM with Enhanced Approximation Techniques The model approximation capability of the RCEM has been enhanced by utilizing various existing approximation techniques and developing new ones. A new approach based on a combined Response Surface Methodology and Artificial Neural Networks strategy was developed to enhance the model approximation capabilities of RCEM (Varadarajan, et al. 1997).

A Generic Robust Design Model To overcome the mathematical limitations of Taguchi methods, we have developed a generic robust design model (Chen et al. 1996) and extended the robust design concept into the early stages of design for making decisions that are robust to the changes of downstream design considerations and decisions that are flexible to be allowed to vary within a range.

Probabilistic-Based Design Model for Developing a Range of Specifications A probabilistic-based design model has been developed for determining a range of or a set of specifications (Chen and Yuan, 1998). The approach is used as a basis for providing the flexibility that allows designs to be readily adapted to changing requirements. This is obtained by developing a range of design solutions that meet a ranged set of design requirements. Meanwhile, designers are allowed to specify the varying degree of acceptability of a ranged set of design requirements based on their preferences

The Robust Concept Exploration Method for Preliminary Engine Design (sponsored by Pratt & Whitney): Our objective in this project was to apply the Robust Concept Exploration Method (RCEM) to configuring gas turbine systems in the early stages of the development (Varadarajan et al., 2000). The developed methods allowed engine designer to: (1) identify a "composite basis engine" which is a derivation of simulated engine designs with the "best" characteristics, (2) identify critical engine design variables, (3) create fast analysis modules to replace the computationally-expensive engine performance simulation tools, and (4) modify the "composite basis engine" through uncertainty analysis, and what-if questions, while meeting multiple quality constraints.

Design and Evaluation Modules for a Collaborative Vehicle Design Framework(sponsored by U.S. Army Tank–Automative and Armanents Command and National Automotive Center). Our objective in this research program was to develop a "Concept Generator" that could efficiently search the design parameter space to initialize the ground vehicle design process. Specifically, we were interested in utilizing the "Concept Generator" to assist the APDF (Automotive Product Development Framework) proposed by the National Automotive Center. Developed methods were applied to robust design for improved vehicle handling under a range of maneuver conditions ( Chen, Garimella, and Michelena, 2000).

Robust Design for Achieving Flexibility in Multidisciplinary Design Optimization An approach has been developed to provide flexibility in resolving the conflicts between the interests of multiple disciplines (Chen and Lewis, 1998). This was implemented by integrating the robust design concept into game theory protocols, in particular the Stackelberg leader/follower protocol.

Optimal Parameter Settings to Overcome Linear Shrinkage of Stereolithography PartsThe objective of this project was to employ an integrated design approach for improving the part accuracy associated with the stereolithography technology. The most significant building parameters that affect sterelithography part geometrical accuracy were identified using the Fishbone Diagram and the Design of Experiments (DOE) techniques. The effects of the environmental parameters were studied during the building phase and the robust design concept is applied to reduce the impact of their variations on part accuracy. A decision support system was developed to choose the optimal settings of building parameters that yield the best accuracy and the least sensitivity to the operating environment.

Pipe Routing Using Genetic Algorithms and Tessellated ObjectsPipe routing is the technique of developing collision-free routes for pipes between two locations in an environment scattered with obstacles. In the past, research has been primarily focused on the use of deterministic optimization techniques to derive the optimal route. Computational efficiency of deterministic techniques is low for highly nonlinear problems like pipe routing. Due to limitations in the representation of 3-D geometry, the shapes of obstacles have been restricted to primitives. In this research a novel approach to overcome these limitations was attempted. A non-deterministic optimization approach based on Genetic Algorithms was proposed to generate pipe routing solution-sets at a much faster rate (Sundurkar and Chen, 1999). Representation of the objects and pipes in the tessellated format offers huge benefits in computation as well as adaptability.