Development and Verification of Engineering Design Iteration Models

Abstract

This thesis develops several models of the engineering design iteration process, and attempts to verify the models in multiple ways. As extensions of the design structure matrix methodthe models are based on mapping the input-output flows of information between discrete tasks in a design process. The information flows include cycles which preclude the use of a purely sequential process. The models are of two basic types, the Sequential Iteration probabilistic type, and the Work Transformation type. Each type of model also includes variations. The Sequential Iteration model can suggest ways that an engineering design iteration cycle can be resequenced in order to lessen expected development time. The Sequential Iteration model assumes that the need for iteration is a probabilistic choice, and the order that the tasks are attempted affects the expected length of time to complete the design process. The Work Transformation model can show which portions of the design problem cause the bulk of time during the iteration process. The Work Transformation model is a discrete time dynamic system description of iteration, where the amount of time taken to complete each iteration stage is a linear function of the amount of work done in the previous iteration stage. The Work Transformation model is verified in two ways: by comparing model predictions with industrial design practice, and by demonstrating model predictions in a controlled experimental setting. Two industrial design settings are considered in the verification process. These are brake system design at General Motors, and electronics module design at Digital Equipment. These settings illustrate that the Work Transformation model can describe the design process and reveal system level interactions which control development time. The Delta Design Game serves as the test bed for a controlled experiment which demonstrates that the Work Transformation model can compare different design strategies. For this game, two different design strategies were identified, and a Work Transformation Matrix was constructed for each strategy, predicting different completion times for each. Experiments were then conducted which demonstrated that the strategy predicted to be faster was indeed so, while the final designs were of comparable overall quality.