Multicriteria decision support to integrate customer and designer preferences for quality improvement and process optimization

Robust design is an engineering tool to improve the quality of a product by determining optimal settings of process design variables so that the aggregate variability of quality characteristics is minimized. Modeling robust design with multiple quality characteristics involves a trade-off optimization among quality characteristics. The majority of the studies that concern with multiple quality characteristics employ multivariate quality loss functions that ignore the relative importance of customer needs and quality characteristics based on the customer and designer preferences. The relative importance and the target values for quality characteristics established in a conceptual design phase embody the preferences of customers and the design engineer, and entail a serious expenditure of time and effort. Further, designing a product solely based on customer preferences may not be enough, because the integration of the design engineer's technical knowledge on functional relationships among customer needs is also of primary importance. Nevertheless, this information is rarely used in the robust design phase.; Since customer preferences are highly important from the viewpoint of the proposed methodology, the assessment of customer-perceived relative importance weights for customer needs is first studied and a new methodology to assess weights is proposed in Chapter 3. The new method is based on the integration of a Markovian model to characterize the buying decision process within the entropy method where the customer rates products' ability to fulfill customer needs using linguistic values. The dependency aspect within the entropy method is identified for the first time and represented by modeling the transitions of linguistic values between successive evaluations using a periodic Markov chain. Fuzzy theory and Monte Carlo simulation are employed to quantify linguistic values into numerical values. The new weight assessment method proposed in Chapter 3 enables a more rigorous analysis of the entropy method and represents one of the first attempts in the literature to incorporate stochastic modeling in entropy method and multiattribute weight assessment.; The development of multicriteria decision-based robust design by integrating the preferences of both the customer and the design engineer into the trade-off optimization among quality characteristics is accomplished in Chapters 4 and 5. Multicriteria decision support is developed for uncorrelated and correlated quality characteristics. The multiobjective formulation for robust design of correlated characteristics is accomplished by developing a new method to model the mean-squared error. In order to explore the full set of Pareto solutions, a compromise programming formulation is selected to structure the multiobjective robust design problem. The proposed robust design framework integrates multiattribute weight assessment, response surface methodology, and multiobjective optimization to enable a holistic approach to robust design.