Quality Function Deployment-based optimization and exploration for ambiguity

Abstract

The present paper suggests a customer-oriented optimization method, called the Quality Function Deployment (QFD)-based optimization method, as an effort to reflect customer's preferences in making a trade-off between multiple objectives. A set of target design objective levels to attain a maximum customer satisfaction level is selected by generating only a part of whole Pareto set. This is enabled by constructing an approximation model for a Pareto surface. The use of approximation can guarantee Pareto-optimality of the target design objective levels and support trade-off from various perspectives by having a clear understanding of the association, with considerably reduced computational cost. The customer satisfaction level is defined as a function of design objectives through the House of Quality of QFD and utility-based attribute function. The utility function elaborates the relationship between customer needs and design objectives. In addition, a strategy to update the constructed approximation model is employed to overcome the inaccuracy caused by the use of approximation. For the purpose of facilitating the construction and the update, the Normal-Boundary Intersection (NBI) method, one of the methods to generate a Pareto set, is used. QFD-based optimization enables an easy generation of various non-inferior alternatives and a more accurate examination of them from a customer's standpoint, which is overlooked in conventional optimization methods. It also allows an exploration of the effects of ambiguity included in the method, which originates from the process of value judgements, with negligible expense and time. The exploration makes use of an orthogonal array and analysis of variance table as an efficient exploration method. This exploration may assist a designer in identifying important factors and examining the robustness against the ambiguity.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 24 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.