| This dissertation develops from quality loss function to warranty loss function in which customer expectation is also considered to be a variable. First, Taguchi's quality loss function for the larger-the-better case, which is different from the smaller-the better and nominal-the-best cases, has been assimilated into the other two by introducing a term called the target-mean ratio. Further topics addressed include the implications of a finite target on the classification of LTB characteristics, a new concept of a "Complementary Characteristic," operating window, complexity, and SN ratio based on complexity.;A warranty is a buyer's confidence owing to the seller's assurance to the buyer that a product will perform as stated or implied. The quality loss function only accounts for immediate issues within manufacturing facilities, whereas the warranty cost occurs during customer use. Therefore, this dissertation develops a methodology that can predict warranty probability and warranty costs on the basis of customer expectation in addition to product performance for smaller-the-better, nominal-the-best, and larger-the-better cases.;In robust engineering, the signal-to-noise ratio is used to improve the robustness of a system. Most products and processes have multiple quality characteristics or output responses. Therefore, this research has been conducted to propose a metric that can be used for multi-response experiments for minimizing quality loss and improving robustness at the same time. The methodology proposed incorporates all three types of characteristics smaller-the-better, nominal-the-best, and larger-the-better and is based on components of quality loss. |
