A parametric approach of computing Taguchi signal-to-noise ratio from incomplete life data

The Taguchi method for experimental design is widely utilized to improve product reliability through the systematic arrangement of multiple runs or treatments, each representing a unique combination of factor levels. In such experiments, lifetime data o f sample units are gathered via life tests, which are typically monitored periodically and terminated at a pre-determined endpoint. The resultant data, characterized by small sample sizes, high censorship, and heterogeneity, preclude direct computation of t he Taguchi signal-to-noise ratio (SNR), a crucial measure of robustness used to evaluate the performance of a run. This paper addresses this issue by introducing a parametric approach that begins by constructing a life distribution model for each run through a three-step parameter estimation procedure. It then calculates the SNR from the fitted distribution using SNR expressions derived from typical life distributions discussed herein. The effectiveness and superiority of the proposed method are demonstrated through two real-world examples, underscoring its applicability and potential benefits in enhancing the reliability analysis of complex products.