Weibull and lognormal Taguchi analysis using multiple linear regression

The paper provides to reliability practitioners with a method (1) to estimate the robust Weibull family when the Taguchi method (TM) is applied, (2) to estimate the normal operational Weibull family in an accelerated life testing (ALT) analysis to give confidence to the extrapolation and (3) to perform the ANOVA analysis to both the robust and the normal operational Weibull family. On the other hand, because the Weibull distribution neither has the normal additive property nor has a direct relationship with the normal parameters (µ, σ), in this paper, the issues of estimating a Weibull family by using a design of experiment (DOE) are first addressed by using an L9 (34) orthogonal array (OA) in both the TM and in the Weibull proportional hazard model approach (WPHM). Then, by using the Weibull/Gumbel and the lognormal/normal relationships and multiple linear regression, the direct relationships between the Weibull and the lifetime parameters are derived and used to formulate the proposed method. Moreover, since the derived direct relationships always hold, the method is generalized to the lognormal and ALT analysis. Finally, the method’s efficiency is shown through its application to the used OA and to a set of ALT data. •It gives the statistical relations and steps to use the Taguchi Method (TM) to analyze Weibull data.•It gives the steps to determine the unknown Weibull family to both the robust TM setting and the normal ALT level.•It gives a method to determine the expected lifetimes and to perform its ANOVA analysis in TM and ALT analysis.•It gives a method to give confidence to the extrapolation in an ALT analysis by using the Weibull family of the normal level.