A New Approach for Parameter Estimation of Mixed Weibull Distribution: A Case Study in Spindle

Abstract

In order to improve the accuracy and efficiency of graphical method and maximum likelihood estimation (MLE) in Mixed Weibull distribution parameters estimation, Graphical-GA combines the advantage of graphical method and genetic algorithm (GA) is proposed. Firstly, with the analysis of Weibull probability paper (WPP), mixed Weibull is identified to data fitting. Secondly, the observed value of shape and scale parameters are obtained by graphical method with least square, then optimizing the parameters of mixed Weibull with GA. Thirdly, with the comparative analysis on graphical method, piecewise Weibull and two-Weibull, it shows graphical-GA mixed Weibull is the best. Finally, the spindle MTBF point estimation and interval estimation are got based on mixed Weibull distribution. The results indicate that graphical-GA are improved effectively and the evaluation of spindle can provide the basis for design and reliability growth.