Abstract: The device life data which involve more than one kind of the failure mode are often fitted by the Weibull mixtures.Due to the complex forms and the multiple parameters,the parameter estimation of these distributions is quite difficult.As to this problem above,on the basis of improving the step size scales and the probability of discovery of the Cuckoo Search (CS) algorithm,a parameter estimation of the Weibull mixtures based on the the proposed method is proposed.At first,the method need to build an optimal model for the aim of minimizing the residual sum of squares.Then,the model is solved by the improving algorithm.In the case study,the life data of the craft windshields are regarded as the fitting objects of the two-fold two-parameter Weibull.The Cuckoo Search (CS) algorithm and other three kinds of the improving Cuckoo Search (CS) algorithm are used to search the best solution for 2 000 times simultaneously.The optimization results are compared and the contrast proves that the accuracy and the success rate of the parameter estimation based on the algorithm which mixes the two improved method is the better one.

Key words: reliability analysis, parameter estimation, least square method, Cuckoo Search (CS) algorithm, algorithm improvement