Research of reliability distribution model of numerical control machine tool based on ANN model and HPSO algorithm

Abstract:
An increasing precision of products often leads to a decrease of failure data. Reliability distribution cannot adequately be described with small data sample. To analyze the reliability distribution, expanding the failure data by artificial neural network is an appropriate method of use. A comprehensive criteria, including generalized regression neural network, hypothesis testing, normalized root mean square error, is presented for selecting the number of components of distribution model. The parameter estimation of the distribution model is solved using maximum likelihood estimation which is improved by hybrid particle swarm optimization. In this paper, the field failure data of four numerical control machine tools were tested and analyzed. The experiment results showed that the two-fold three-parameter Weibull distribution is superior to other models and more suitable for modeling failure data of NC machine tools.