The Method About Complex System Reliability Growth Segmentde Evaluation Based On Bayes Information Fusion

The quality of a large complex system will be related to the success of a large project of a region and a country, and to the international status in a certain area, and quality problem may be a threat to the life and property security. Reliability is one of the important intrinsic properties of product quality, which run through the whole process of the product or system, from development, shape to be put into use. Therefore, it is important to attach much importance to the reliability management. In development stage, reliability growth test can lead to reliability growth. The popular evaluation method is Duane, AMSAA and bayes reliability evaluation method.For the great complex system, there are thousands of components. The complex system has little similar system and difficult to understand the failure mechanism of the system. This paper research about how to identify the abrupt change point and reliability evaluation, and the research is mainly about the situation that abrupt change point lead to reliability growth rate slow down. Based on growth trend graph to build a segmentation model, and then based on maximum entropy method to get the prior distribution. Using the data of install and integration stage of a certain great complex system to prove the method is usable and effective.Firstly, introduces the research background, research purpose and meaning. In the second chapter, introduces the related concepts and index of reliability and review the model and method of reliability growth management. The third chapter point out the reason that lead to reliability growth’s abrupt change. Establish the reliability growth segmentation model, which can reflect corrective action’s effects on growth characteristics. This segmentation model can be used to solve much more problems. The fourth chapter’s research is based on the earlier study of multistage system reliability growth, build the bayes reliability evaluation model based on the maximum entropy method, and prove the validity of the model through the case. Finally make a summary and outlook about the models built in the paper. The result indicates that identify abrupt change point is useful for this research. Maximum entropy method is useful for insuring the prior parameters. And bayes method is useful for multistage data fusion. The main conclusion of the research as follows:(1) The segmentation model can be used to identify the abrupt change point that lead to the reliability growth rate to speed up, slow down and mixed situation.(2) The segmentation model can indicate the speed change regulation, and help to know more about the system reliability change character.(3) The evaluation method can fuse multistage information, and get more calculated result.