Research On Chaos Prediction Of ATP Software Reliability

With the rapid development of computer technology,the scale of software is more and more bigger,the complexity is getting higher,and the function is becoming stronger and stronger.It is precisely because of these characteristics of the software that the software system will inevitably fail during operation.For some areas that require higher security and reliability software,such as rail transit,aviation,and nuclear power,research on software reliability models and software failure behavior predictions is necessary.For evaluating the reliability of predictive software,the commonly used method is to model the reliability of the software based on the analysis of the software failure mechanism,and then use the reliability model to predict the software failure behavior.Through the analysis of the causes of software failure,this paper considers that software failure behavior is not only random,but also chaotic.Therefore,chaos theory and software reliability model are combined to predict software failure behavior.The research work of this paper mainly consists of the following aspects:(1)There are problems with existing software reliability models.Most software reliability models are based on probabilistic or stochastic processes.However,the reliability model based on stochastic process is based on the premise of certain assumptions about software failure behavior,which determine the accuracy of the reliability model.If the assumption is quite different from the actual situation,the accuracy of the model will decrease.The software reliability chaos model can avoid the assumptions,and start from the software failure data to mine the inherent law of software failure.(2)Analysis of chaotic characteristics of software failure behavior.In the past,the reason why the software failure behavior was random was that the software usage profile was random.However,this paper considers that the use profile of the software is somewhat certain.In daily operation,the input of the software is quite deterministic and will not be randomly selected from the infinite input field;during the test,the tester will not select the test case from the infinite input field,and the test case is selected quite.The regularity is not completely random.On the other hand,tester's thinking way may be influenced due to their inner consciousness or external environment.This is similar to the fact that chaotic systems are both deterministic and sensitive to initial values.(3)Chaotic time series analysis of software failure data.The chaotic recognition algorithm of chaotic time series data and method of choosing embedding dimension and time delay of reconstructed phase space are introduced.The G-P algorithm and the C-C method are selected to obtain the embedding dimension and time delay of the ATP(Automatic Train Protection)software failure data.The chaoticity is discriminated by calculating the maximum Lyapunov exponent of the ATP software failure data.(4)Verification of ATP software reliability chaotic model.The ATP software failure data is used to verify the chaotic reliability model.The chaotic predictions method based on the maximum Lyapunov exponential and BP neural network are used to predict the software failure behavior.The prediction results are compared with that of the J-M model and the G-O model.The results show that the chaotic reliability model is better than the stochastic process model in predicting software failure behavior.There are 32 figures,2 table and 74 references in this paper.