Research On Reliability Model Of Embedded Software

As the high reliability and security, stable development, the embedded software is widelyused in all areas of people’s lives. According to the characteristics of the embedded software,how to establish a software reliability model has been one of the most significant problems inreliability industry.The development of navigation technology has meaning for nationalsecurity, politics and economy. Navigation software systems are widely applied to aerospaceindustry, marine and other realms, its reliability is essential for the safety of delivery device.Embedded software is the main component of these critical software systems. How to predictnavigation software reliability by embedded software reliability model is a focus of currentresearch work.The study is started by the characteristics of embedded software impacts on reliabilitymodeling. The goal is building software reliability models with its own characteristics topredict its reliability.The content is as following:First of all, the major issues of current research work are expounded. It is proposed thattwo characteristics should not be ignored. One is that the testing environment and theoperating environment is different, the other is that the high accuracy of prediction is the basicrequirement. These are two point of departure of this research.Secondly, the first feature of embedded software influencing on software reliabilitymodeling is analyzed quantitatively to prove that the impact of difference exists and can notbe ignored. Then a method to build software reliability model considering the difference ofthe testing environment and the operating environment is presented. Selecting the best fittingmodel on failure data in the testing phase, proposing a new function of environmental factorwhich contacts with failure detection ration of the testing and operational phase, softwarereliability model of the operational phase is established finally. The feasibility of modelingmethod and new environmental factor function is proved on a published data, and alsoproposed model is simple and high precision.Thirdly, for the second feature of embedded software, neural network technology isproposed to solve the problems of the high prediction accuracy, and failure data with noise.The generalization ablilty of neural networks and its improvement is investigated deeply, andthe neural network modeling process is improved. Software reliability model is established onthe published data, and the training performance, fitting and forecasting of model is verified. Fourthly, the characteristics of embedded software electronic chart are studied, and twokinds of embedded software reliability model established are applied to predict its reliability.And then, the feature of two models is summarized.At last, the further work is proposed from reliability data collection and early reliabilityprediction model.