NHPP models use Non-homogeneous Poisson Process to describe failure process ofsoftware, so that to assess present reliability and predict reliability trend. With applyingsimply and good fit effect, NHPP models are the most widely used SRGM. Yet the amount ofmodels is expanding, each model has different assumption and application environment, somodel selection is becoming a challenge to engineers. Recent researches focus on looking forcommon characteristics and developing universal framework of models, and embody therelevant function according to software development and testing environment to obtainspecific model when applying. Existing framework is too obsolete to include newachievements of software reliability research.The paper gives a new NHPP framework based on analysis of research methods of classicmodels. It first introduces some basic concepts about software reliability models and relativemathematical knowledge. Then it enumerates several typical software models and theirparameter estimations. At the end, considering the present NHPP software reliability growthmodels, this paper proposes an improved framework, which considers fault correlations andsoftware components construction, as well as imperfect software debugging, time-dependentfault detection rate, and time-dependent fault removal efficiency. Many existing NHPPsoftware reliability growth models are special cases of the framework. |