Algebraic Method For Complex Software Reliability Assessment And Its Empirical Study

Software reliability modeling and assessment is an important research direction in the field of software engineering.As computer software enters various fields of human activities,the requirements for software quality are getting higher and higher.Software reliability is the most important indicator representing security and stability of operation.It plays a crucial role in measuring the quality of software development and ensuring the dependence of software system,especially for a class of safety-critical software used in transportation,power,aerospace,defense and other fields.By introducing new methods related to algebraic theory and focusing on how to effectively carry out reliability modeling and assessment in prophase of software development,the dissertation mainly solves the difficulty of describing and calculating the reliability structure of complex software,and further proposes a reliability analysis method suitable for the evolution process of software structure.Besides,the improvement of traditional software reliability growth model and the empirical study of reliability structural model are also discussed.The main work of this dissertation is as follows:(1)Research background and research significance of software reliability assessment are expounded,and the relationship between software reliability models and software reliability assessment is clarified.The two main models—software reliability growth model and software reliability structural model—are discussed from the aspects of modeling ideas,popular models and scope of application.The shortcomings of the current software reliability modeling methods are summarized,and the problems that need to be solved further are pointed out.(2)Aiming at reliability prediction in the later period of software development,the improvement of reliability models based on non-homogeneous Poission process is studied by combining testing effort and fault detection rate.Firstly,the basic assumptions and mathematical expressions of the model which based on non-homogeneous Poission process and considering testing effort are discussed in detail.The modeling implication of testing effort is explained,and major styles of the testing effort function are enumerated.Secondly,the effect of fault detection rate on model improvement is discussed,and the fault detection rate function that can be fitted to different test stages is introduced.Based on that,a mathematical model which combines the factors of testing effort and fault detection rate is constructed in order to improve the performance of software reliability growth models.Experiments show that the improved model is superior to other representative reliability growth models in key performance indicators such as prediction accuracy and goodness of fit.(3)Aiming at reliability estimate in prophase of software exploitation,the role of algebraic method in structural reliability modeling and calculation is studied.Firstly we use algebraic tool to abstract the features of basic structural style into algebraic paradigm forms,and attach reliability calculation criteria to each algebraic paradigm.So the paradigm becomes a generic pattern that can be matched.Then,an algebraic model composed of expressions is established for the complex software architecture.Expressions can be constituted by the combination and nesting of basic paradigms to express the complexity of local and overall software structures.On this basis,an analytic algorithm for the algebraic model is proposed,which automatically identifies and matches all kinds of paradigms by layer in order to solve the overall reliability of a software system.Compared with additional structural reliability modeling methods,the algebraic method better solves the problem of modeling and describing complex software structures,and further forms a general reliability calculation method.Experiments show that the reliability assessment results of the algebraic method are applicable,and show good applicability in real-time and automatic calculation of reliability in prophase of software exploitation.(4)Aiming at reliability analysis and assessment in the process of software evolution,the evolution description based on an algebraic method and the real-time reliability evaluation of evolutionary behavior are studied.Firstly,the essence of software evolution and its influence on reliability are analyzed,and the problems of simple incremental evolution and structural evolution are discussed.Then the software evolution process is presented as an atomic sequence of operations by defining the evolutionary atomic operations.The algebraic method is utilized to track and explore the reliability impact of each step in the sequence,and a reliability assessment process framework oriented to software evolution is formed.The results of practical examples show that the framework can effectively analyze the key points and overall trends of evolution,and is suitable for feedback and constraint evolutionary design in order to improve the quality of software products.(5)Aiming at the problem that structural reliability model is not easy to be applied due to the lack of modeling parameters in the early development period of actual software projects,the numerical estimation methods of module reliability based on polynomial regression and cyclic neural network are studied respectively.Combining algebraic approach with the above module reliability estimates,the overall reliability assessment of several versions of the open source project was completed.Empirical research aims to solve how to effectively implement reliability modeling and assessment.According to the experimental intermediate results and final results,and compared with numerous reliability growth models using testing failure data,it is proved that the algebraic method is feasible to evaluate the reliability of actual projects from the perspective of structural analysis.