Distribution Of Applied Research, Genetic Algorithm-based Software Reliability

Software reliability allocation is a new branch of software engineering. Under the assurance of the software reliability target, we analyze the inner part of the software and separate the software into different elements.To maximize the reliability attained by the customer at certain software development budget,we allocate each element according to the related software reliability target.And the software elements are defined as operation,subsystem,object or the entity which can be allocated.In this thesis, we concentrate on the research and realization of the multi-module software reliability allocation technique.Three model frameworks of the software reliability allocation are discussed and analyzed in detail,which are as follows: RCCM model framework, BCRM model framework and BCUM model framework. The cost functions of software development are discussed and all frameworks are illustrated and experimented with detailed examples.Genetic algorithm is applied to the software reliability allocation.Also, the allocation technique based on genetic algorithm is compared with the conventional allocation techniques. For standard genetic algorithm has the defects of slowly converging and easily falling into local extremum,the author designed and realized the adaptive multi-population parallel genetic algorithm (AMPGA) to solve the reliability allocation problem of large and complicated software systems.Finally,we experimented on the comity center subsystem, delivery center subsystem and system management subsystem of the project:″The JiangSu Province Postal Logistics Information System″.