摘要软件系统规模日益庞大，结构日益复杂，从而导致软件可靠性问题日益突出。关键领域软件的可靠性必须得到严格保证，而软件可靠性参数估计是利用软件可靠性模型开展软件可靠性评估的关键性工作，因而受到研究者的高度重视。83815

传统上两种最常用的参数估计方法是极大似然法和最小二乘法，但是传统数值方法经常会面临不能收敛或者迭代过程过分依赖初值等问题，因此需要寻找更好的软件可靠性模型参数估计方法。近年来，一种新的思路是将群体智能算法应用到可靠性模型的参数估计中去。

本文基于粒子群算法对软件可靠性模型进行参数估计，针对现有的利用粒子群算法中使用的适应值函数存在的问题，结合极大似然估计构造了一种新的适应值函数。本文还针对算法在软件可靠性模型参数估计的具体应用中存在的问题，提出了问题解的剔除方法和利用先验知识的方法。通过对比5组经典软件可靠性数据的仿真结果，本文提出的算法比现有算法在软件可靠性参数估计和预计方面具有更好的适用性，对于失效数的估计比现有算法有很大的提高。

毕业论文关键词：粒子群算法；软件可靠性模型；参数估计；

Abstract With the increasingly large scale of software system, the structure of software is increasingly complex, resulting in software reliability problem increasingly prominent。 Key areas must be strict to ensure the reliability of software, the software reliability parameters estimation is the use of software reliability model in software reliability assessment of critical work, therefore attaches great importance to by the researchers。

Traditionally, two of the most commonly used methods of parameter estimation is maximum likelihood method and least squares, but the traditional numerical method often face can't convergence or iterative process to rely too much on initial value problems, so need to find a better software reliability model parameter estimation method。 In recent years, a kind of new train of thought is the swarm intelligence algorithm applied to the reliability of the model parameter estimation。

Based on particle swarm optimization (PSO) algorithm to parameter estimation of software reliability model, to solve the existing problems in using particle swarm optimization (PSO) algorithm, this paper construct a new fitness function combined with maximum likelihood estimation。 In view of the algorithm in software reliability model parameter estimation problems that exist in the specific application, this paper also proposes the solution of eliminating method and the method of using prior knowledge。 By comparing with the five classic software reliability data sets of simulation results, the proposed algorithm is better than existing algorithms in software reliability parameters estimation and has better applicability and is expectation to estimate the number of failure than the existing algorithm is greatly improved。

Keywords:Particle swarm optimization (PSO);Software reliability model;Parameter estimation。

目  录

第一章 绪论 1

1。1 研究背景 1

1。2 粒子群算法的发展历史及由来 1

1。3 软件可靠性 3

1。4 论文安排 5

第二章 基本概念 6

2。1 软件可靠性模型 6

2。1。1 软件可靠性评估的基本理论 6

2。1。2 JELINSKI-MORANDA模型 6

2。1。3 GOEL-OKUMOTO模型