序贯测试动态优化与多目标优化算法研究

发布时间：2018-11-07 12:30

【摘要】：由于电子系统结构日趋复杂,对其性能及状态的测试也变得愈加困难,因此,需要在系统设计时就将可测性作为一项指标纳入设计范畴。序贯测试问题的目的是生成一组总体测试代价最小的测试序列以识别系统的故障,对减少系统后期维护成本具有现实意义,本文对序贯测试问题进行了深入研究,对不同方面的序贯测试问题提出了解决方法。目前进行可测性辅助设计的软件较少,且大多都是基于客户端/服务器架构(C/S架构),本文使用基于浏览器/服务器(B/S架构)架构对相关功能进行设计与实现。本文主要研究工作如下:1.介绍了序贯测试问题及常用于解决序贯测试问题的AO*算法,由于AO*算法的性能主要取决于所选的启发函数,根据实际需要介绍了两种不同的启发函数:可求取系统最优测试代价的基于霍夫曼编码的启发函数和用于快速估算系统测试代价的基于信息熵的启发函数,并用两个实例说明了这两种启发函数的区别及各自的使用范围。2.由于序贯测试的相关参数在系统实际使用中经常发生改变,本文提出根据具体发生的变化对已有的故障诊断树进行修改的算法,该算法利用了已有信息对原故障诊断树进行调整,相对于重新生成新的故障诊断树的效率要高,且该算法需要判断是否需要对原故障诊断树进行调整,适合在参数经常发生波动的情况下提高效率。3.首先说明了多目标优化问题,并将实际序贯测试问题遇到的多目标问题与经典多目标问题相结合,并介绍一种基于多目标极值的遗传规划优化算法来解决这一问题,这种算法在遗传操作的基础上增加了分组、分配适应度、择优等操作,在种群进化的过程中选出非支配解,最终输出一组非支配解供系统的设计人员参考使用。4.介绍了软件的总体设计,对软件结构有了系统的了解,并介绍了开发软件使用的技术及数据结构。然后对软件的主要功能进行了详细介绍,系统建模模块可用于系统自动建模或是手动输入参数,序贯测试模块可以生成系统的故障诊断树和相关的参数报告,动态变化针对已生成的序贯诊断树进行修改,多目标优化可产生一组非支配解。
[Abstract]:Because the structure of electronic system is becoming more and more complex, it is more difficult to test the performance and state of electronic system. Therefore, it is necessary to include testability as an index in the design of electronic system. The purpose of the sequential test problem is to generate a set of test sequences with the lowest total test cost to identify the system faults, which is of practical significance to reduce the maintenance cost of the system in the later stage. In this paper, the sequential test problem is studied in depth. The solutions to the sequential test problems in different aspects are presented. At present, there are few software for testability aided design, and most of them are based on client / server architecture (C / S architecture). In this paper, browser / server (B / S) architecture is used to design and implement the related functions. The main work of this paper is as follows: 1. This paper introduces the sequential test problem and the AO* algorithm, which is often used to solve the sequential test problem. Because the performance of the AO* algorithm mainly depends on the selected heuristic function, According to the practical needs, two different heuristic functions are introduced: the Hoffman coding heuristic function which can be used to calculate the optimal test cost of the system and the information entropy based heuristic function used to estimate the system test cost quickly. Two examples are given to illustrate the difference between the two kinds of heuristic functions and their scope of use. 2. 2. Because the related parameters of sequential testing often change in the actual use of the system, this paper proposes an algorithm to modify the existing fault diagnosis tree according to the specific changes. The algorithm makes use of the existing information to adjust the original fault diagnosis tree, which is more efficient than the reconstruction of the new fault diagnosis tree, and the algorithm needs to determine whether the original fault diagnosis tree needs to be adjusted. Suitable in the case of frequent fluctuations in parameters to improve efficiency. 3. First, the multi-objective optimization problem is explained, and the multi-objective problem is combined with the classical multi-objective problem, and a genetic programming optimization algorithm based on multi-objective extremum is introduced to solve the problem. On the basis of genetic operation, this algorithm adds some operations, such as grouping, assigning fitness, selecting optimal operation, and selects the non-dominated solution in the process of population evolution. Finally, it outputs a group of non-dominated solutions for the reference of the designers of the system. 4. This paper introduces the overall design of the software, and gives a systematic understanding of the software structure, and introduces the technology and data structure used in the development of the software. Then the main functions of the software are introduced in detail. The system modeling module can be used to model the system automatically or input parameters manually, and the sequential test module can generate the fault diagnosis tree and related parameter reports of the system. The dynamic change is modified to the generated sequential diagnostic tree, and a set of non-dominated solutions can be generated by multi-objective optimization.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18;TP311.52

【参考文献】