基于离散粒子群优化算法的多值属性系统故障诊断策略

针对传统离散粒子群优化（PSO）算法仅能搜索多值属性系统（MVAS）最小完备测试集的问题，通过重塑离散PSO算法，提出一种测试序列寻优算法—PSO-测试(TS)算法。在多值D矩阵和五元组的基础上，公式化处理MVAS的诊断策略。重塑离散粒子群的过程，将离散PSO算法与MVAS的故障诊断策略融合。设置PSO-TS算法的自身认知和社会知识阶段的计算规则，并通过引入交换序提升PSO-TS算法中粒子的多样性。采用实例和随机仿真实验验证PSO-TS算法。研究结果表明：与MV-Rollout和MV-IG算法相比，PSO-TS算法的期望测试费用少，能够获得较优的诊断策略，但是运行时间较长。

Fault Diagnosis Strategy for Multi-valued Attribute System Based on a Discrete Particle Swarm Optimization Algorithm

To solve the problem that the traditional discrete particle swarm optimization (DPSO) algorithm can only find the minimum complete test set for a multivalued attribute system (MVAS), particle swarm optimization for test sequencing (PSO-TS) algorithm is proposed. The diagnosis strategy for MVAS is formulated based on multi-valued D matrix and five-tuple. The implementation process of DPSO is remodeled, and DPSO algorithm is combined with the fault diagnosis strategy of MVAS. Subsequently, a set of calculation rules for self-cognition and the social knowledge are set, and the exchange order is introduced to increase particle diversity. The PSO-TS algorithm is verified using experiments and stochastic simulations. Compared with the MV-Rollout algorithm and MV-IG algorithm, the PSO-TS algorithm can obtain an optimal fault diagnosis strategy with a relatively lower expected test cost, but with a longer running time.