基于遗传算法解决多故障诊断问题

摘要

Since multiple fault diagnosis problem has enormous searching space, a good heuristic method, Genetic Algorithms (GA) was used to solve this problem. Yet trials with different parameter settings of the GA lead to results of different reliability. This paper was focused on comparing reliability of these different trials based on different parameter settings. In conclusion, within certain ranges of parameter settings, trials of tournament selection with smaller tournament size (2 - 4), two-point crossover with higher crossover probability (0.8), higher mutation probability (0.011), larger population size (160), tend to lead to higher reliability.%针对于多故障诊断问题搜索空间大的难点，采用了启发式搜索方法-遗传算法来解决该问题。由于遗传算法有多种参数配置方法，不同的参数设置会导致结果的信度高低不同。通过遗传算法的实现，并对其不同参数配置的实验结果进行比较分析，得出了该算法在多故障诊断问题中取得最佳性能的参数条件，即竞争规模较小（2-4）的竞争选择、较高概率（0.8）的两点交叉、较高变异概率（0.011）及较大种群数量（160）。