论文部分内容阅读
主要研究不可靠测试下多信号模型的多故障诊断问题。最优的多故障诊断是计算复杂度完全类(NP-Complete)问题,因此大型系统的诊断一般只能用次优的随机搜索算法。次梯度优化算法能够在虚警概率较小时给出较好的结果,但如果测试个数很多且虚警概率较大时,该算法就不能消除虚警的影响,会使估计的故障覆盖所有失败的测试,而不是找到系统真实的故障。针对这一问题,提出了能够同时考虑虚警和误警的目标函数,使算法能排除虚警的测试准确定位故障,并用改进的遗传算法搜索故障部件提高诊断速度。仿真诊断结果表明,同时发生故障的部件个数较少时,遗传算法的诊断速度明显优于次梯度优化算法,而且能够更有效地抑制虚警的影响。
The main research is multi-signal model multi-fault diagnosis under unreliable test. Optimal multi-fault diagnosis is a Computational Complexity Complete (NP-Complete) problem, so large-scale system diagnostics typically use only sub-optimal random search algorithms. The sub-gradient optimization algorithm can give better results when the false alarm probability is smaller, but the algorithm can not eliminate the influence of the false alarm if the test number is large and the false alarm probability is large, which will make the estimated fault cover all failures Instead of finding a real fault on the system, In order to solve this problem, an objective function that can consider both false alarm and false alarm is proposed. The algorithm can eliminate the false alarm test and locate the fault accurately. The improved genetic algorithm can search the faulty part and improve the diagnosis speed. The simulation results show that the genetic algorithm is superior to the sub-gradient optimization algorithm in the speed of diagnosis, and can suppress the false alarm more effectively.