基于HMM的电子设备状态监测与健康评估

State monitoring and health evaluation of electronic equipment using HMM

为了克服隐马尔科夫模型(hidden Markov model,HMM)在训练时波氏(Baum-Welch,B-W)算法易陷入局部最优解的不足,采用多智能体遗传算法(multi-agent genetic algorithm,MAGA)对其进行参数估计,设计了染色体编码方法和遗传操作方式。利用Viterbi算法的状态估计和状态回溯能力对温控放大器进行状态监测和健康评估研究,仅需建立一个HMM,大幅度减少了HMM作为分类器使用时的模型训练计算量。仿真结果表明,MAGA优化的HMM具有更好的状态监测性能,采用Viterbi算法得到的状态概率值对设备进行健康评估有效可行。

For overcoming the deficiency that the Baum-Welch （B-W） algorithm is easy to fall into local optimal solution, a multi-agent genetic algorithm （MAGA） is used to estimate parameters of the hidden Markov model （HMM）. Chromosome coding method and genetic operation mode are designed. State monitoring and health evaluation of the temperature control amplifier are researched utilizing the state estimation and retrospect ability of the Viterbi algorithm. Only one HMM is established, which greatly reduces the calculation of model training of HMM as a categorizer. The simulation results show that the HMM optimized by MAGA has a better state monitoring performance, and it is practical to evaluate health situation of equipments using state probability obtained by the Viterbi algorithm.