基于改进HMM和LS-SVM的机载设备故障预测研究

针对传统故障预测方法不能直接预测设备状态的不足，提出了将改进隐马尔科夫模型（HMM）和最小二乘支持向量机（LS—SVM）相结合的机载设备故障预测方法。首先，采用多智能体遗传算法对HMM参数进行训练优化，克服了B-W算法易陷入局部最优解的缺陷；其次，分别研究设计了设备是否具有使用阶段状态退化过程数据2种情况下的故障预测算法流程；最后，以飞机发动机温控放大器为应用对象进行仿真计算。结果表明，该算法不仅预测精度高，而且预测结果直接与设备状态相关，易于理解分析。

Fault Forecast of Airborne EquipmentsBased on Improved HMM and LS-SVM

For the deficiency that the traditional fault forecast methods cannot predict the states of equipments, a fault forecast method based on improved hidden Markov model （HMM） and least square support vector machine （LS-SVM） was presented. Multi-agent genetic algorithm （MAGA） was used to estimate parameters of HMM for overcoming the problem that Baum-Welch algorithm fall into local optimal solution easily. Two fault prognostic algorithms were designed separately according to the situations whether the equipment had the state degradation process data of using stage. The simulation results showed that these two algorithms were of high forecast precision, and the forecast results directly related to the states of equipment were easy to be understood and analyzed.