基于改进粒子群优化LSSVM的飞机发动机故障诊断

Fault diagnosis for aero-engine based on improved particle swarm algorism optimizing support vector machine

针对以往飞机发动机故障诊断方法由于故障样本少而导致的诊断精度低,提出了一种基于最小二乘支持向量机(least squares support vector machine,LSSVM)的飞机发动机故障诊断方法。首先,给出了基于LSSVM对飞机发动机进行故障诊断的模型;然后,为了提高LSSVM的诊断性能,采用改进的粒子群算法对LSSVM的参数进行训练,并定义了最终基于改进粒子群优化SVM的具体诊断算法;最后,通过飞机发动机故障诊断实例仿真实验证明了文中方法能正确地实现故障分类,具有较高的故障诊断精度,且与其他方法相比,具有较优的适应度和较快的收敛速度。

Aiming at the low diagnosis accuracy of traditional fault diagnosis reasoning method for Aero En gine, a diagnosis method based on LSSVM （least squares support vector machine） was proposed. Firstly, the model of diagnosis for LSSVM was given, then in order to improve the performance of LSSVM, the improved particle swarm algorism was used to train the parameters of LSSVM, and the specific algorism for using the improved particle swarm algorism optimizing SVM was defined. Finally, the aircraft generator fault diagnosis experiment shows the method can realize fault diagnosis correctly with the high diagnosis accuracy, and corn pared with the other methods, it has the optimal fitness and rapid convergence rate.