基于QAPSO-SVM的航空发动机故障诊断

Aeroengine Fault Diagnosis Using Quantum Adaptive PSO-SVM

为了提高支持向量机(SVM)对航空发动机气路故障诊断的准确率以及在诊断过程中的可靠性,引入量子自适应粒子群优化算法(QAPSO)对其惩罚因子、核函数参数以及松弛变量进行寻优,并利用优化后的支持向量机对航空发动机气路故障进行诊断。以PW4056发动机的EHM(Equipment Health Monitoring)状态监控系统为例,将QAPSO-SVM算法、QPSO-SVM、PSO-SVM、SVM、BP和Elman神经网络用于航空发动机气路故障诊断中,从发动机故障诊断的性能参数、抗噪能力、训练时间以及诊断精度四方面进行综合比较分析。仿真结果表明:QAPSO-SVM方法在多种故障模式的诊断中准确度都要优于其他两种优化的SVM以及BP与Elman神经网络;QAPSO-SVM的平均诊断准确度与抗噪能力也要优于其他方法,并且在故障诊断性能参数不同的情况下,也能够很好地对其进行诊断,展现出较强适应能力。

In order to improve the accuracy of aeroengine gas path fault diagnosis and solve the problem of Support Vector Machine(SVM) instability in aeroengine gas path diagnosis,a Quantum Adaptive Particle Swarm Optimization(QAPSO) algorithm is used to optimize the parameters of the traditional SVM,then the optimized p support vector machine is used to diagnose the aeroengine gas path fault.Taking the Equipment Health Monitoring(EHM) status monitoring system of PW4056 engine as an example,the QAPSO-SVM algorithm and QPSO-SVM,PSO-SVM,SVM,BP and Elman neural network are used to diagnose the gas path fault and compare synthetically from the selected performance parameters,ability of anti-noise,training time and the accuracy of diagnosis of four aspects.The simulation results show that the QAPSO-SVM method is more accurate than the other two optimized SVM and two neural networks in the diagnosis of each failure mode.The average diagnostic accuracy and anti-noise ability of QAPSO-SVM is also more accurate than other methods.What’s more,the QAPSO-SVM also has good diagnosis ability and strong adaptability as the election fault diagnosis performance parameters are changed.