基于概率神经网络的发动机故障诊断

 用反向传播神经网络 (BPNN)和概率神经网络 (PNN)对航空发动机若干原型故障进行定性的诊断,并将仿真结果进行了比较。仿真结果表明,当测量参数不包含噪声或噪声较小时,两种网络都具有很高的诊断准确率;当测量参数的噪声较大时,则概率神经网络的诊断准确率远大于反向传播神经网络,显示了概率神经网络较强的诊断鲁棒性。此外,概率神经网络能够充分利用故障先验知识,并考虑代价因子的作用,从而把误诊断可能带来的损失减小到最低程度。

PROBABILISTIC NEURAL NETWORKS BASED ENGINE FAULT DIAGNOSIS

In this paper, both back\|propagation neural networks (BPNN) and probabilistic neural networks (PNN) are applied to qualitative aero-engine prototype fault diagnosis, and the simulated results are compared with each other. The simulated results show that when the measurements do not contain any noise or the noises are comparatively small, the success rates of diagnosis of both BPNN and PNN are quite high; when noises rise, the success rate of PNN is much higher than that of BPNN.