基于径向基神经网络的航空发动机气路故障诊断

Diagnosis of Aero-engine Gas Path Fault Based on Radial-Basis Function Neural Network

摘要:

基于径向基神经网络对民用高涵道比航空发动机风扇、增压级、高压压气机、高压涡轮、低压涡轮5大气路部件的效率降低故障进行诊断；采用Gasturb进行故障训练样本和测试样本库的生成,诊断结果显示,采用径向基神经网络进行航空发动机气路故障诊断的计算时间短、精度较高,不仅能定性的定位故障部位,而且在大多数情况下可以定量的给出该部件的性能衰退程度；某些情况下诊断结果与测试样本不尽一致,但都是方程的合理解,这是因为航空发动机的数学模型是一个多解的复杂方程,一个总性能的衰减对应着多组部件性能衰退的组合；随噪声幅值加大,诊断精度变差,同时研究发现诊断精度受噪声影响的敏感系数在不同的噪声幅值水平下是不同的。

Abstract:

Fault diagnosis of five gas path parts such as fan, booster, high-pressure compressor(HPC), high-pressure turbine(HPT), low-pressure turbine(LPT)'s efficiency degradation have been conducted based on Radial-Basis Function(RBF) neural network. The training and testing samples have been generated by Gasturb software. Diagnose result showed RBF neural network has advantage of less-time-costing and high-precision. RBF neural network can not merely isolate the fault parts, also it can determinate the degradation of components performance. In some instances, the diagnostics result doesn't agree with testing sample, but it also is reasonable solution, because the mathematical model of aero-engine is so complicated that the mathematical equation have more than one reasonable solutions. A degradation of gross performance may be caused by several combinations of components performance degradation. With increasing amplitude of noise, precision of diagnostics became worse. Sensitivity coefficient of diagnostics-precision corrupted by noise is variable with different amplitude of noise.