基于粒子滤波的涡扇发动机健康诊断

Turbofan Engine Fault Diagnosis using Particle Filter Technique

采用粒子滤波算法,设计了一种无需涡扇发动机线性化模型的故障诊断方法。通过含有高斯白噪声的转速测量信号准确估计出相应转速值,构造残差,并设定合适的阈值,实现了故障诊断。压气机健康诊断仿真结果表明:高压压气机(HPC)效率突变30%会导致残差信号1 000倍以上的变化,可以明显检测到故障的发生及发生时刻;当HPC效率突变量为2%,残差信号将变成正常时的6倍,残差信号还可以反映退化程度。因此基于非线性粒子滤波的发动机健康诊断方法可较好地辨识发动机性能退化、故障及失效。

A method of fault diagnosis without linear models for turbofan engine based on particle filter has been designed.Firstly,the value of particular parameters from sensors that contain noise and affected by disturbance would be estimated. Then,set threshold and construct residual signal from the estimated parameters and measured value according to the characteristic of particle filter. Simulation shows that the 30% deviation of HPC＇s efficiency will lead to the change of 100 000% for residual signal. 2% deviation of HPC＇s efficiency would lead to 600% change of the residual signal,and the change of residual＇s change could be used to determine the degree of degeneration. The residual based on particle filter could distinguish degeneration from faults to achieve the objective of fault diagnosis.