基于多模型的航空发动机传感器混合故障诊断方法

Hybrid Fault Diagnosis Method for Aero-Engine Sensor Based on Multiple Model

关键词:  航空发动机  传感器  多模型  故障诊断  长短期记忆网络  混合卡尔曼滤波

Key words:  Aero-engine  Sensor  Multiple model  Fault diagnosis  Long short-term memory  Hybrid Kalman filter

摘要:

本文在多模型架构下，提出一种航空发动机传感器在线混合故障检测与隔离算法。利用长短期记忆网络逼近航空发动机建模误差、健康参数变化、过程噪声和测量噪声等不确定性源引起的真实发动机与机载模型之间的偏差。将传感器测量输出与不确定性值的偏差用于一种基于多模型的混合卡尔曼滤波器组算法中，利用贝叶斯方法计算每个传感器在健康模式和不同故障模式下的条件概率，然后根据最大概率准则进行传感器故障检测与隔离，克服了阈值难以选取的问题。针对某型涡扇发动机传感器发生偏置故障、漂移故障和间歇性故障的情形进行仿真验证，并对比了不同传感器之间的检测与隔离精度。结果表明：所提出的方法可以在更高水平的退化下诊断出发动机传感器常见的故障，混合方法对不同不确定性源具有鲁棒性。

Abstract:

Under the multi-model architecture， an online hybrid fault detection and isolation algorithm for aero-engine sensor is proposed. The long short-term memory network is used to approximate the deviation between the real engine and the on-board engine model caused by the uncertainty sources such as aero-engine modelling error， health parameter changes， process noise and measurement noise. The deviation between the sensor measurement output and the uncertainty value is used in a hybrid Kalman filters algorithm based on multiple model method. The Bayesian approach is used to calculate the conditional probability of each sensor under health mode and different fault modes， and then the maximum probability criterion is used for sensor fault detection and isolation to overcome the problem of difficult threshold selection. The simulation is carried out to verify the sensor bias fault， drift fault and intermittent fault of a turbofan engine， and the detection and isolation accuracy of different sensors is compared. The results show that the proposed method can diagnose common faults of aero-engine sensor at higher levels of degradation， and the hybrid method is robust to different sources of uncertainty.