基于多域特征优化的航空发动机传感器智能故障诊断

Intelligent Fault Diagnosis of Aeroengine Sensor Based on Optimized Multi-Domain Features

关键词:  航空发动机  传感器故障诊断  深度置信网络（DBN）  改进亨利气体溶解度优化算法（BHGSO）  多域特征

Key words:  Aeroengine  Sensor fault diagnosis  Deep belief network（DBN）  Boosted Henry gas solubility optimization algorithm （BHGSO）  Multi-domain feature

摘要:

为解决对航空发动机传感器故障诊断时单域特征反映故障信息不全面问题，提出了一种基于优化的多域特征进行智能故障诊断的方法。该方法提取了传感器信号的时、频域特征和形态信息，共同组成多域特征，从多维度描述传感器的健康状况；并提出了一种新的元启发式算法—改进亨利气体溶解度优化算法（Boosted Henry gas solubility optimization，BHGSO）进行特征选择，尽量以最低维度但知识丰富的高品质信息来训练故障识别模型，以减轻计算负担，并提高诊断可靠性；最后将特征向量作为传感器的健康指标，基于深度置信网络（Deep belief network，DBN）实现智能故障诊断。仿真结果表明，该研究提出的方法能够对航空发动机传感器进行有效的故障诊断，且具有较高的准确度和较小的计算负担。

Abstract:

In order to solve the problem of incomplete fault information reflected by single-domain features for aeroengine sensor fault diagnosis， a method based on optimized multi-domain features for intelligent fault diagnosis is proposed. The method extracts multi-domain features including time domain， frequency domain and morphological information， which together form multi-domain features to describe the health condition of the sensor from multiple dimensions. Afterwards， a new meta-heuristic algorithm， the boosted Henry gas solubility optimization （BHGSO） algorithm is proposed for feature selection to train the fault identification model with the lowest dimensional but knowledge-rich high quality feature information as much as possible to reduce the computational burden. Finally， intelligent fault diagnosis is performed using deep belief network （DBN） based on the feature vectors， which are used as indicators of the sensor’s health. The simulation results show that the proposed method can effectively diagnose faults in aeroengine sensors with high accuracy and low computational burden.