基于GA-AANN神经网络的SDQ算法的航空发动机传感器数据预处理

Aero-Engine Sensor Data Preprocessing Based on SDQAlgorithm of GA-AANN Neural Network

关键词:  航空发动机传感器  故障诊断  SDQ算法  遗传算法  AANN神经网络

Key words:  Aero-engine sensor  Fault diagnosis  SDQ algorithm  Genetic algorithm  AANN neural network

摘要:

为实现对输入健康管理系统的航空发动机传感器数据进行数据鉴定、故障诊断以及去除噪声信号干扰，提出了一种航空发动机传感器数据预处理方法。针对双通道传感器航空涡扇发动机，搭建了以合理性检验模块和解析冗余检验模块为主要内容的SDQ算法模型，利用遗传算法优化的AANN神经网络实现传感器的解析冗余检验。采用蒙特卡罗仿真方法，将改进的SDQ算法与一种基于最小二乘法的SDQ算法进行对比仿真验证。结果表明，本文提出的SDQ算法在发动机稳态条件下对阶跃故障和漂移故障隔离的平均正确率分别提高了1.7%和19.1%，在发动机动态条件下对阶跃故障和漂移故障隔离的平均正确率分别提高了12.5%和33.8%。且在多传感器故障诊断和除噪方面性能优异，处理后的传感器信号平均信噪比提高了8.27dB。

Abstract:

In order to realize the data identification， fault diagnosis and noise interference of the aero-engine sensor data of the input health management system， a method of aero-engine sensor data preprocessing was proposed. Aiming at the air turbofan engine of dual-channel sensor， a SDQ algorithm model with reasonableness checks module and analytical redundancy checks module as the main content was established， and the AANN neural network optimized by genetic algorithm was used to realize the analytical redundancy checks of the sensor. The improved SDQ algorithm was compared with a SDQ algorithm based on the least squares method using the Monte Carlo simulation method. The simulation results prove that the average correct rate of step fault and drift fault isolation of the improved SDQ algorithm increased by 1.7% and 19.1% respectively when the engine in steady states， and the average correct rate of step fault and drift fault isolation of the improved SDQ algorithm increased by 12.5% and 33.8% respectively when the engine in dynamic states. The algorithm also has excellent performance in multi-sensor fault diagnosis and noise reduction， and the average signal-to-noise ratio of the processed sensor signal increased by 8.27dB.