基于T-S模糊KPCA模型的分布式控制系统传感器故障诊断

Fault Diagnosis of Sensors in Distributed Control System Based on T-S Fuzzy KPCA Model

关键词:  航空发动机  分布式控制系统  T-S模糊模型  核主元分析法  传感器故障诊断

Key words:  Aero-engine  Distributed control system  T-S fuzzy model  Kernel principal component analysis  Sensor fault diagnosis

摘要:

为减小航空发动机多工况的工作特性和分布式控制系统非线性网络环境对故障诊断系统的影响，针对航空发动机分布式控制系统，提出一种基于T-S模糊KPCA模型的传感器故障诊断方法。首先采用C均值模糊聚类法，以油门杆角度为样本标签，对样本空间进行模糊分类，再通过模糊相似矩阵剔除各样本子空间的野值点；其次建立标称工况的KPCA模型，并利用训练样本对非标称工况的隶属度函数进行辨识，得到全工况T-S模糊KPCA模型；最后利用统计量[T2]和SPE对传感器故障进行检测，并采用数据重构方法对故障传感器进行隔离定位。仿真结果表明该方法对发动机的任意稳定工况具有自适应能力，能够在非线性网络环境下对正常样本和故障样本保持较低的虚警率和漏报率。当多个传感器同时发生故障时，能够准确找到故障源，实现对故障传感器的隔离。

Abstract:

To reduce the effect of multi-operating conditions characteristic of aero-engine and the nonlinear network environment of distributed control system(DCS)on fault diagnosis system，a sensor fault diagnosis method for aero-engine distributed control system based on T-S fuzzy KPCA model was proposed. The fuzzy C-means clustering algorithm was applied to classify the sample space，in which the throttle lever angle was taken as the sample label. Then the outlier points were removed using the fuzzy similar matrix. Secondly，the KPCA model of nominal operating conditions was built. And the membership function of non-nominal operating conditions was identified with the training samples，then the T-S fuzzy KPCA model of full condition was built. Finally，the statistics [T2] and SPE were applied to detect the sensors fault. The faulted sensors were separated and located by data reconstruction. Simulation result shows that the proposed method is adaptive to any stable operating condition of aero-engine. Lower false alarm rate and missing alarm rate are retained under the nonlinear network conditions. When multiple sensors go wrong synchronously，this method could find and isolate the fault sensors.