基于多残差描述的鲁棒故障诊断方法

The Robust Fault Diagnosis Method Based on Multi-error Description

论文采用自适应阈值、多残差描述故障特征等方法研究鲁棒故障诊断问题。该方法首先采用多层感知机网络建立系统正常工况解析模型,根据不同的故障模式在过程操作单元和控制回路之间故障传播的途径不同,采用多残差描述方法和基于加权移动残差设计各种历史故障特征矩阵,通过与历史故障特征矩阵的匹配度快速诊断故障源。最后利用DAMADICS阀门基准实验平台的19种不同故障模式验证提出的故障检测与诊断方法的有效性。

This paper focuses on the problem of designing a robust fault diagnosis scheme with neural networks, adaptive threshold and fault feature description. The system models are built using multi-layer perception neural network. Then fault feature is described by multi-residual according to fault spread track, and robust fault diagnosis method based on weighted sliding residual is developed. Finally, simulation result of 19 kinds valve faults on prove the proposed method is effective.