基于模糊神经网络的气体密封故障诊断

Fault Diagnosis of Gas Seal Based on Fuzzy Neural Network

针对传统的依靠单一参数对密封状态监测的不足,利用模糊神经网络处理问题的能力,综合多种密封参数的信息,建立针对密封状态监测的模糊神经网络系统,提出基于模糊神经网络的故障诊断的方法。该方法确定泄漏量、端面温度、气膜厚度、阻封气泄漏量作为密封的监测参数,并确定各个参数的隶属函数;通过大量的历史数据、MAT-LAB模拟数据和专家知识分析得到各个特征参数值和修正值,建立5种密封状态的输出模式;通过隶属函数实现输入样本的模糊化;通过MATLAB编程来设计、优化神经网络结构,利用历史数据训练神经网络。通过实例分析验证了建立的模糊神经网络的实效性。

According to the defect of seal condition monitoring depending on single parameter, the seal condition monitoring system with fuzzy neural network（FNN） which well applies to setting information problems was established regarding the different information of seal parameter. A fault diagnosis method based on FNN was put forward. Spillage, face temperature,gas film thickness and quench gas spillage that are given the membership functions were selected as the seal monitoring parameter. With sufficient historical data, the MATLAB simulation data and expert knowledge, the characteristic parameters and modified value were defined, the output models of five seal conditions were established. The membership functions were applied to transfer input data into fuzzy data. The design, optimization and training of the neural network were based on the MATLAB program and the historical data. The fuzzy neural network was tested by the original data and proved to be effective and reliable.