TORNADO燃机故障诊断的神经网络模型研究

为改善单神经网络收敛速度慢的问题,本文采用组合神经网络建模方法,建立了燃机性能仿真网络模型,并进行实验验证。建模过程中,将建模对象划分为三个相对独立的子网络,利用获取的实验数据组成训练域对网络进行训练,建立起一套可用于燃气轮机控制系统仿真及故障诊断的组合神经网络模型。实验结果表明:该模型平均输出误差约为3%-6%,计算时间小于100ms,可用于基于模型的燃气轮机诊断系统。

Study of Neural Network Model for TORNADO Gas Turbine Fault Diagnosis

In order to improve the problem that single neural network model has slow convergence speed,we adopt combination neural network modeling method. We established a network model of the gas turbine performance simulation,and it was validated by experiment. In the course of modeling,the modeling object is divided into three relatively independent sub-networks. We use the training domain obtained by experimental data to train the network,and set up a combined neural network model that can be used for gas turbine control system simulation and fault diagnosis. The experimental results show that the average output error of this model is about 3% -6%,the calculation time is less than 100ms,and it can be used for model-based gas turbine diagnosis system.