神经网络在化工过程故障诊断中的应用

针对现代复杂的化工生产过程，提出一种基于神经网络的故障诊断方法。并分别将典型的BP算法和改进后的BP算法用于TE（Tennessee Eastman）模型的故障诊断中。经过诊断结果的比较，得出标准的BP算法在实际应用中具有收敛速度慢等缺点；自适应学习速率动量梯度下降的BP算法以及用L-M（Levenberg-Marquardt）法先对BP网络进行优化的BP算法具有收敛速度快、不易陷入局部极小值等优点，其中又以L-M优化BP算法效果最好。结合rIE模型的仿真结果可以看出，L-M优化BP算法在工业实际中具有很大的优势。

Application of Neural Networks to Chemical Fault Diagnosis

The modern complicated chemical production is discussed and a method based on neural networks in fault diagnosis is proposed.The typical BP algorithm and improved BP algorithm are applied to the fault diagnosis of Tennessee Eastman(TE) model.The comparing results show that the typical BP ANN has the disadvantages of low training speed,and the gradient descent momentum and adaptive learning rate BP ANN and the LevenbergMarquardt BP ANN optimizing the network firstly have advantages of high training speed,hardly trapping into the partial minimum,and etc.The simulation of TE model shows that this method has great power in the industry.