一种基于时序预报神经网络的故障预报方法及其应用

A FAULT PREDICTION METHOD VIA TIME SERIES PREDICTIVE NEURAL NETS AND ITS APPLICIATION

提出一种基于时序预报神经网络的工业过程故障预报方法，同时给出了描述神经网络预报和外推能力的表达方式，并以氯碱电解工艺的现场数据验证了这种故障预报方法的有效性．实验结果表明，该方法可成功地用以实现氯中含氢的２４小时预报．

A novel method of fault prediction in industrial.systems which is based on the predictive neural network trained using time series data, is proposed in this paper.prediThen descriptions of the ability of prediction or generalization of time series predictive neural networks are discussed. Also we apply the new approach to the prediction of hydrogen-in-chlorine in electrolytic soda industry to prevent an explosion caused by the mixture of them. The results show that the new method is effective for 24 hour prediction.