Early Fault Warning of Induced Draft Fans Based on 1DCNN-GRU Model and Bayesian Test
In order to realize the condition monitoring and fault warning of induced draft fans in power plants,one-dimensional convolutional neural network(1DCNN)and gate recurrent unit(GRU)are combined,and a bearing condition monitoring and warning model based on 1DCNN-GRU is proposed.Firstly,the correlation variables are evaluated by distance correlation coefficient,and the modeling variables are selected by minimal-redundancy-maximal-relevance algorithm.1DCNN is used for feature extraction,and the prediction model of induced draft fan bearing vibration velocity is obtained by GRU.Then,Bayesian test is introduced to make full use of prior information to capture the early creep of the fault signal and timely identify unit anomalies to realize early warning.Finally,an induced draft fan of a 1 000 MW unit is taken as an example to verify the model.The experimental results show that the prediction accuracy of the proposed model is better than that of other neural network models,and the early warning method is more timely and more effective than the common sliding window method.
Fault warninginduced draft fanneural networkBayesian test
万方数据
维普
