In order to monitor the slate of complex industrial process in real time and predict the fault trend accurately, this paperpresents a fault prediction method based on denoising auto eneoder(DAE)and temporal convolutional network (TCN).Firstly,therandom forest algorithm is used to filter out the features related to faults. Then, the nonlinear features of input data are extracted and theoriginal features of input data are reconstructed, and the squared prediction error(SPE)statistics is established based on thereconstruetion error to reflect the slate characteristies of the faults. Finally, considering that the derivative of ReL.U activation funetion inthe residual module of TCN is zero in the negative interval, which may cause certain neurons to fail to activate, a Swish activationfunetion and filter response normalization-based temporal convolutional network (SFTCN) is proposed. By construeting the obtained SPEinto time series, the SPE predietion can he realized based on the SFTCN. Experiments are conducted with the data of Tennessee Eastman (TE) process and the life-ceycle vibration data of rolling bearings measured by the center for intelligent maintenance systems of theUniversity of Michigan. Results show that eompared with the unmodified TCN, the average absolute pereentage eror of the proposedmethod is redueed by at least 20.9%, which has high application value.