Deep learning has been widely used in the field of fault diagnosis, among which convolution neural networks (CNN) and long short-term memory networks (LSTM) are typical models. As a common method of multi-sensor signal fault diagnosis, CNN can obtain good diagnosis effect, but it can not diagnose unknown composite fault. In order to solve this problem, a CNN-LSTM-FCM(fuzzy C-means) model was proposed. Since LSTM was more sensitive to the time signals with the connection between the front and the back, it was combined with CNN to realize the diagnosis of unknown signals. The decoupling of composite fault was achieved through the probability classification output. The CNN-LSTM-FCM model itself had optimized parameter design, which further improved the diagnosis accuracy. The chemical process fault measurement data was used for experiments, and the results showed that the diagnostic accuracy of CNN-LSTM-FCM model could reach 97.15%, which was superior to both CNN model and LSTM model, thus having a high application value in fault diagnosis.