An End-to-End model based on CNN-LSTM for Industrial Fault Diagnosis and Prognosis

As the “Industry 4.0” and “Made in China 2025” bringing forward, the researches on the combination of information technology and manufacturing industry become more and more important to realize intelligent manufacturing. One of those researches is industrial fault diagnosis and prognosis of PHM (Prognostics Health Management). Most of the industrial data are time series with multiple features collected from different sensors and a good feature presentation is very important for accurate results. In order to extract features automatically and get accurate results, we propose an end-to-end method, CNN-LSTM, for industrial data based on deep learning and transfer learning, where CNN extracts features automatically and LSTM analyzes the new feature sequences from CNN. We evaluate our method on the dataset of wind turbines blade-icing problem and get good results compared with other models. It shows we propose an effective end-to-end method for industrial fault diagnosis and prognosis.