Predictive Maintenance: detect Faults from Sensors with CRNN and Spectrograms

Predictive Maintenance is an active field of research in every area. Especially in recent years, a great boom is registered in Machine Learning solutions. Some business invests a great amount of money to develop solutions which can predict in advance a possible fault occurring in the business activities. An interesting domain for these studies is the IoT industry, where we have sensors which monitor the working status of a machine or a particular engine part. The classical approach for this kind of problem involves usually the adoption of time series models in conjunction with signal process techniques which enable us to extract value from high-frequency data.