Intelligent Embedded Load Detection at the Edge on Industry 4.0 Powertrains Applications

In the context of Industry 4.0, there has been great focus in developing intelligent sensors. Deploying them, condition monitoring and predictive maintenance have become feasible solutions to minimize operating and maintenance costs while also increasing safety. A critical aspect is the applied load to the supervised machinery system. Vibration data can be used to determine the current condition, but this needs signal processing specially developed and adapted to the monitored machine part for feature extraction. Artificial intelligence (AI) can, on one hand, simplify the development of such special purpose processing and on another hand be used to monitor and classify machine conditions by learning features directly from data. By bringing the AI computation as close as possible to the sensor (Edge-AI), data bandwidth can be minimized, system scalability and responsiveness can be improved and real-time requirements can be fulfilled. This work describes how Edge-AI on a STM32-bit microcontroller can be implemented. Our experimental setup demonstrates how AI can be effectively used to detect and classify the load on a powertrain. In order to do this, we use a MEMS capacity accelerometer to sense vibrations of the system. Also, this work demonstrates how Deep Neural Networks (DNN) for signal classification are build and trained by using an open-source deep learning framework and how the code library for the microcontroller is automatically generated afterwards by using STM32Cube. AI toolchain. We compare the classification accuracy of a memory compressed DNN against an uncompressed DNN.