A Fault Diagnosis Platform of Actuators on Embedded IoT Microcontrollers

In the process of monitoring and fault diagnosis of complex electromechanical equipment, the close coupling between the fault diagnosis process and the front-end equipment can effectively reduce the occurrence of serious faults and significantly improve the economic benefits. In this paper, an Internet of Things (IoT) framework for monitoring and diagnosing industrial equipment is designed and implemented for complex electromechanical equipment running in real-time. All the procedures are physically implemented on a hardware prototype, which includes hardware selection, software configuration, transplanting of machine learning (ML) model and data communication. The framework of the physical platform is universal and flexible. It can be deployed in various monitoring scenarios, and flexibly customize the deployed artificial intelligence (AI) models according to their applications. Three typical machine learning algorithms of SVM, ANN and LSTM models are transplanted to STM32 MCU to compare the results. Finally, the proposed method is experimentally validated on NASA Electro-mechanical actuators (EMAs) data set.