Predictive maintenance forecasting in asynchronous motors with supervised machine learning

Asynchronous motors (ASM) are widely used in the most critical processes in industry. Failure of ASMs in these processes is not desired as it will cause cost and production losses. To prevent or minimize this, it is important to make predictive maintenance estimates with continuous monitoring of ASMs. In this study, it is aimed to make predictive maintenance estimation of ASMs using supervised machine learning (SML). For this purpose, the temperature and vibration variables of the ASM were transferred to the Atmega328P microcontroller (MCU) with a positive temperature constant (PTC) thermistor and MPU6050 accelerometer. First, the temperature and vibration information of a robust ASM was taught to the Atmega328P MCU with an SML-based algorithm. Then, when the taught information exceeds the maximum 20% value, the information to generate a warning is entered. Then, to test the SML algorithm, vibration and temperatures at different values were applied to the ASM and the operation of the SML-based ASM predictive maintenance system was monitored. The results of continuous comparison of the information taught to the SML with the incoming information are sent to the mobile device in the form of alerts by Bluetooth communication by the Atmega328P MCU. Thanks to the control software of this Android-based mobile device, it is ensured that the ASM is stopped, started and the information flow to the nearby operator is continuous. Thus, making an SML-based ASM predictive maintenance estimation and alerting operators with a mobile device has been successfully accomplished.