An IOT-based Multi-Sensor Digital Twin for Predictive Maintenance of Industrial Motors

More factories need maintenance that's reliable and cheap. That need pushes factories to use Digital Twin technology with Artificial Intelligence (AI). We present a multisensor AI-driven Digital Twin design for maintenance of a 37 kW (50 HP) industrial induction motor. The Digital Twin design uses vibration sensors, temperature sensors and current sensors to record the induction motor’s wear, thermal wear and electrical wear. I set up the Masibus scanner to read sensor data in real time. The Masibus 85XX+ universal scanner then sends the data using Modbus RTU over an RS-485 link. I built a Digital Twin application in a Python environment. The Digital Twin application shows the operating parameters as they happen. The Digital Twin application also runs calculations. I apply feature engineering techniques to the sensor data. Feature engineering techniques add features and rate-of-change indicators. Feature engineering techniques turn the data into inputs for a Random Forest regression model. The Random Forest regression model predicts the Remaining Life (RUL). The Random Forest regression model gives a Remaining Life (RUL) prediction that I can use to plan maintenance. I tested the system performance against ISO 10816 Class II vibration standards. I found that the proposed model gives a prediction accuracy with an R Squared (R2) of 0.91 and a Mean Absolute Error (MAE) of 2.45 hours. I see that the multi-sensor approach makes the system performance stable and reliable compared to single-parameter monitoring techniques. I also integrated the system performance with the ThingSpeak IoT platform. The integration lets the ThingSpeak IoT platform provide monitoring and historical data logging. The integration makes the framework able to grow and fit Industry 4.0 applications.