Machine Learning for Sensorless Temperature Estimation of a BLDC Motor (ADAS)

BLDC motors find applications in various industries, such as consumer electronics, electric vehicles, and household appliances. Effective temperature management is crucial to maximizing the performance and lifetime of motors. Traditional methods for monitoring temperatures are based on physical sensors. Such sensors have a natural tendency to degrade over time, making them complex and costly. The authors, therefore, propose a novel, sensorless approach towards temperature estimation through machine learning algorithms. With input current, voltage, speed, and time data analyzed XGBoost, Random Forest, Support Vector Regressor, and Linear Regression models predict the motor temperature accurately in real time. Such models identify the patterns within large operational datasets, ensuring precise thermal estimations without physical sensors. The results of the research indicate that sensors can be replaced by machine learning, thereby making motor control systems simpler more reliable, and more durable. This has great implications in the field of enhancement of efficiency in motor functions and extension through advanced techniques in temperature predictions.