Predicting Faults in Robotic Arms Using Machine Learning: A Study On Passive Detection Methods

Abstract

Robotic arms are widely used in industries and keeping them reliable is very important. Traditional maintenance often waits until problems become visible, but new methods use data from sensors to detect faults early. This article reviews research from 2020 to 2025 that studies how signals from sensors such as accelerometers, gyroscopes, and motor currents, can be used with machine learning to find faults. The main types of faults like gear, bearing, motor, and collision issues and the techniques used to detect them are compared. The review shows that deep learning methods are becoming popular, but classical approaches and hybrid models are still useful. Overall, this study highlights how passive sensors combined with machine learning can improve predictive maintenance and make robotic arms safer and more dependable.