Machine Learning Based Diagnosis of Shaft Failures In Heavy Vehicles

The study focuses on the use of ML techniques in diagnosing failures in shafts within mechanical systems while complementary methods to traditional fault detection techniques based on manual inspection and signal processing, methods including but not limited to SVM and NN. These models prove fruitful in the classification of main types of shaft failure: misalignment, overload, and fatigue failure with the help of important signal types such as vibration, torque, and speed using techniques of feature extraction such as RMS and FFT. The work has obtained an accuracy of more than 93.97% in classification with a comparison of the efficacy of SVM and NN models. It culminates into the conclusive findings on the significant improvements ML makes over traditional techniques in the predictive accuracy and reliability optimized to maintenance of mechanical systems. Future work will focus on integrating these models directly with IoT devices in real-time using edge computing to further improve on the real-time effectiveness of predictive maintenance.