Audio Based Detection of Saw Blade Sharpness Using Machine Learning

Any mechanical system like a drill, a cutting machine, vehicle engine can subject to faults during operation. These machines are intended to work within a tolerance range. Therefore, even though some of the faults may not stop the system from working, if these faults persist, they can lead to the breakdown of the system. Primarily, the operating sound of the mechanical machine indicates the status of the machine. Therefore, audio signals can be employed for the detection of any impending fault. The goal of this project is to develop machine learning algorithms for system consists of a circular saw blade used for cutting the printed circuit boards (PCB). This blade gets blunt after every few weeks. As the sharpness of the blade decreases, the quality of the cut deteriorates, and the blade starts to consume more power. When the blade is utilized completely, there is a slight change in the blade sound. Using these sound samples, a machine learning algorithm is developed to classify the audio; either blade is new or has reached the end of its life cycle. The machine learning classifiers are also compared with a classifier programmed by using FFT (Fast Fourier Transform) method.