CHATTER DETECTION SYSTEM AND METHODS IN SOUND SENSOR-BASED MACHINING PROCESS USING SUPPORT VECTOR MACHINE

Abstract

This invention discloses a system and method for detecting chatter in a lathe machining process based on a microphone sound sensor using a support vector machine (SVM) algorithm. The system consists of a dynamic cardioid microphone sound sensor with a response range of 500–20,000 Hz, while an electronic panel processes the signal through a sound card, a Raspberry Pi microcontroller, and a display / HMI. The signal processing method includes two-stage filtering using a band-pass filter and a 4th-order causal butterworth filter, followed by time–frequency analysis using a Short-Time Fourier Transform (STFT) to produce a local spectral representation. From each STFT frame, three frequency features are extracted, namely Power Spectral Density (PSD), 85% spectral roll-off, and Shannon entropy, which are then combined as SVM input to distinguish chatter and non-chatter conditions.The test results show that SVM provides the best performance with an F1-score of 0.9893 and an accuracy of 0.9805, surpassing other models such as Logistic Regression, Passive-Aggressive, Gaussian Naive Bayes, and Linear Perceptron. This system is suitable for application as a chatter detection solution with stable performance and efficient computation, to support predictive maintenance on industrial lathes.