Real-time tool condition monitoring with the internet of things and machine learning algorithms

Machine Learning and the Internet of Things (IoT) has an essential role in machine health monitoring and maintenance. An intelligent system that can predict the tool condition is created by collecting data from various sensors. With the rich insights derived from the collected data and integrated through the IoT system, predictive analytics is gaining prominence in Tool Condition Monitoring systems (TCMs). The research reported in this paper uses an accelerometer sensor and a microphone positioned on the spindle head to collect vibration and sound signals associated with eight different tool conditions. Then, the statistical and wavelet features are extracted from the collected signals. The T-test is employed to identify the dominant features. These dominant features are used to train the various machine learning classifiers to predict the tool condition. Finally, the statistical features from sound and vibration are combined and trained to explore sensor fusion techniques. The Ensemble Trees is selected as the best-performing algorithm for online IoT-based TCMs. This model predicts tool conditions with an accuracy of 98% and provides alert messages and indications through Thingspeak. By integrating multiple sensors and employing advanced analytical techniques, this approach aims to improve the accuracy and effectiveness of TCMs, enabling proactive maintenance and optimization of tool usage.