REPOSITORY > RESULTS

Abstract

In modern manufacturing, cutting tools play a pivotal role in ensuring product quality,
process reliability, and cost efficiency. Premature or delayed tool replacement can lead to
unnecessary expenses, compromise product quality, machine safety and production
stability. Conventional approaches, such as fixed replacement intervals or visual
inspections, often fail to provide accurate insights into the actual tool condition, especially
in high-throughput industrial environments.
This doctoral dissertation proposes a comprehensive approach to tool life prediction
based on vibration analysis using triaxial accelerometer, combined with machine learning
algorithms. The developed system enables real-time monitoring of tool wear, allowing
predictive decisions that minimize production downtime and enhance overall process
efficiency. The solution integrates sensor signal acquisition, feature extraction, model
training, and operator-friendly visualization to support decision-making in real-world
machining conditions.
Experimental validation conducted in an industrial setting demonstrates a strong
correlation between vibration patterns and the tool’s wear state. The predictive models
achieve high accuracy, enabling timely alerts for tool replacement and supporting the
optimization of machining parameters. The system is designed to be cost-effective,
adaptable to various machines and processes, and easily integrated into existing
manufacturing infrastructure.
The findings contribute to the development of intelligent, sustainable manufacturing
systems and offer a practical tool for increasing productivity, reducing material waste, and
enhancing operational decision-making through data-driven insights.