An experimental study of multi-sensor tool wear monitoring and its application to predictive maintenance

Wear in cutting tools is a critical issue that can lead to reduced product quality, increased production costs, and unexpected downtime. To mitigate these challenges, the implementation of tool wear monitoring systems and predictive maintenance strategies has gained significant attention in recent years. Early detection or prediction of tool wear is vital to optimize tool life and maintain the manufacturing processes efficiently. This paper presents a method to determine the tool wear progression based on the collaboration of direct and indirect monitoring techniques. By analyzing the monitoring of data from force, vibration, and current sensors to estimate the tool wear state, and correlating this information with a photographic database of the tool wear progression used to create an image recognition system, it is possible to classify the tool wear at any moment into three states: Good, Moderate, and Worn. A case study was conducted to test the advantages and limitations of the proposed method. The case study also shows that the improvement of the prediction of the tool wear state might be useful in the decision-making of whether the tool life can be extended, or the tool must be replaced, as well as in the detection of anomalies during the machining process, aiming its improvement and the reduction of operational costs.