LAUSR

Numerical control (NC) turret constitutes a key component for NC lathe. Due to its strategic role in the successful machining, it is important to monitor the NC turret conditions to recognize the need for maintenance, thereby avoiding unplanned downtime. Signal segmentation is a very basic work in such a multi-action system condition monitoring. This paper borrows the idea of speech segmentation and proposes a simple vibration segmentation approach to obtain action-known segments. In this approach, we formulate a segment optimization problem by a new fitness function, and present a three-layer genetic algorithm to decouple the problem. The advantage of the approach is the little requirement of expert knowledge, which is friendly to users. The case study of NC turret from the experiment bench uses fourfold cross-validation to indicate that the approach has a good performance on segment production, action recognition, and the two combination. The ablation tests demonstrate the contribution of the component of the approach.