Tool wear monitoring for cavity milling based on vibration singularity analysis and stacked LSTM

Tool wear monitoring (TWM) system plays an important role since it ensures the accuracy of manufacturing and workpiece quality, especially in aerospace manufacturing. Due to the challenge of various paths from the milling process of large cavity-like structural parts and impact of tool wear on the surface quality, there remains an urgent need for a high-precision and robust TWM approach. This article addresses this issue by employing a stacked network in conjunction with a feature extraction method in which it combines vibration singularity analysis with correlation analysis. The singularity of the original vibration signal, estimated by the Holder exponent (HE), is analyzed to eliminate the influence of the milling path, and the sensitive features based on HEs are extracted and reduced via Pearson's correlation coefficient (PCC) analysis. Subsequently, a stacked long short-term memory neural network (LSTM) trained by these features has been applied to estimate tool wear, which is verified by a dataset obtained from the processing site. Experimental results indicate that the proposed method has improved the accuracy of tool wear prediction, which outperforms the developed methods such as LSTM, bi-direction LSTM (BiLSTM) and its stacked model, partial least squares regression (PLSR) model, and support vector regression (SVR) model optimized by the whale optimization algorithm (WOA). Meanwhile, this method lays the foundation for using vibration signal to monitor tool wear in cavity milling.

Automated summary

This article introduces a tool wear monitoring system that combines stacked network and feature extraction methods to improve the accuracy of tool wear prediction through vibration singularity analysis and correlation analysis.