Force control of grinding process based on frequency analysis

TL;DR

This paper introduces a novel force estimation method during grinding that leverages frequency analysis via Mel spectrograms and neural networks, effectively handling hysteresis drift and noise.

Contribution

It presents a new external force estimation approach using frequency analysis and neural networks, improving robustness against drift in grinding processes.

Findings

Frequency-based analysis outperforms amplitude-based methods under noisy conditions.

The proposed method is robust against hysteresis-induced drift.

Neural network integration enhances force estimation accuracy.

Abstract

Hysteresis-induced drift is a major issue in the detection of force induced during grinding and cutting operations. In this paper, we propose an external force estimation method based on the Mel spectrogram of the force obtained from a force sensor. We focus on the frequent strong correlation between the vibration frequency and the external force in operations with periodic vibrations. The frequency information is found to be more effective for an accurate force estimation than the amplitude in cases with large noise caused by vibration. We experimentally demonstrate that the force estimation method that combines the Mel spectrogram with a neural network is robust against drift.