In-process monitoring of the ultraprecision machining process with convolution neural networks

ABSTRACT

In-process monitoring and quality control are the most critical aspects of the manufacturing industry, especially in ultra-precision machining (UPM) at an industrial scale. However, in-process ensuring product quality has been difficult, as any subtle change in the process influences the UPM process dynamics and the process outcome. In order to meet the increasingly soaring demand for precision components, intelligent monitoring of the machining process is essentially important and much needed. Capturing complex signal patterns through conventional signal processing for the UPM process is often challenging due to the comparably high noise levels in the industrial environment. Signals obtained during UPM are inherent transients and non-stationary, necessitating extensive and accurate features for classification. Accurate detection of anomalies may allow for quick corrective actions, reducing the degree of damage. Earlier research revealed multi-sensor analysis, which yields richer signal feature information, but the unavoidable sensor failure in conjunction with heterogeneous sensing made it challenging. In order to address the challenges, this paper investigates the feasibility of convolution neural network (CNN) for classifying abnormal and normal machining in the UPM process. The vibrational signals obtained from B&J 4533-B accelerometer during diamond turning are transformed into time-frequency-based log-spectrogram images. These images are classified using CNN, and the results show that a proposed convolutional neural network algorithm has demonstrated an accuracy of 85.92% in classifying images and thus the corresponding in-process machining status.

KEYWORDS:

• In-process monitoring
• convolutional neural network (CNN)
• time-frequency analysis
• vibrational signal
• ultra-precision machining

Disclosure statement

No potential conflict of interest was reported by the author(s).

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.