Machine tool operation multi-source sensing data multi-domain deep mining method and system

Abstract

The application relates to the field of industrial manufacturing monitoring, and discloses a machine tool operation multi-source sensing data multi-domain deep mining method and system. The method comprises the following steps: collecting vibration, acoustic emission and force signals through a multi-source sensor and extracting features; segmenting the features to generate a preliminary data set; analyzing the preliminary data set and determining scene classification labels; optimizing the sampling window and filtering parameters according to the scene classification labels, obtaining an optimization threshold and a window length; denoising based on the optimization parameters to obtain deep features; mining and weighting the features by using deep learning and an attention mechanism to generate refined features; adjusting the feature weights through a feedback mechanism to generate an abnormal monitoring feature set. The application realizes scene adaptive processing and deep correlation mining of multi-source signals, solves the problems of inaccurate feature extraction, parameter solidification and insufficient utilization of multi-source information, and improves the accuracy, adaptability and early warning capability of the monitoring system.

Images