A feature extraction method based on machine tool spindle

By establishing a finite element model on the machine tool spindle and a sensor arrangement driven by modal perception, and optimizing VMD by combining a physical consistency evaluation function, the problems of sensor redundancy and modal loss are solved, and efficient signal processing and feature extraction are achieved.

CN121542711BActive Publication Date: 2026-05-26ZHEJIANG UNIV
View PDF 4 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
ZHEJIANG UNIV
Filing Date
2026-01-20
Publication Date
2026-05-26

AI Technical Summary

Technical Problem

In existing technologies, sensor placement relies on experience, resulting in significant redundancy that is difficult to adaptively eliminate. The VMD method lacks physical constraints, leading to high signal processing complexity and loss of modal information.

Method used

By establishing a whole-machine finite element model of the machine tool spindle, selecting sensor positions based on modal vibration modes, and combining modal perception and physical consistency evaluation functions, redundant signals are adaptively eliminated, the VMD decomposition order is optimized, and low-dimensional physical information feature vectors are extracted.

Benefits of technology

It realizes physical modal interpretation and redundancy elimination of sensor deployment, reduces costs, improves the accuracy and interpretability of signal processing, and solves the problems of modal over-splitting and key mode loss.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121542711B_ABST
    Figure CN121542711B_ABST
Patent Text Reader

Abstract

This invention discloses a feature extraction method based on machine tool spindles, relating to the field of condition monitoring, including: (1) establishing a whole-machine finite element model and obtaining target modal frequencies and mode shapes; (2) implementing a multi-sensor layout strategy based on modal perception and spatial constraints; (3) acquiring multi-field signals, eliminating non-steady-state data and performing asynchronous time-integration; (4) constructing a channel evaluation function to automatically eliminate redundant and insensitive channels; (5) constructing a physical consistency evaluation function, using inherent frequency constraints for VMD decomposition and adaptively selecting the optimal order; and (6) combining modal perception weighted fusion features to construct a physical information feature vector. This invention solves the problems of blind sensor layout and poor feature interpretability by guiding data processing through physical priors, significantly improving the accuracy of condition representation and laying the foundation for constructing a machine tool spindle system condition monitoring model based on machine learning / deep learning.
Need to check novelty before this filing date? Find Prior Art