A physical knowledge-enhanced feature representation method for fault diagnosis

CN121706036BActive Publication Date: 2026-05-26QILU UNIVERSITY OF TECHNOLOGY (SHANDONG ACADEMY OF SCIENCES) +1

View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office: CN · China
Patent Type: Patents(China)
Current Assignee / Owner: QILU UNIVERSITY OF TECHNOLOGY (SHANDONG ACADEMY OF SCIENCES)
Filing Date: 2026-02-14
Publication Date: 2026-05-26

AI Technical Summary

Technical Problem

Existing technologies in oil well fault diagnosis, especially in complex conditions with multiple fault coupling and similar characteristics, lack sufficient feature utilization and diagnostic accuracy. Traditional methods are time-consuming, labor-intensive, and have poor real-time performance, while machine learning and deep learning-based methods struggle to achieve accurate diagnosis of subtle local differences.

Method used

A physical knowledge-enhanced feature representation method is adopted. By acquiring and preprocessing dynamometer data, a neural network model with a hybrid attention mechanism is designed. This model combines multi-domain, multi-scale physical knowledge features and grayscale images of the dynamometer diagram to achieve complementary information fusion. Finally, a Sigmoid classifier is used for fault diagnosis.

Benefits of technology

It achieves more comprehensive and discriminative feature characterization under complex operating conditions with multiple fault coupling and similar characteristics, thereby improving the diagnostic accuracy and real-time monitoring capability of oil well conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

Figure CN121706036B_ABST

Patent Text Reader

Abstract

This invention relates to the field of oil well fault diagnosis technology, and in particular provides a physical knowledge-enhanced feature representation fault diagnosis method. The method includes acquiring raw data and preprocessing it to obtain preprocessed data; designing physical knowledge-enhanced feature representations to obtain fused physical-image features; constructing a neural network model based on a hybrid attention mechanism to form a nonlinear mapping between physical-image features and fault label data; classifying the data using a Sigmoid classifier to obtain oil well operating condition fault results; evaluating the performance of the neural network model based on the hybrid attention mechanism to obtain evaluation results; and performing online real-time monitoring of oil well operating conditions. Under complex operating conditions with multiple coupled faults and similar features, this method achieves more comprehensive and discriminative feature representation.

Need to check novelty before this filing date? Find Prior Art