A boiler weld defect identification method, electronic equipment and storage medium

By combining multi-source information from defect detection models and large language models, and utilizing consistency scoring and strategy correction, the problems of missed detection and false alarm rate in boiler weld defect identification were solved, achieving a balance between high recall and low false alarm rate, and improving the accuracy and reliability of identification results.

CN122134720BActive Publication Date: 2026-07-24CHONGQING SPECIAL EQUIP TESTING & RES INST (CHONGQING SPECIAL EQUIP ACCIDENT EMERGENCY INVESTIGATION & PROCESSING CENT)
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
CHONGQING SPECIAL EQUIP TESTING & RES INST (CHONGQING SPECIAL EQUIP ACCIDENT EMERGENCY INVESTIGATION & PROCESSING CENT)
Filing Date
2026-04-30
Publication Date
2026-07-24

AI Technical Summary

Technical Problem

Existing boiler weld defect identification technologies suffer from high risk of missed detection, high false alarm rate, inconsistent defect judgment, and difficulty in meeting industrial safety requirements. In particular, when using deep learning models, it is difficult to achieve a balance between high recall rate and low false alarm rate.

Method used

By acquiring X-ray digital images and metadata, a defect detection model is used to obtain the probability and features of defect types. Combined with a large language model, a search intent is generated. Input is then fed into industry standards and a historical case database for consistency scoring and strategy correction, ensuring the accuracy and reliability of defect identification.

Benefits of technology

It effectively reduces the false alarm rate without lowering the recall rate, improves the accuracy and reliability of defect identification, and meets industrial safety requirements.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122134720B_ABST
    Figure CN122134720B_ABST
Patent Text Reader

Abstract

The application provides a boiler weld defect identification method, an electronic device and a storage medium. The method comprises the following steps: acquiring a radiographic digital image and corresponding metadata of a to-be-evaluated sheet; obtaining a first parameter set based on a preset defect detection model; obtaining a second parameter set of each suspected defect and a retrieval intention based on a pre-trained large language model according to the first parameter set and the metadata; inputting the retrieval intention into an industry standard database and a historical case database respectively to obtain standard evidence and historical case evidence; determining an initial defect based on the first parameter set, the second parameter set, the standard evidence and the historical case evidence; obtaining a consistency score based on the first parameter set, the second parameter set, the standard evidence and the historical case evidence; and correcting the initial defect through the consistency score to obtain an identification result of a boiler weld defect. The application improves the accuracy and reliability of the defect identification result.
Need to check novelty before this filing date? Find Prior Art