A multi-modal data fusion processing method for industrial internet

By adopting an adaptive search for the optimal feature extraction architecture and a cross-spatial alignment mechanism, the problems of modality adaptability and incompleteness in multimodal data processing are solved, achieving high-quality data fusion and secure utilization, which is suitable for industrial internet equipment monitoring.

CN122432967APending Publication Date: 2026-07-21HUBEI DATA GROUP CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
HUBEI DATA GROUP CO LTD
Filing Date
2026-04-13
Publication Date
2026-07-21

AI Technical Summary

Technical Problem

Existing technologies struggle to adaptively extract heterogeneous industrial features, overcome the dual incompleteness of multimodal data, ensure semantic structure alignment, and achieve data quality assessment and security during fusion processing. They also pose risks such as modality loss, label loss, information redundancy and conflict, and data privacy leakage.

Method used

By adaptively searching for the optimal feature extraction architecture for each modality, combining a cross-spatial structure alignment mechanism to handle data incompleteness, and employing a multi-dimensional quality assessment system and automatic sensitive data discovery, high-quality fusion and secure utilization of multimodal data are achieved.

Benefits of technology

It improves the accuracy and security of multimodal data processing, is suitable for monitoring industrial Internet equipment, and solves the problems of poor modal adaptability, incompleteness, fusion conflicts and data privacy leakage, thus realizing the efficient and compliant use of data.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122432967A_ABST
    Figure CN122432967A_ABST
Patent Text Reader

Abstract

The application provides a kind of multi-modal data fusion processing method for industrial internet equipment monitoring, to solve the technical pain points such as heterogeneous sensing data feature adaptation difference, mode loss and label incompleteness, fusion feature conflict and production sensitive data identification difficulty under complex environment of industrial field.The method collects heterogeneous multi-modal data including vibration, acoustic emission, industrial vision and equipment log, searches the optimal auto-encoder architecture of each mode;Relying on the mode-label hybrid driving core mechanism to realize the accurate interpolation of missing working condition data;Through cross-space structure alignment to ensure the semantic consistency of original space and reconstructed complete space, improve the robustness of fusion representation;Further, a multi-dimensional quality evaluation system is constructed to identify private and confidential data in the production process of equipment.The application significantly improves the representation accuracy and compliance safety of industrial equipment state monitoring, and is suitable for typical industrial internet scenarios such as intelligent factory fault diagnosis and production safety monitoring.
Need to check novelty before this filing date? Find Prior Art