Data anomaly detection method and system for embedded digital twin framework

By combining a digital twin framework with embedded deployment, dual-core detection, and lightweight incremental learning, this approach addresses the issues of decreased model accuracy, insufficient adaptability, and catastrophic forgetting in embedded data anomaly detection, achieving efficient, stable anomaly detection and rapid response.

CN122451735APending Publication Date: 2026-07-24温州中壹技术研究院有限公司
0 Cites 0 Cited by

Patent Information

Application Number
CN202610609444.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-05-06
Publication Date
2026-07-24

AI Technical Summary

Technical Problem

Existing embedded data anomaly detection methods suffer from several problems, including inconsistent initial modeling, high complexity of embedded deployment, significant differences in device operating environments leading to decreased model accuracy, insufficient adaptability of single detection checks to complex anomaly patterns, sensitivity to distribution drift and sudden events, susceptibility to missed detections and false alarms, inability of lightweight incremental learning to continuously adapt to changes in operating conditions, insufficient sample utilization, and catastrophic forgetting.

Method used

A comprehensive data anomaly detection path is adopted, which combines a digital twin framework with embedded deployment, dual-core detection, and lightweight incremental learning. Through hardware resource configuration, virtual model loading, and multi-source interface binding, combined with feature comparison of fast detection kernel and intelligent detection kernel and fusion of multi-dimensional anomaly factors, and combined with online parameter updates and small-batch incremental training, adaptive dual-core data anomaly detection is achieved.

Benefits of technology

It achieves high-precision, low-latency anomaly detection in resource-constrained environments, reduces error rate, improves equipment reliability and operational intelligence, and can quickly respond to complex anomaly modes and adapt to changes in operating conditions, maintaining detection accuracy and stability.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention discloses a data anomaly detection method and system for an embedded digital twin framework, and relates to the technical field of embedded artificial intelligence application, and the method comprises the steps: initializing a digital framework; deploying an embedded system; self-adaptive dual-core detection is carried out, a rapid detection core based on a reasoning enhancement rule engine and an intelligent detection core based on an improved physical constraint lightweight space-time residual model are combined, multi-dimensional abnormal factor fusion is achieved, and a self-adaptive dual-core data anomaly detection model is obtained; lightweight incremental learning is carried out, online model updating is realized through real-time data acquisition, difference weight distribution and small-batch training, and an incremental optimized detection model is obtained; and data anomaly detection. According to the invention, real-time, accurate and evolvable anomaly detection of complex working conditions can be realized in an embedded resource limited environment, and the safety and long-term reliability of equipment operation are improved.
Need to check novelty before this filing date? Find Prior Art