Fault prediction method and device and electronic equipment thereof

By employing a multi-feature set and sub-model strategy, feature selection and model structure are designed for different fault types, solving the problem of insufficient prediction accuracy and sensitivity in optical module fault prediction, and achieving efficient and accurate fault prediction and improved stability.

CN122137754APending Publication Date: 2026-06-02BEIJING BAIDU NETCOM SCI & TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
BEIJING BAIDU NETCOM SCI & TECH CO LTD
Filing Date
2026-02-05
Publication Date
2026-06-02

AI Technical Summary

Technical Problem

In optical module fault prediction, existing technologies struggle to simultaneously improve prediction accuracy and sensitivity with a single model, leading to performance compromises when handling multiple faults and impacting the stability and long-term performance of the optical module.

Method used

A sub-model strategy with multiple feature sets and target fault prediction models is adopted. Through feature extraction and feature optimization, exclusive feature selection and model structure are designed for different fault types. Combined with feature importance evaluation algorithm, efficient and accurate fault prediction of optical modules can be achieved.

Benefits of technology

It improves the stability and long-term performance of optical modules, provides feasible early fault prediction capabilities, reduces fault recovery time and post-fault handling costs, and enhances network stability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122137754A_ABST
    Figure CN122137754A_ABST
Patent Text Reader

Abstract

This disclosure provides a fault prediction method, apparatus, and electronic device thereof, relating to the field of Internet of Things (IoT) technology, particularly to cloud computing and data processing technologies. The specific implementation scheme is as follows: acquiring candidate telemetry data of an optical module; extracting features from the candidate telemetry data using different combinations of feature terms to obtain multiple feature sets; inputting the multiple feature sets into a target fault prediction model, and using multiple fault prediction sub-models of the target fault prediction model to perform inference prediction on the multiple feature sets to obtain the fault prediction result of the optical module. This disclosure avoids performance compromises when a single model handles multiple faults, achieving more efficient and accurate fault prediction, improving the stability and long-term performance of optical modules, providing implementable early fault prediction capabilities for large-scale data center optical modules, and enabling continuous improvement of network stability without additional hardware modifications.
Need to check novelty before this filing date? Find Prior Art