Method and system for predicting remaining life of equipment based on abrasive particle monitoring and physical constraints

By constructing a multivariate feature map of abrasive particles and a deeply coupled PINN model, combined with an adaptive weighted loss function and a two-stage strategy, the problems of lack of physical constraints and small sample size in abrasive particle monitoring methods are solved, and efficient and reliable prediction of equipment remaining life is achieved.

CN122287402APending Publication Date: 2026-06-26XI AN JIAOTONG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
XI AN JIAOTONG UNIV
Filing Date
2026-05-28
Publication Date
2026-06-26

AI Technical Summary

Technical Problem

Existing abrasive monitoring methods lack physical constraints, which makes the prediction results prone to violating the equipment degradation law. Furthermore, the model has poor generalization ability under small sample conditions, making it difficult to effectively capture wear mode changes.

Method used

By establishing the wear rate-abrasive particle generation mapping equation and the abrasive particle type transition probability evolution equation, a multivariate feature map is constructed. A deeply coupled PINN model and a multi-scale neural operator are designed, and an adaptive weighted joint loss function and a two-stage pre-training-adversarial fine-tuning strategy are combined to achieve multi-physical constraint prediction of abrasive particle type, size distribution and quantity.

Benefits of technology

It improves the physical consistency and stability of predictions, reduces lifetime rebound, and enhances prediction accuracy and generalization ability under small sample conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122287402A_ABST
    Figure CN122287402A_ABST
Patent Text Reader

Abstract

This invention discloses a method and system for predicting the remaining life of equipment based on abrasive particle monitoring and physical constraints. Specifically, it involves: acquiring the type, size distribution, and quantity of abrasive particles in the lubricating oil; establishing a wear rate-abrasive particle generation mapping equation and an abrasive particle type transition probability evolution equation based on Archard wear theory; constructing a deeply coupled PINN model composed of a solution network and a multi-physical residual coding network using multivariate temporal features as input; obtaining a life-feature sensitivity field; constructing a multi-scale neural operator for abrasive particles to efficiently solve for the remaining life of the equipment; designing an adaptive weighted joint loss function based on physical residual gradient balancing; and a two-stage strategy of physical-guided prior pre-training and adversarial fine-tuning to solve the gradient competition and sample scarcity problems among multiple physical constraints of abrasive particles. This invention achieves physically consistent and highly reliable prediction of the remaining life of equipment under conditions of few samples by constructing a deeply coupled architecture of high-dimensional physical priors and neural networks.
Need to check novelty before this filing date? Find Prior Art