Review on crack monitoring technology based on physics-informed neural networks
GUO Xiang
SONG Zhigong
Abstract：Structural health monitoring(SHM)plays a crucial role in ensuring infrastructure safety,extending service life,and reducing maintenance costs.However,existing monitoring methods usually rely on sensor data and traditional physical models,which exhibit certain limitations when addressing problems such as complex mechanical behaviors,environmental change,and data scarcity.As an emerging deep learning approach,physics-informed neural network(PINN)integrates physical laws with data-driven characteristics,offering a novel pathway to overcome the constraints of conventional methods.Firstly,the methods that incorporate the physical principles into machine learning frameworks and the effectiveness of these methods in structural health monitoring were thoroughly discussed.Secondly,multiple approaches for integrating physical knowledge with machine learning models were discussed,highlighting their respective advantages and limitations.Finally,the application of PINN in SHM was comprehensively reviewed,particularly their potential in damage identification,crack propagation analysis,and life prediction.Compared to traditional methods,PINN demonstrates significant advantages in handling complex structural problems.In the training process of embedding physical equations(e.g.mechanical governing equations)into the neural network,PINN not only effectively handles the problems of data scarcity and overfitting,but also improves the accuracy and reliability of structural health assessments.
Keywords：Physics-informed neural networkStructural health monitoringCrackFatigue assessment methodReview
Publication Date：2025-12-15
Online Publishing Date：2026-09-12(First online date of this platform, not the publication date of the document)
Pages：13( 18-30 )