A physics-informed wave tomography framework for defect reconstruction: A collaborative network scheme

Liu, Hairui, Li, Qi, Qian, Zhi, Li, Peng, Qian, Zhenghua and Liu, Dianzi (2025) A physics-informed wave tomography framework for defect reconstruction: A collaborative network scheme. Journal of Nondestructive Evaluation, 44 (3). ISSN 0195-9298

PDF (PIWT- a collaborative network) - Accepted Version Restricted to Repository staff only until 10 June 2026. Request a copy

Abstract

It is challenging for guided wave tomography methods to intelligently solve problems in the area of structural defect detection, as this requires more data to achieve the high-accuracy reconstruction of defects. To meet this end, a physics-informed wave tomography framework (PIWT) with a collaborative network scheme is proposed in this paper to reconstruct defects in metal plates with high levels of accuracy and efficiency. First, taking the spatial coordinate information of the point source and sampling points as the inputs of the deep learning collaborative network, a physical principle-based prediction framework is established by minimizing the loss functions to realize the mapping of inputs to outputs, which are represented as the travel time and wave velocity in two collaborative networks for defect reconstruction. To effectively guide the convergence direction of the collaborative network for efficient computations, the Helmholtz equation and source condition are leveraged as the constraints on PIWT to realize the defect reconstruction. As the developed approach belongs to the class of mesh-free methods, its superiority over the conventional mesh-based ultrasonic Lamb wave tomography imaging (ULWTI) technique is demonstrated for defect reconstruction throughout the numerical and experimental examples in terms of accuracy. Moreover, the effects of pre-training on the accelerated convergence and accuracy of the PIWT framework are discussed to allow the training with few epochs and also help effectively achieve real-time high-precision defect reconstruction in the fields of non-destructive testing and structural health monitoring, thus offering a promising solution for broader engineering applications.

Item Type:	Article
Additional Information:	Data Availability: No datasets were generated or analysed during the current study.
Uncontrolled Keywords:	defect reconstruction,guided wave,machine learning,pinn,wave tomography,mechanics of materials,mechanical engineering ,/dk/atira/pure/subjectarea/asjc/2200/2211
Faculty \ School:	Faculty of Science > School of Engineering, Mathematics and Physics
UEA Research Groups:	Faculty of Science > Research Groups > Sustainable Energy Faculty of Science > Research Groups > Materials, Manufacturing & Process Modelling
Related URLs:	http://www.scopus.com/inward/record.url?...
Depositing User:	LivePure Connector
Date Deposited:	30 May 2025 14:30
Last Modified:	23 Jun 2025 14:30
URI:	https://ueaeprints.uea.ac.uk/id/eprint/99368
DOI:	10.1007/s10921-025-01210-z

Actions (login required)

View Item