Xu Zhang
Xu Zhang has a total of 11 co-authors. Here are the top co-authors by number of shared publications:
- Gang Liu
Shandong Provincial Key Laboratory of Oil, Gas and New Energy Storage and Transportation Safety, China University of Petroleum (East China), Qingdao, Shandong, China
4 shared publications - Lei Chen
College of Pipeline and Civil Engineering, China University of Petroleum (East China), Qingdao 266580, China
3 shared publications Zehui Zhang
3 shared publicationsLei Chen
2 shared publicationsYuxuan Dong
2 shared publicationsAocheng Wang
1 shared publication
Latest Publications
- Open Access
- 11 July 2026
- Published by Elsevier in SSRN Electronic Journal
A Frequency-Domain Physics-Informed Fourier Neural Operator for Guided Wave Modeling in Multi-Plate Welded Structures
Constructing high-fidelity wavefield databases for intelligent ultrasonic guided wave testing in complex industrial structures (e.g., multi-plate assemblies with intersecting welds) is bottlenecked by FEM simulation cost. Existing neural operator surrogates predominantly adopt time-domain autoregressive formulations, suffering from error accumulation and high memory overhead. This paper proposes a frequency-domain Physics-Informed Fourier Neural Operator (PI-FNO) that directly maps spatially distributed structural parameters (local thickness, Lamé constants, and density) together with the excitation frequency, to the steady-state complex displacement field, entirely bypassing temporal iteration. Physical consistency is enforced via 3D elastodynamic constraints and a physics-aware spatial reweighting (PASR) module targeting weld regions. The framework is validated on a typical "3-plate 2-weld" petrochemical tank floor unit. Ablation over six model configurations demonstrates that the full PI-FNO with 3D Navier constraints and PASR achieves the lowest N-MSE of 0.0795 globally and 0.0579 at weld regions, with SSIM of 0.940, substantially outperforming both the pure data-driven FNO baseline and an equivalent U-Net. Under data-limited conditions, the Navier PI-FNO achieves comparable accuracy to the baseline with only half the training samples. Physical validation combining f-k dispersion analysis and time-domain LDV comparison confirms that the PI-FNO accurately captures mode conversion and time-of-flight kinematics, demonstrating reliable sim-to-real transferability.
- 11 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessA Frequency-Domain Physics-Informed Fourier Neural Operator for Guided Wave Modeling in Multi-Plate Welded Structures
Constructing high-fidelity wavefield databases for intelligent ultrasonic guided wave testing in complex industrial structures (e.g., multi-plate assemblies with intersecting welds) is bottlenecked by FEM simulation cost. Existing neural operator surrogates predominantly adopt time-domain autoregressive formulations, suffering from error accumulation and high memory overhead. This paper proposes a frequency-domain Physics-Informed Fourier Neural Operator (PI-FNO) that directly maps spatially distributed structural parameters (local thickness, Lamé constants, and density) together with the excitation frequency, to the steady-state complex displacement field, entirely bypassing temporal iteration. Physical consistency is enforced via 3D elastodynamic constraints and a physics-aware spatial reweighting (PASR) module targeting weld regions. The framework is validated on a typical "3-plate 2-weld" petrochemical tank floor unit. Ablation over six model configurations demonstrates that the full PI-FNO with 3D Navier constraints and PASR achieves the lowest N-MSE of 0.0795 globally and 0.0579 at weld regions, with SSIM of 0.940, substantially outperforming both the pure data-driven FNO baseline and an equivalent U-Net. Under data-limited conditions, the Navier PI-FNO achieves comparable accuracy to the baseline with only half the training samples. Physical validation combining f-k dispersion analysis and time-domain LDV comparison confirms that the PI-FNO accurately captures mode conversion and time-of-flight kinematics, demonstrating reliable sim-to-real transferability.
- 13 October 2024
- Published by Elsevier in Thin-Walled Structures
- Vol. 205
An improved modal tracking algorithm for dispersion analysis of arbitrary prestressed plates
- 13 October 2024
- Published by Elsevier in Thin-Walled Structures
- Vol. 205
An improved modal tracking algorithm for dispersion analysis of arbitrary prestressed plates
- 15 June 2024
- Published by Elsevier in Thin-Walled Structures
- Vol. 202
A novel multi-step superposition model for the dispersion analysis of multiaxial prestressed plate-like structures
- 15 June 2024
- Published by Elsevier in Thin-Walled Structures
- Vol. 202
A novel multi-step superposition model for the dispersion analysis of multiaxial prestressed plate-like structures
- 29 February 2024
- Published by Elsevier in International Journal of Mechanical Sciences
- Vol. 271
Thermo-acoustoelastic determination of third-order elastic constants using coda wave interferometry
- 29 February 2024
- Published by Elsevier in International Journal of Mechanical Sciences
- Vol. 271
Thermo-acoustoelastic determination of third-order elastic constants using coda wave interferometry