Numerical and Machine Learning Prediction of Blast and Fragment Propagation in RC Frames Under Internal Explosion
- 1College of Civil Engineering, Tongji University, Shanghai 200092, China
Abstract
Each component of reinforced concrete (RC) frames absorbs and dissipates blast energy, mitigating damage to adjacent rooms. Finite element modeling and machine learning are combined to analyze the propagation of blast waves and wall fragments in RC frames. A validated model simulates explosion scenarios with five TNT equivalents and examines blast loading and dynamic responses of adjacent rooms. Orthogonal experimental design is adopted to analyze the attenuation effects of structural parameters (room dimensions, venting coefficient, wall strength and thickness), TNT equivalent, and the number of propagation rooms on power parameters propagated to adjacent rooms. The results reveal that the peak overpressure and impulse of the adjacent rooms decrease by approximately two and one orders of magnitude, respectively, compared with the preceding room. The ejection velocity of wall fragments decays exponentially with the number of propagation rooms, which is the primary factor attenuating the power parameters, followed by the masonry wall strength and thickness. A backpropagation neural network-based surrogate model is developed to predict the attenuation coefficients of the power parameters (peak overpressure, duration, and wall fragment ejection velocity), achieving prediction errors below 20%. The proposed prediction framework provides a computationally efficient alternative to full-scale numerical simulation for rapid damage assessment.
Related
- A numerical method for two-dimensional Boussinesq-Full Dispersion systems modelling internal wave propagation
- Mathematical and Numerical Analysis of Tsunami Generation and Propagation: The Initial Tsunami Condition
- Direct Vertex-Derivative Coefficient Propagation for Homogeneous Numerical Integration (HNI)
- Fragment-Level Graph Encoding for Interpretable Molecular Property Prediction
- Frequentist prediction intervals for random-effects meta-analysis via confidence-distribution propagation
BibTeX
@article{qiu2026numerical,
title = {{Numerical and Machine Learning Prediction of Blast and Fragment Propagation in RC Frames Under Internal Explosion}},
author = {Tao Qiu and Junrun Li},
journal = {Mathematics},
year = {2026},
month = nov,
doi = {10.3390/math14214070},
url = {https://www.mdpi.com/2227-7390/14/21/4070},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Numerical and Machine Learning Prediction of Blast and Fragment Propagation in RC Frames Under Internal Explosion
AU - Tao Qiu
AU - Junrun Li
PY - 2026
DA - 2026/11/06
JO - Mathematics
DO - 10.3390/math14214070
UR - https://www.mdpi.com/2227-7390/14/21/4070
AB - Each component of reinforced concrete (RC) frames absorbs and dissipates blast energy, mitigating damage to adjacent rooms. Finite element modeling and machine learning are combined to analyze the propagation of blast waves and wall fragments in RC frames. A validated model simulates explosion scenarios with five TNT equivalents and examines blast loading and dynamic responses of adjacent rooms. Orthogonal experimental design is adopted to analyze the attenuation effects of structural parameters (room dimensions, venting coefficient, wall strength and thickness), TNT equivalent, and the number of propagation rooms on power parameters propagated to adjacent rooms. The results reveal that the peak overpressure and impulse of the adjacent rooms decrease by approximately two and one orders of magnitude, respectively, compared with the preceding room. The ejection velocity of wall fragments decays exponentially with the number of propagation rooms, which is the primary factor attenuating the power parameters, followed by the masonry wall strength and thickness. A backpropagation neural network-based surrogate model is developed to predict the attenuation coefficients of the power parameters (peak overpressure, duration, and wall fragment ejection velocity), achieving prediction errors below 20%. The proposed prediction framework provides a computationally efficient alternative to full-scale numerical simulation for rapid damage assessment.
SN - 2227-7390
ER -