Physics-Informed Cascaded Neural Network for Predicting Rear Plate Equivalent Hole Diameter in Honeycomb Sandwich Panels under Hypervelocity Impact

To meet the requirements of spacecraft protection and structural optimization under hypervelocity impact, a physics-informed cascaded neural network (PICNN) model was developed to improve the prediction accuracy of the rear plate equivalent hole diameter in honeycomb sandwich panels. A damage response dataset for honeycomb sandwich panels was constructed from experimental data reported in the published literature. Outlier removal and local interpolation-based data augmentation along the impact velocity direction were then performed to improve data quality and expand the coverage of impact conditions. Dimensionless input features were reconstructed through dimensional analysis, and a physics-informed feature representing the subsequent damage driving capability was constructed by incorporating the front plate hole response, thereby enabling the physics-guided transfer from front plate damage to rear plate hole prediction. The effects of different regression loss functions, including mean squared error loss, mean absolute error (MAE) loss, and Huber loss, on model performance were compared to determine the optimal loss function for rear plate hole diameter prediction. The results show that data augmentation improved model prediction performance and that MAE loss achieved the best overall performance. After the physics-informed feature was introduced, the PICNN reduced the root-mean-square error by approximately 19.3% and increased the coefficient of determination from 0.95 to 0.97, corresponding to a relative improvement of approximately 2.1% compared with the model without the physics layer. Compared with traditional empirical models and the random forest model, the PICNN achieved superior performance in both rear plate perforation classification and rear plate equivalent hole diameter prediction. The proposed model provides an effective approach for rear plate damage prediction in honeycomb sandwich panels and offers a predictive basis for protective performance evaluation and structural optimization design.

Authors

Institutions

Publication Details

Journal
Journal of Aerospace Engineering
Published
2026-11-18
DOI
https://doi.org/10.1061/jaeeez.aseng-7307
Primary Topic
High-Velocity Impact and Material Behavior
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Physics-Informed Cascaded Neural Network for Predicting Rear Plate Equivalent Hole Diameter in Honeycomb Sandwich Panels under Hypervelocity Impact

Hao Li, Hengwei Xu, Yuke Wang, Yonggang Lu et al.
2 links
Journal of Aerospace Engineering
High-Velocity Impact and Material Behavior
article

Physics-Informed Cascaded Neural Network for Predicting Rear Plate Equivalent Hole Diameter in Honeycomb Sandwich Panels under Hypervelocity Impact

Hao Li, Hengwei Xu, Yuke Wang, Yonggang Lu, Bin Jia, Zhengcao Lu
article en

Abstract

To meet the requirements of spacecraft protection and structural optimization under hypervelocity impact, a physics-informed cascaded neural network (PICNN) model was developed to improve the prediction accuracy of the rear plate equivalent hole diameter in honeycomb sandwich panels. A damage response dataset for honeycomb sandwich panels was constructed from experimental data reported in the published literature. Outlier removal and local interpolation-based data augmentation along the impact velocity direction were then performed to improve data quality and expand the coverage of impact conditions. Dimensionless input features were reconstructed through dimensional analysis, and a physics-informed feature representing the subsequent damage driving capability was constructed by incorporating the front plate hole response, thereby enabling the physics-guided transfer from front plate damage to rear plate hole prediction. The effects of different regression loss functions, including mean squared error loss, mean absolute error (MAE) loss, and Huber loss, on model performance were compared to determine the optimal loss function for rear plate hole diameter prediction. The results show that data augmentation improved model prediction performance and that MAE loss achieved the best overall performance. After the physics-informed feature was introduced, the PICNN reduced the root-mean-square error by approximately 19.3% and increased the coefficient of determination from 0.95 to 0.97, corresponding to a relative improvement of approximately 2.1% compared with the model without the physics layer. Compared with traditional empirical models and the random forest model, the PICNN achieved superior performance in both rear plate perforation classification and rear plate equivalent hole diameter prediction. The proposed model provides an effective approach for rear plate damage prediction in honeycomb sandwich panels and offers a predictive basis for protective performance evaluation and structural optimization design.

Journal of Aerospace EngineeringVol. 40(2)
University of Science and Technology of China (CN), Southwest University of Science and Technology (CN), China Academy of Engineering Physics (CN)
Industry, innovation and infrastructure
High-Velocity Impact and Material Behavior
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon