Shixue Liang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Shixue Liang has a total of 40 co-authors. Here are the top co-authors by number of shared publications:
Xing Lin
Zhejiang Sci-Tech Univ
6 shared publications- Yuanxie Shen
School of Civil Engineering and Architecture, Zhejiang Sci-Tech University, Hangzhou 310018, China
6 shared publications - Zhengyu Fei
School of Civil Engineering and Architecture, Zhejiang Sci-Tech University, Hangzhou 310018, China
4 shared publications Junning Wu
School of Civil Engineering and Architecture, Zhejiang Sci-Tech University, Hangshou 310018, China
3 shared publications- Yan-Ping Liang
Department of Civil Engineering, Hangzhou City University and Zhejiang Engineering Research Center of Intelligent Urban Infrastructure, Hangzhou 310015, P. R. China
3 shared publications - Yiqing Cai
School of Civil Engineering and Architecture, Zhejiang Sci-Tech University, Hangzhou 310018, China
3 shared publications
Latest Publications
- 1 January 2027
- Published by American Society of Civil Engineers (ASCE) in Journal of Computing in Civil Engineering
- Vol. 41 (1)
Surrogate Modeling of Stochastic Hysteresis Response in Irregular RC Shear Walls Using CNN-KAN and Manifold Random Fields
Accurately predicting the hysteretic response of irregular reinforced concrete (RC) shear walls is essential for seismic performance assessment. However, spatial variability in material properties and geometric complexity introduce significant stochasticity, making conventional stochastic finite element (SFEM) methods computationally inefficient due to difficulties in random field assignment and the high cost of nonlinear analyses. This study proposes an efficient deep learning framework that uses a convolutional neural network-Kolmogorov–Arnold network (CNN-KAN) model to predict stochastic hysteresis responses of irregular shear walls with manifold random fields. The random field is constructed using the isometric mapping (isomap) method to accurately represent the spatial variability of material properties in SFEM. Then, hysteresis curves obtained from SFEM analysis are used to train the CNN-KAN model, the CNN-long short-term memory (LSTM) model, and the CNN-multilayer perceptron (MLP) model. While all three models achieve high prediction accuracy, only CNN-KAN successfully preserves the stochastic diversity of structural responses. The traditional CNN-MLP and CNN-LSTM models suffer from an averaging effect, failing to capture the inherent randomness. The trained CNN-KAN model is then used as a surrogate model for Monte Carlo simulation in large-scale sample analysis. The results demonstrated that it can efficiently generate hysteresis curves and energy dissipation distributions to study the influence of spatial correlation on structure stochastic responses. The proposed model significantly enhances computational efficiency, reducing the computing time to only 1/300 of the conventional SFEM analysis. Further, the proposed framework enables efficient uncertainty quantification and is extensible to other complex structural systems involving spatial randomness.
- 1 January 2027
- Published by American Society of Civil Engineers (ASCE) in Journal of Computing in Civil Engineering
- Vol. 41 (1)
Surrogate Modeling of Stochastic Hysteresis Response in Irregular RC Shear Walls Using CNN-KAN and Manifold Random Fields
Accurately predicting the hysteretic response of irregular reinforced concrete (RC) shear walls is essential for seismic performance assessment. However, spatial variability in material properties and geometric complexity introduce significant stochasticity, making conventional stochastic finite element (SFEM) methods computationally inefficient due to difficulties in random field assignment and the high cost of nonlinear analyses. This study proposes an efficient deep learning framework that uses a convolutional neural network-Kolmogorov–Arnold network (CNN-KAN) model to predict stochastic hysteresis responses of irregular shear walls with manifold random fields. The random field is constructed using the isometric mapping (isomap) method to accurately represent the spatial variability of material properties in SFEM. Then, hysteresis curves obtained from SFEM analysis are used to train the CNN-KAN model, the CNN-long short-term memory (LSTM) model, and the CNN-multilayer perceptron (MLP) model. While all three models achieve high prediction accuracy, only CNN-KAN successfully preserves the stochastic diversity of structural responses. The traditional CNN-MLP and CNN-LSTM models suffer from an averaging effect, failing to capture the inherent randomness. The trained CNN-KAN model is then used as a surrogate model for Monte Carlo simulation in large-scale sample analysis. The results demonstrated that it can efficiently generate hysteresis curves and energy dissipation distributions to study the influence of spatial correlation on structure stochastic responses. The proposed model significantly enhances computational efficiency, reducing the computing time to only 1/300 of the conventional SFEM analysis. Further, the proposed framework enables efficient uncertainty quantification and is extensible to other complex structural systems involving spatial randomness.
- 30 October 2026
- Published by World Scientific Pub Co Pte Ltd in International Journal of Structural Stability and Dynamics
Influence of Random Geometric Imperfection Morphology on Localized Buckling of Axially Compressed Cylindrical Shells
Initial geometric imperfections strongly affect the buckling resistance of cylindrical shells, yet the role of their spatial morphology in localized buckling remains insufficiently understood. This study investigates axially compressed cylindrical shells with random imperfections normalized to a common root-mean-square (RMS) amplitude of [Formula: see text]. A total of 1[Formula: see text]600 geometrically nonlinear finite element analyses are performed for eight correlation lengths to examine the critical load factor, response localization, and imperfection–hotspot correspondence. The critical load factor varies non-monotonically with correlation length: the lowest values occur in the short-to-intermediate range, whereas the largest local roughness occurs at the shortest correlation length. Pointwise and regional analyses show that severe imperfection regions correspond most strongly with localized high-response zones over the short-to-intermediate range. Very short correlation fields exhibit rough short-wave patterns, while long correlation fields are smoother and associated with more distributed responses and weaker spatial correspondence. Within the prescribed imperfection-amplitude level, these results show that changes in spatial morphology can produce markedly different buckling responses even when the global RMS amplitude is fixed.
- 30 October 2026
- Published by World Scientific Pub Co Pte Ltd in International Journal of Structural Stability and Dynamics
Influence of Random Geometric Imperfection Morphology on Localized Buckling of Axially Compressed Cylindrical Shells
Initial geometric imperfections strongly affect the buckling resistance of cylindrical shells, yet the role of their spatial morphology in localized buckling remains insufficiently understood. This study investigates axially compressed cylindrical shells with random imperfections normalized to a common root-mean-square (RMS) amplitude of [Formula: see text]. A total of 1[Formula: see text]600 geometrically nonlinear finite element analyses are performed for eight correlation lengths to examine the critical load factor, response localization, and imperfection–hotspot correspondence. The critical load factor varies non-monotonically with correlation length: the lowest values occur in the short-to-intermediate range, whereas the largest local roughness occurs at the shortest correlation length. Pointwise and regional analyses show that severe imperfection regions correspond most strongly with localized high-response zones over the short-to-intermediate range. Very short correlation fields exhibit rough short-wave patterns, while long correlation fields are smoother and associated with more distributed responses and weaker spatial correspondence. Within the prescribed imperfection-amplitude level, these results show that changes in spatial morphology can produce markedly different buckling responses even when the global RMS amplitude is fixed.
- 10 September 2026
- Published by Elsevier in Probabilistic Engineering Mechanics
- Vol. 86
Regression mode collapse in deep learning surrogates for stochastic structural response prediction
- 10 September 2026
- Published by Elsevier in Probabilistic Engineering Mechanics
- Vol. 86
Regression mode collapse in deep learning surrogates for stochastic structural response prediction
- Open Access
Sustainable Mix Design of Sugarcane Bagasse Ash Concrete via AutoML-Assisted Multi-Objective Optimization
The environmental impact of cement production has become a growing global concern due to its substantial contribution to CO2 emissions. Sugarcane bagasse ash (SCBA), as a supplementary cementitious material, offers a sustainable alternative by partially replacing cement and reducing the carbon footprint of concrete. However, determining optimal mix proportions that balance mechanical strength, cost efficiency, and environmental benefits remains a complex challenge. In this study, a multi-objective optimization framework was developed by integrating automated machine learning (Auto-ML) with the NSGA-III algorithm. A surrogate model for predicting compressive strength was constructed using the TPOT-based Auto-ML tool, achieving high predictive accuracy with R2 values of 0.993 and 0.908 for the training and test datasets, respectively. NSGA-III was then employed to derive Pareto-optimal mix designs, enabling simultaneous optimization of strength, cost, and CO2 emissions. To validate the proposed multi-objective optimization framework, SCBA concrete specimens were prepared using the optimized mix proportions and tested under uniaxial compression. The experimental results exhibited good agreement with the predicted values, with deviations within an acceptable margin, thereby confirming the accuracy and reliability of the framework. This study provides a practical approach for the intelligent design of low-carbon SCBA concrete, contributing to the advancement of sustainable construction practices.
Open AccessSustainable Mix Design of Sugarcane Bagasse Ash Concrete via AutoML-Assisted Multi-Objective Optimization
The environmental impact of cement production has become a growing global concern due to its substantial contribution to CO2 emissions. Sugarcane bagasse ash (SCBA), as a supplementary cementitious material, offers a sustainable alternative by partially replacing cement and reducing the carbon footprint of concrete. However, determining optimal mix proportions that balance mechanical strength, cost efficiency, and environmental benefits remains a complex challenge. In this study, a multi-objective optimization framework was developed by integrating automated machine learning (Auto-ML) with the NSGA-III algorithm. A surrogate model for predicting compressive strength was constructed using the TPOT-based Auto-ML tool, achieving high predictive accuracy with R2 values of 0.993 and 0.908 for the training and test datasets, respectively. NSGA-III was then employed to derive Pareto-optimal mix designs, enabling simultaneous optimization of strength, cost, and CO2 emissions. To validate the proposed multi-objective optimization framework, SCBA concrete specimens were prepared using the optimized mix proportions and tested under uniaxial compression. The experimental results exhibited good agreement with the predicted values, with deviations within an acceptable margin, thereby confirming the accuracy and reliability of the framework. This study provides a practical approach for the intelligent design of low-carbon SCBA concrete, contributing to the advancement of sustainable construction practices.
- 9 January 2026
- Published by SAGE Publications in Advances in Structural Engineering
- Vol. 29 (12) , 2438-2454
Punching shear behavior of Slab–Column joints with partial UHPC enhancement: Experimental, numerical, and data-driven predictions
Reinforced concrete (RC) slab–column joints are prone to brittle punching shear failure. Enhancing their punching shear capacity is therefore critical for improving structural safety of flat slab structures. Ultra-high performance concrete (UHPC), characterized by high strength and ductility, provides a promising approach to enhancing the punching shear performance of slab–column joints. This study evaluates the effectiveness of partial UHPC application in enhancing punching shear performance. Four specimens with varying UHPC areas were tested to assess load–displacement behavior and failure modes. To extend the investigation, 216 validated finite element (FE) models were developed to examine the effects of UHPC area, material strengths, reinforcement ratio, and slab depth. A comprehensive dataset combining simulation and test results was used to train machine learning (ML) models. Among four algorithms, XGBoost achieved the best performance (RMSE = 21.87, R 2 = 0.98). SHAP and partial dependence plot analyses revealed UHPC area as the most influential factor, followed by UHPC strength and slab depth. In contrast, reinforcement ratio and column size showed limited influence, diverging from trends observed in conventional RC slab-column joints. These findings highlight the structural benefits of partial UHPC enhancement and offer new insight into optimizing slab–column joint design.
- 9 January 2026
- Published by SAGE Publications in Advances in Structural Engineering
- Vol. 29 (12) , 2438-2454
Punching shear behavior of Slab–Column joints with partial UHPC enhancement: Experimental, numerical, and data-driven predictions
Reinforced concrete (RC) slab–column joints are prone to brittle punching shear failure. Enhancing their punching shear capacity is therefore critical for improving structural safety of flat slab structures. Ultra-high performance concrete (UHPC), characterized by high strength and ductility, provides a promising approach to enhancing the punching shear performance of slab–column joints. This study evaluates the effectiveness of partial UHPC application in enhancing punching shear performance. Four specimens with varying UHPC areas were tested to assess load–displacement behavior and failure modes. To extend the investigation, 216 validated finite element (FE) models were developed to examine the effects of UHPC area, material strengths, reinforcement ratio, and slab depth. A comprehensive dataset combining simulation and test results was used to train machine learning (ML) models. Among four algorithms, XGBoost achieved the best performance (RMSE = 21.87, R 2 = 0.98). SHAP and partial dependence plot analyses revealed UHPC area as the most influential factor, followed by UHPC strength and slab depth. In contrast, reinforcement ratio and column size showed limited influence, diverging from trends observed in conventional RC slab-column joints. These findings highlight the structural benefits of partial UHPC enhancement and offer new insight into optimizing slab–column joint design.
- Open Access
- 1 January 2026
- Published by Tech Science Press in Computer Modeling in Engineering & Sciences
- Vol. 148 (3) , 1-10
A Two-Stage Machine Learning and Finite Element Framework for System Failure Analysis of Reinforced Concrete Flat-Slab Structures
Punching shear failure at a single slab–column joint can trigger rapid load redistribution and cascading damage in reinforced concrete flat-slab systems, making component-level reliability assessment insufficient for evaluating progressive-collapse risk. Moreover, direct Monte Carlo simulation (MCS) coupled with nonlinear finite element (FE) analysis is computationally prohibitive for rare-event system reliability problems involving multiple uncertainties. To address these challenges, this study proposes a two-stage machine learning–finite element (ML–FE) framework for system failure assessment of reinforced concrete flat-slab structures. In Stage I, an ML surrogate trained on 610 experimental slab–column joint tests predicts punching shear resistance and screens potentially critical realizations from a large MCS sample space. In Stage II, only the screened realizations are propagated into a nonlinear OpenSees model to simulate load redistribution, sequential joint failure, and system-level damage evolution. System reliability is then evaluated under three candidate failure criteria representing progressively more severe performance states, from initial local punching to widespread structural damage. The results show that the estimated system failure probability is strongly dependent on the adopted failure definition: local punching events frequently develop into moderate system damage, whereas only a smaller fraction progresses to severe widespread failure. For the largest Monte Carlo case, the Stage-I screening filtered approximately 99.94% of safe realizations and reduced the estimated computational time from about 26,000 h for direct MCS–FE analysis to approximately 15 h. The proposed framework contributes a computationally scalable link between experimentally informed component resistance and mechanics-based system response, enabling multi-level reliability assessment while also revealing characteristic failure-propagation paths and structural robustness of reinforced concrete flat-slab systems.
- 1 January 2026
- Published by Tech Science Press in Computer Modeling in Engineering & Sciences
- Vol. 148 (3) , 1-10
Open AccessA Two-Stage Machine Learning and Finite Element Framework for System Failure Analysis of Reinforced Concrete Flat-Slab Structures
Punching shear failure at a single slab–column joint can trigger rapid load redistribution and cascading damage in reinforced concrete flat-slab systems, making component-level reliability assessment insufficient for evaluating progressive-collapse risk. Moreover, direct Monte Carlo simulation (MCS) coupled with nonlinear finite element (FE) analysis is computationally prohibitive for rare-event system reliability problems involving multiple uncertainties. To address these challenges, this study proposes a two-stage machine learning–finite element (ML–FE) framework for system failure assessment of reinforced concrete flat-slab structures. In Stage I, an ML surrogate trained on 610 experimental slab–column joint tests predicts punching shear resistance and screens potentially critical realizations from a large MCS sample space. In Stage II, only the screened realizations are propagated into a nonlinear OpenSees model to simulate load redistribution, sequential joint failure, and system-level damage evolution. System reliability is then evaluated under three candidate failure criteria representing progressively more severe performance states, from initial local punching to widespread structural damage. The results show that the estimated system failure probability is strongly dependent on the adopted failure definition: local punching events frequently develop into moderate system damage, whereas only a smaller fraction progresses to severe widespread failure. For the largest Monte Carlo case, the Stage-I screening filtered approximately 99.94% of safe realizations and reduced the estimated computational time from about 26,000 h for direct MCS–FE analysis to approximately 15 h. The proposed framework contributes a computationally scalable link between experimentally informed component resistance and mechanics-based system response, enabling multi-level reliability assessment while also revealing characteristic failure-propagation paths and structural robustness of reinforced concrete flat-slab systems.
- 20 November 2025
- Published by Elsevier in Structures
- Vol. 82
Experimental and numerical study of flexural behavior of reinforced concrete slabs containing recycled slurry micropowder as cement replacement
- 20 November 2025
- Published by Elsevier in Structures
- Vol. 82
Experimental and numerical study of flexural behavior of reinforced concrete slabs containing recycled slurry micropowder as cement replacement
- Open Access
- 1 January 2025
- Published by Tech Science Press in Computer Modeling in Engineering & Sciences
- Vol. 145 (1) , 215-239
A CGAN Framework for Predicting Crack Patterns and Stress-Strain Behavior in Concrete Random Media
Random media like concrete and ceramics exhibit stochastic crack propagation due to their heterogeneous microstructures. This study establishes a Conditional Generative Adversarial Network (CGAN) combined with random field modeling for the efficient prediction of stochastic crack patterns and stress-strain responses. A total dataset of 500 samples, including crack propagation images and corresponding stress-strain curves, is generated via random Finite Element Method (FEM) simulations. This dataset is then partitioned into 400 training and 100 testing samples. The model demonstrates robust performance with Intersection over Union (IoU) scores of 0.8438 and 0.8155 on training and testing datasets, and R2 values of 0.9584 and 0.9462 in stress-strain curve predictions. By using these results, the CGAN is subsequently implemented as a surrogate model for large-scale Monte Carlo (MC) Simulations to capture the key statistical characteristics such as crack density and spatial distribution. Compared to conventional FEM-based methods, this approach reduces the computational cost to about 1/250 while maintaining high prediction accuracy. The methodology establishes a viable pathway for probabilistic fracture analysis in quasi-brittle materials, balancing computational efficiency with physical fidelity in capturing material stochasticity.
- 1 January 2025
- Published by Tech Science Press in Computer Modeling in Engineering & Sciences
- Vol. 145 (1) , 215-239
Open AccessA CGAN Framework for Predicting Crack Patterns and Stress-Strain Behavior in Concrete Random Media
Random media like concrete and ceramics exhibit stochastic crack propagation due to their heterogeneous microstructures. This study establishes a Conditional Generative Adversarial Network (CGAN) combined with random field modeling for the efficient prediction of stochastic crack patterns and stress-strain responses. A total dataset of 500 samples, including crack propagation images and corresponding stress-strain curves, is generated via random Finite Element Method (FEM) simulations. This dataset is then partitioned into 400 training and 100 testing samples. The model demonstrates robust performance with Intersection over Union (IoU) scores of 0.8438 and 0.8155 on training and testing datasets, and R2 values of 0.9584 and 0.9462 in stress-strain curve predictions. By using these results, the CGAN is subsequently implemented as a surrogate model for large-scale Monte Carlo (MC) Simulations to capture the key statistical characteristics such as crack density and spatial distribution. Compared to conventional FEM-based methods, this approach reduces the computational cost to about 1/250 while maintaining high prediction accuracy. The methodology establishes a viable pathway for probabilistic fracture analysis in quasi-brittle materials, balancing computational efficiency with physical fidelity in capturing material stochasticity.