Hybrid data-driven approaches to predicting the compressive strength of ultra-high-performance concrete using SHAP and PDP analyses

Case Studies in Construction Materials, Год журнала: 2024, Номер 20, С. e02991 - e02991

Опубликована: Фев. 19, 2024

Ultra-high-performance concrete (UHPC) is a cutting-edge and advanced constructions material known for its exceptional mechanical properties durability. Recently, machine learning (ML) methods play pivotal role in predicting the compressive strength (CS) of UHPC evaluating dominant input parameters suitable mix design. This research, three hybrid models were utilized: Random Forest (RF), AdaBoost (AB), Gradient Boosting (GB) algorithms with particle swarm optimization (PSO), namely AB-PSO, RF-PSO, GB-PSO, to predict perform SHAP (Shapley additive explanation) analysis. To build predictive ML models, dataset 810 experimental data points was collected from published literature. Additionally, interaction plots generated visualize impact each feature on specific prediction made by model. Our results indicate that better than traditional GB-PSO model showed high accuracy among models. The had higher precision compared other two It achieved R2 values 0.9913 during training stage 0.9804 testing CS. analysis revealed age, fiber, cement, silica fume, superplasticizer significant influence strength, while comparatively lower. PDP (Partial Dependence Plots) amount individually variables can be calculated simply designed These findings are valuable construction applications offer essential insights design engineers builders, aiding their understanding significance component UHPC.

Язык: Английский

---

Compressive strength evaluation of cement-based materials in sulphate environment using optimized deep learning technology

Developments in the Built Environment, Год журнала: 2023, Номер 16, С. 100298 - 100298

Опубликована: Дек. 1, 2023

Strength serves as a vital performance metric for assessing long-term durability of cement-based materials. Nevertheless, there is scarcity models available predicting residual strength in-situ structures made materials exposed to sulphate conditions. To address this challenge, study presents novel approach using deep learning predict the degradation compressive under marine environments. Specifically, convolutional neural network (DCNN) established, consisting two layers, one pooling layer, and fully connected layers. In innovative model, contents cement, water-to-cement ratio, sand, concentration exposure temperature are selected inputs, while output subjected deterioration. improve forecast capability, particle swarm optimization adopted optimizing hyperparameters DCNN, which can be implemented by reducing discrepancy between model prediction measured strength. Finally, experimental data used establish evaluate proposed method. The results show that learning-based predictive has best suffering from attack via comparison with other commonly models. outcome research offers potential solution remaining undergo practical attack.

Язык: Английский

---

Compressive strength prediction of high-strength concrete using hybrid machine learning approaches by incorporating SHAP analysis

Asian Journal of Civil Engineering, Год журнала: 2023, Номер 24(8), С. 3243 - 3263

Опубликована: Май 25, 2023

Язык: Английский

---

A study on the macro-properties of PCB fiber-reinforced concrete from recycled electronic waste and validation of results using RSM and ANN

Asian Journal of Civil Engineering, Год журнала: 2023, Номер 24(6), С. 1667 - 1680

Опубликована: Фев. 5, 2023

Язык: Английский

---

A hybrid strategy of AutoML and SHAP for automated and explainable concrete strength prediction

Case Studies in Construction Materials, Год журнала: 2023, Номер 19, С. e02405 - e02405

Опубликована: Авг. 17, 2023

The precise prediction of concrete compressive strength is essential for ensuring safe and reliable infrastructure design construction. However, traditional empirical models often struggle to accurately predict due the complex nonlinear relationship between properties target strength. This study introduces an AutoML-SHAP (Automatic Machine Learning - SHapley Additive exPlanations) strategy, designed automatically provide insightful interpretations predictive outcomes. AutoML model uses K-fold bagging multilayer stacking automate selection hyperparameter tuning. integration SHAP offers synergistic benefits, facilitating development a precise, efficient, comprehensively interpretable model. Results demonstrate that outperforms other machine learning predicting without human intervention. established within 174 s exhibits comparable performance with R2 = 0.96, RMSE 3.63, MAE 2.41. provides global explanation impact mixing parameters on strength, local feature contribution each prediction, making process transparent reliable. Feature dependence analysis reveals influence tendency

Язык: Английский

---

Predicting concrete strength through packing density using machine learning models

Engineering Applications of Artificial Intelligence, Год журнала: 2023, Номер 126, С. 107177 - 107177

Опубликована: Сен. 25, 2023

Язык: Английский

---

A comparative study of machine learning models for construction costs prediction with natural gradient boosting algorithm and SHAP analysis

Asian Journal of Civil Engineering, Год журнала: 2024, Номер 25(4), С. 3301 - 3316

Опубликована: Фев. 9, 2024

Язык: Английский

---

Optimized machine learning models for prediction of effective stiffness of rectangular reinforced concrete column sections

Structures, Год журнала: 2024, Номер 62, С. 106155 - 106155

Опубликована: Март 13, 2024

Язык: Английский

---

Explainable AutoML models for predicting the strength of high-performance concrete using Optuna, SHAP and ensemble learning

Frontiers in Materials, Год журнала: 2025, Номер 12

Опубликована: Янв. 21, 2025

Accurately predicting key engineering properties, such as compressive and tensile strength, remains a significant challenge in high-performance concrete (HPC) due to its complex heterogeneous composition. Early selection of optimal components the development reliable machine learning (ML) models can significantly reduce time cost associated with extensive experimentation. This study introduces four explainable Automated Machine Learning (AutoML) that integrate Optuna for hyperparameter optimization, SHapley Additive exPlanations (SHAP) interpretability, ensemble algorithms Random Forest (RF), Extreme Gradient Boosting (XGB), Light (LGB), Categorical (CB). The resulting interpretable AutoML O-RF, O-XGB, O-LGB, O-CB are applied predict strengths HPC. Compared baseline model from literature, O-LGB achieved improvements predictive performance. For it reduced Mean Absolute Error (MAE) by 87.69% Root Squared (RMSE) 71.93%. 99.41% improvement MAE 96.67% reduction RMSE, along increases R 2 . Furthermore, SHAP analysis identified critical factors influencing cement content, water, age curing age, water-binder ratio, water-cement ratio strength. approach provides civil engineers robust tool optimizing HPC reducing experimentation costs, supporting enhanced decision-making structural design, risk assessment, other applications.

Язык: Английский

---

An explainable boosting-based ensemble machine learning model for predicting the properties of date palm fiber reinforced concrete

Sustainable Chemistry and Pharmacy, Год журнала: 2025, Номер 44, С. 101949 - 101949

Опубликована: Фев. 15, 2025

Язык: Английский

---