BULLETIN OF THE CHINESE CERAMIC SOCIETY ›› 2026, Vol. 45 ›› Issue (9): 3070-3082.DOI: 10.16552/j.cnki.issn1001-1625.2026.0229
• Cement and Concrete • Previous Articles Next Articles
LI Shuo(
), QIN Yanhui(
), LI Meng’en, MA Haoyuan, YIN Jinshuai, ZHENG Lulu, DENG Suwen
Received:2026-03-16
Revised:2026-04-08
Online:2026-09-15
Published:2026-10-09
Contact:
QIN Yanhui
CLC Number:
LI Shuo, QIN Yanhui, LI Meng’en, MA Haoyuan, YIN Jinshuai, ZHENG Lulu, DENG Suwen. Full-Age Prediction and Interpretability Study of Concrete Compressive Strength Based on Machine Learning[J]. BULLETIN OF THE CHINESE CERAMIC SOCIETY, 2026, 45(9): 3070-3082.
| Indicator | Mean | Standard value | Minimum | First quartile | Median | Third quartile | Maximum | |
|---|---|---|---|---|---|---|---|---|
| Input variable | Cement content/(kg·m-3) | 281.91 | 102.96 | 89.00 | 197.50 | 271.00 | 355.68 | 540.00 |
| Slag content/(kg·m-3) | 72.63 | 85.60 | 0 | 0 | 26.00 | 135.70 | 359.40 | |
| Fly ash content/(kg·m-3) | 50.21 | 60.64 | 0 | 0 | 0 | 106.98 | 200.10 | |
| Water content/(kg·m-3) | 178.00 | 25.06 | 75.10 | 160.60 | 181.71 | 192.94 | 247.00 | |
| Plasticizer content/(kg·m-3) | 5.75 | 5.81 | 0 | 0 | 5.40 | 9.61 | 40.00 | |
| Coarse aggregate content/(kg·m-3) | 998.12 | 103.13 | 801.00 | 936.00 | 998.00 | 1 056.00 | 1 477.30 | |
| Fine aggregate content/(kg·m-3) | 764.48 | 88.71 | 389.50 | 720.00 | 772.10 | 821.00 | 992.60 | |
| Curing age/d | 42.16 | 59.17 | 1.00 | 7.00 | 28.00 | 49.50 | 365.00 | |
| Output variable | Compressive strength/MPa | 37.04 | 16.44 | 2.33 | 24.80 | 36.53 | 47.81 | 82.60 |
Table 1 Dataset parameter statistics
| Indicator | Mean | Standard value | Minimum | First quartile | Median | Third quartile | Maximum | |
|---|---|---|---|---|---|---|---|---|
| Input variable | Cement content/(kg·m-3) | 281.91 | 102.96 | 89.00 | 197.50 | 271.00 | 355.68 | 540.00 |
| Slag content/(kg·m-3) | 72.63 | 85.60 | 0 | 0 | 26.00 | 135.70 | 359.40 | |
| Fly ash content/(kg·m-3) | 50.21 | 60.64 | 0 | 0 | 0 | 106.98 | 200.10 | |
| Water content/(kg·m-3) | 178.00 | 25.06 | 75.10 | 160.60 | 181.71 | 192.94 | 247.00 | |
| Plasticizer content/(kg·m-3) | 5.75 | 5.81 | 0 | 0 | 5.40 | 9.61 | 40.00 | |
| Coarse aggregate content/(kg·m-3) | 998.12 | 103.13 | 801.00 | 936.00 | 998.00 | 1 056.00 | 1 477.30 | |
| Fine aggregate content/(kg·m-3) | 764.48 | 88.71 | 389.50 | 720.00 | 772.10 | 821.00 | 992.60 | |
| Curing age/d | 42.16 | 59.17 | 1.00 | 7.00 | 28.00 | 49.50 | 365.00 | |
| Output variable | Compressive strength/MPa | 37.04 | 16.44 | 2.33 | 24.80 | 36.53 | 47.81 | 82.60 |
| Model | Parameter | Optimal parameter |
|---|---|---|
| LightGBM | Learning_rate | 0.2 |
| N_estimators | 1 000 | |
| Max_depth | 6 | |
| Num_leaves | 127 | |
| Min_child_samples | 30 | |
| Subsample | 0.9 | |
| Colsample_bytree | 0.9 | |
| Reg_alpha | 1 | |
| Reg_lambda | 0.5 | |
| N_jobs | 1 | |
| XGBoost | Learning_rate | 0.1 |
| N_estimators | 10 000 | |
| Max_depth | 3 | |
| Min_child_weight | 1 | |
| Subsample | 0.9 | |
| Colsample_bytree | 0.6 | |
| Gamma | 0.1 | |
| Reg_alpha | 0 | |
| Reg_lambda | 3 | |
| CatBoost | Learning_rate | 0.15 |
| Iterations | 1 000 | |
| Depth | 3 | |
| L2_leaf_reg | 1 | |
| Border_count | 128 | |
| Random_strength | 2 | |
| Bagging_temperature | 2 | |
| RF | N_estimators | 200 |
| Max_depth | 10 | |
| Min_samples_split | 5 | |
| Min_samples_leaf | 1 | |
| Max_features | 0.7 | |
| Bootstrap | False |
Table 2 Best parameters of models
| Model | Parameter | Optimal parameter |
|---|---|---|
| LightGBM | Learning_rate | 0.2 |
| N_estimators | 1 000 | |
| Max_depth | 6 | |
| Num_leaves | 127 | |
| Min_child_samples | 30 | |
| Subsample | 0.9 | |
| Colsample_bytree | 0.9 | |
| Reg_alpha | 1 | |
| Reg_lambda | 0.5 | |
| N_jobs | 1 | |
| XGBoost | Learning_rate | 0.1 |
| N_estimators | 10 000 | |
| Max_depth | 3 | |
| Min_child_weight | 1 | |
| Subsample | 0.9 | |
| Colsample_bytree | 0.6 | |
| Gamma | 0.1 | |
| Reg_alpha | 0 | |
| Reg_lambda | 3 | |
| CatBoost | Learning_rate | 0.15 |
| Iterations | 1 000 | |
| Depth | 3 | |
| L2_leaf_reg | 1 | |
| Border_count | 128 | |
| Random_strength | 2 | |
| Bagging_temperature | 2 | |
| RF | N_estimators | 200 |
| Max_depth | 10 | |
| Min_samples_split | 5 | |
| Min_samples_leaf | 1 | |
| Max_features | 0.7 | |
| Bootstrap | False |
| Group | Indicator | Model | |||
|---|---|---|---|---|---|
| RF | XGBoost | LightGBM | CatBoost | ||
| Full dataset | R2 | 0.93 | 0.95 | 0.95 | 0.96 |
| RMSE/MPa | 4.55 | 3.58 | 3.58 | 3.29 | |
| MAE/MPa | 3.19 | 2.34 | 2.48 | 2.25 | |
| No fly ash | R2 | 0.95 | 0.98 | 0.98 | 0.93 |
| RMSE/MPa | 3.71 | 2.52 | 2.79 | 4.44 | |
| MAE/MPa | 2.79 | 1.76 | 1.94 | 2.70 | |
| No slag | R2 | 0.92 | 0.91 | 0.89 | 0.95 |
| RMSE/MPa | 4.69 | 4.21 | 4.71 | 3.39 | |
| MAE/MPa | 3.15 | 2.70 | 2.98 | 2.30 | |
| Early age | R2 | 0.78 | 0.86 | 0.87 | 0.92 |
| RMSE/MPa | 6.75 | 5.48 | 5.34 | 3.47 | |
| MAE/MPa | 4.56 | 3.25 | 3.33 | 2.51 | |
| Medium age | R2 | 0.73 | 0.82 | 0.83 | 0.87 |
| RMSE/MPa | 7.31 | 5.99 | 5.93 | 5.35 | |
| MAE/MPa | 4.89 | 4.14 | 4.02 | 3.89 | |
| Late age | R2 | 0.88 | 0.93 | 0.92 | 0.93 |
| RMSE/MPa | 4.10 | 3.54 | 3.64 | 2.90 | |
| MAE/MPa | 3.09 | 2.32 | 2.46 | 2.23 | |
Table 3 Error evaluation metrics of models
| Group | Indicator | Model | |||
|---|---|---|---|---|---|
| RF | XGBoost | LightGBM | CatBoost | ||
| Full dataset | R2 | 0.93 | 0.95 | 0.95 | 0.96 |
| RMSE/MPa | 4.55 | 3.58 | 3.58 | 3.29 | |
| MAE/MPa | 3.19 | 2.34 | 2.48 | 2.25 | |
| No fly ash | R2 | 0.95 | 0.98 | 0.98 | 0.93 |
| RMSE/MPa | 3.71 | 2.52 | 2.79 | 4.44 | |
| MAE/MPa | 2.79 | 1.76 | 1.94 | 2.70 | |
| No slag | R2 | 0.92 | 0.91 | 0.89 | 0.95 |
| RMSE/MPa | 4.69 | 4.21 | 4.71 | 3.39 | |
| MAE/MPa | 3.15 | 2.70 | 2.98 | 2.30 | |
| Early age | R2 | 0.78 | 0.86 | 0.87 | 0.92 |
| RMSE/MPa | 6.75 | 5.48 | 5.34 | 3.47 | |
| MAE/MPa | 4.56 | 3.25 | 3.33 | 2.51 | |
| Medium age | R2 | 0.73 | 0.82 | 0.83 | 0.87 |
| RMSE/MPa | 7.31 | 5.99 | 5.93 | 5.35 | |
| MAE/MPa | 4.89 | 4.14 | 4.02 | 3.89 | |
| Late age | R2 | 0.88 | 0.93 | 0.92 | 0.93 |
| RMSE/MPa | 4.10 | 3.54 | 3.64 | 2.90 | |
| MAE/MPa | 3.09 | 2.32 | 2.46 | 2.23 | |
| [1] | 金伟良, 赵羽习. 混凝土结构耐久性研究的回顾与展望[J]. 浙江大学学报(工学版), 2002, 36(4): 371-380+403. |
| JIN W L, ZHAO Y X. State-of-the-art on durability of concrete structures[J]. Journal of Zhejiang University (Engineering Science), 2002, 36(4): 371-380+403 (in Chinese). | |
| [2] |
LIU J, ZHANG B D, LI Y, et al. Modeling and analysis of fiber-reinforced high-performance concrete strength prediction based on nonlinear programming[J]. Construction and Building Materials, 2022, 322: 126421.
DOI URL |
| [3] | 严久鑫. 混凝土强度快速发展的影响机制研究[D]. 徐州: 中国矿业大学, 2021. |
| YAN J X. Research on influence mechanism of strength development of fast-strength concrete[D]. Xuzhou: China University of Mining and Technology, 2021 (in Chinese). | |
| [4] | 余晗之, 周理安, 刘 宇, 等. 基于砂浆膜厚度的GA-BP神经网络混凝土抗压强度预测研究[J]. 硅酸盐通报, 2026, 45(2): 437-448. |
| YU H Z, ZHOU L A, LIU Y, et al. Prediction of concrete compressive strength by GA-BP neural network based on mortar film thickness[J]. Bulletin of the Chinese Ceramic Society, 2026, 45(2): 437-448 (in Chinese). | |
| [5] |
LI J Q, WU Z M, SHI C J, et al. Durability of ultra-high performance concrete: a review[J]. Construction and Building Materials, 2020, 255: 119296.
DOI URL |
| [6] | 余 睿, 陈飞翔, 范定强, 等. 基于机器学习技术的超高性能混凝土性能预测与精细化设计[J]. 硅酸盐学报, 2025, 53(5): 1165-1176. |
| YU R, CHEN F X, FAN D Q, et al. Machine learning-based performance prediction and precision design of ultra-high performance concrete[J]. Journal of the Chinese Ceramic Society, 2025, 53(5): 1165-1176 (in Chinese). | |
| [7] | 刘 晓, 王思迈, 卢 磊, 等. 机器学习预测混凝土材料耐久性的研究进展[J]. 硅酸盐学报, 2023, 51(8): 2062-2073. |
| LIU X, WANG S M, LU L, et al. Development on machine learning for durability prediction of concrete materials[J]. Journal of the Chinese Ceramic Society, 2023, 51(8): 2062-2073 (in Chinese). | |
| [8] | 苏 强, 胡以婵. 基于RF算法的混凝土工作性能预测方法研究[J]. 价值工程, 2025, 44(12): 43-45. |
| SU Q, HU Y C. Research on concrete workability prediction method based on RF algorithm[J]. Value Engineering, 2025, 44(12): 43-45 (in Chinese). | |
| [9] |
CHUN P J, UJIKE I, MISHIMA K, et al. Random forest-based evaluation technique for internal damage in reinforced concrete featuring multiple nondestructive testing results[J]. Construction and Building Materials, 2020, 253: 119238.
DOI URL |
| [10] | 朱文邦, 郑秀梅, 杨增增, 等. 基于遗传算法优化BP神经网络的沙漠砂混凝土强度预测[J]. 混凝土, 2024(5): 48-51+56. |
| ZHU W B, ZHENG X M, YANG Z Z, et al. Strength prediction of desert sand concrete based on genetic algorithm optimized BP neural network[J]. Concrete, 2024(5): 48-51+56 (in Chinese). | |
| [11] | 徐潇航, 胡张莉, 刘加平, 等. 基于机器学习回归模型的三峡大坝混凝土强度预测[J]. 材料导报, 2023, 37(2): 41-49. |
| XU X H, HU Z L, LIU J P, et al. Concrete strength prediction of the Three Gorges Dam based on machine learning regression model[J]. Materials Reports, 2023, 37(2): 41-49 (in Chinese). | |
| [12] | 章伟琪, 王辉明. 混凝土抗压强度的可解释深度学习预测模型[J]. 东北大学学报(自然科学版), 2024, 45(5): 738-744+752. |
| ZHANG W Q, WANG H M. Interpretable deep learning prediction model for compressive strength of concrete[J]. Journal of Northeastern University (Natural Science), 2024, 45(5): 738-744+752 (in Chinese). | |
| [13] |
ASADI SHAMSABADI E, ROSHAN N, HADIGHEH SALI, et al. Machine learning-based compressive strength modelling of concrete incorporating waste marble powder[J]. Construction and Building Materials, 2022, 324: 126592.
DOI URL |
| [14] |
KANG M C, YOO D Y, GUPTA R. Machine learning-based prediction for compressive and flexural strengths of steel fiber-reinforced concrete[J]. Construction and Building Materials, 2021, 266: 121117.
DOI URL |
| [15] | 曹 源. 基于贝叶斯优化的CatBoost模型在混凝土抗压强度预测中的研究[J]. 混凝土, 2025(7): 87-94. |
| CAO Y. Investigation into the prediction of concrete compressive strength using a Bayesian-optimized CatBoost model[J]. Concrete, 2025(7): 87-94 (in Chinese). | |
| [16] | 薛 飞. 基于SMOTE-XGBoost算法的混凝土强度预测[J]. 混凝土与水泥制品, 2024(8): 32-36. |
| XUE F. Prediction of strength of concrete based on SMOTE-XGBoost algorithm[J]. China Concrete and Cement Products, 2024(8): 32-36 (in Chinese). | |
| [17] | 张 健. 具有异常值检测功能的基于LightGBM模型的混凝土坍落度回归预测方法[J]. 城市建设理论研究(电子版), 2025(15): 52-54. |
| ZHANG J. A LightGBM-based concrete slump regression prediction method with outlier detection function[J]. Theoretical Research in Urban Construction, 2025(15): 52-54 (in Chinese). | |
| [18] | 覃 源, 薛 存, 李 遥, 等. 盐冻耦合作用下水工混凝土耐久性及寿命预测[J]. 水力发电学报, 2024, 43(2): 110-122. |
| QIN Y, XUE C, LI Y, et al. Durability and lifespan predictions of hydraulic concrete under salt freezing coupling effect[J]. Journal of Hydroelectric Engineering, 2024, 43(2): 110-122 (in Chinese). | |
| [19] | 刘聪林, 李 盛, 崔晓宁, 等. 融合XGBoost和SHAP的混凝土抗压强度预测分析模型[J]. 水利水电技术(中英文), 2025, 56(2): 246-258. |
| LIU C L, LI S, CUI X N, et al. Prediction and analysis of concrete compressive strength based on XGBoost and SHAP[J]. Water Resources and Hydropower Engineering, 2025, 56(2): 246-258 (in Chinese). | |
| [20] |
PARUTHI S, VERMA R, SHARMA N, et al. Ensemble machine learning models for predicting strength of concrete with foundry sand and coal bottom ash as fine aggregate replacements[J]. Scientific Reports, 2025, 15: 38331.
DOI |
| [21] | 程 琳, 袁喜娜, 马春辉, 等. 基于可解释机器学习的混凝土重力坝变形安全监控模型[J]. 水利水电科技进展, 2025, 45(3): 77-85. |
| CHENG L, YUAN X N, MA C H, et al. Deformation safety monitoring model of concrete gravity dam based on interpretable machine learning[J]. Advances in Science and Technology of Water Resources, 2025, 45(3): 77-85 (in Chinese). | |
| [22] | 康天蓓, 梁 玉, 梁意博, 等. 基于多机器学习模型的再生混凝土抗盐冻性能预测[J]. 材料导报, 2025, 39(19): 104-114. |
| KANG T B, LIANG Y, LIANG Y B, et al. Prediction of salt-freeze resistance of recycled aggregate concrete based on multiple machine learning models[J]. Materials Reports, 2025, 39(19): 104-114 (in Chinese). | |
| [23] |
BREIMAN L. Random forests[J]. Machine Learning, 2001, 45(1): 5-32.
DOI |
| [24] | CHEN T Q, GUESTRIN C. XGBoost: a scalable tree boosting system[C]// Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. San Francisco California USA. ACM, 2016: 785-794. |
| [25] | KE G L, MENG Q, FINLEY T, et al. LightGBM: a highly efficient gradient boosting decision tree[C]// NIPS’17: Proceedings of the 31st International Conference on Neural Information Processing Systems, December 4-9, 2017, Long Beach, California, USA. New York: Curran Associates Inc., 2017: 3149-3157. |
| [26] | PROKHORENKOVA L, GUSEV G, VOROBEV A, et al. CatBoost: unbiased boosting with categorical features[C]// NIPS’18: Proceedings of the 32nd International Conference on Neural Information Processing Systems, December 3-8, 2018, Montréal, Canada. New York: Curran Associates Inc., 2018: 6639-6649. |
| [27] | LUNDBERG S M, LEE S I. A unified approach to interpreting model predictions[C]// Proceedings of the 31st International Conference on Neural Information Processing Systems. ACM, 2017: 4768-4777. |
| [28] |
YEH I C. Modeling of strength of high-performance concrete using artificial neural networks[J]. Cement and Concrete Research, 1998, 28(12): 1797-1808.
DOI URL |
| [1] | LIANG Qimin, WANG Zhe, MEI Yingjie, SUN Ao, LI Pengfei. Workability Prediction of Self-Compacting Concrete Based on CatBoost Optimization Algorithm [J]. BULLETIN OF THE CHINESE CERAMIC SOCIETY, 2025, 44(9): 3178-3187. |
| Viewed | ||||||
|
Full text |
|
|||||
|
Abstract |
|
|||||