硅酸盐学报

基于机器学习的小样本超高性能混凝土性能预测及可解释性分析

PDF下载

张增起 ¹^,² , 杜伟杰 ² , 刘宇 ³ , 马善亮 ² , 刘晓明 ¹^,² , 荆锐 ⁴

硅酸盐学报 | 第二十八届中国科协年会智能建造关键无机非金属材料–AI驱动水泥基材料创新专题-研究论文 2026,54(10): 3436-3452

收起

硅酸盐学报 |第二十八届中国科协年会智能建造关键无机非金属材料–AI驱动水泥基材料创新专题-研究论文 2026 , 54 (10) : 3436 -3452

基于机器学习的小样本超高性能混凝土性能预测及可解释性分析

全屏

张增起¹^,², 杜伟杰², 刘宇³, 马善亮², 刘晓明¹^,², 荆锐⁴

作者信息

^1.北京科技大学绿色低碳钢铁冶金全国重点实验室，北京　100083

^2.北京科技大学冶金与生态工程学院，北京　100083

^3.中建一局集团建设发展有限公司，北京　100102

^4.唐山工业职业技术大学，河北　唐山　063299

作者简介:

张增起（1991-），男，博士，教授。

Performance Prediction and Interpretability Analysis of Ultra-High-Performance Concrete with Small Sample Size Based on Machine Learning

Zengqi ZHANG¹^,², Weijie DU², Yu LIU³, Shanliang MA², Xiaoming LIU¹^,², Rui JING⁴

Affiliations

^1.State Key Laboratory of Advanced Metallurgy, University of Science and Technology Beijing, Beijing 100083, China

^2.School of Metallurgical and Ecological Engineering, University of Science and Technology Beijing, Beijing 100083, China

^3.China Construction First Group Construction and Development Co., Ltd., Beijing 100102, China

^4.Tangshan Polytechnic University, Tangshan 063299, Hebei, China

ZHANG zengqi (1991-), male, Ph.D. Professor. E-mail: zhangzq@ustb.edu.cn

出版时间: 2026-09-01 doi: 10.14062/j.issn.0454-5648.20260296

文章导航

摘要

收起

为在小样本条件下实现超高性能混凝土（UHPC）性能的准确预测，并揭示多组分胶凝材料对UHPC力学性能的影响规律，融合了条件生成对抗网络（CTGAN）、贝叶斯优化的极限梯度提升树（BO-XGBoost）模型以及可解释性机器学习方法，建立了一套用于多响应预测与分析的框架。以160条实验数据为原始数据集，采用CTGAN生成合成样本进行数据增强，建立以矿物掺合料类型和掺量、减水剂掺量和养护龄期为输入，水膜层厚度与抗压强度为输出的BO-XGBoost预测模型，并结合沙普利可加性解释值（SHAP）和部分依赖图（PDPs）分析各变量对输出性能的影响规律。结果表明，利用CTGAN生成的数据在分布形态与变量相关性上与真实数据保持高度一致，有效缓解了小样本带来的数据不足的问题。BO-XGBoost模型在测试集上对水膜层厚度和抗压强度的预测决定系数（R²）分别达到0.986和0.976，误差分布集中且对称，表现出优异的预测精度与泛化能力。SHAP与PDPs分析表明了不同矿物掺合料对水膜层厚度与抗压强度的影响具有显著差异。粉煤灰微珠和硅灰在中低掺量下对水膜层厚度具有正向作用，而多种掺合料在高掺量条件下因颗粒堆积致密化表现为负向影响。对于抗压强度，超细矿渣粉、硅灰和粉煤灰微珠整体呈正向贡献，矿渣粉与粉煤灰在高掺量条件下因活性受限表现为负向作用。该框架为小样本条件下UHPC的配合比设计与性能优化提供了有效的数据驱动手段。

关键词

条件生成对抗网络 / 超高性能混凝土 / 数据增强 / 水膜层厚度 / 抗压强度

Abstract

收起

Introduction

Ultra-high-performance concrete (UHPC) is widely applied in bridges, prefabricated structures, and advanced construction owing to its high strength, dense microstructure, and excellent durability performance. Its performance is strongly governed by the type and dosage of supplementary cementitious materials (SCMs), including silica fume (SF), ground granulated blast-furnace slag (GGBS), superfine GGBS (sGGBS), fly ash (FA), and fly ash microspheres (sFA), as well as polycarboxylate superplasticizer (PCE) dosage and curing age (CA). However, the relationships among mixture parameters, water distribution, hydration, and strength development are highly nonlinear, and conventional empirical models often fail to provide accurate predictions.

Water film thickness (WFT) is a key microscopic parameter reflecting the free water covering the surfaces of solid particles, directly influencing particle lubrication, dispersion, and fresh-state behaviour in high-solid-volume UHPC systems. Compressive strength (CS) is another essential index for mechanical evaluation. Accurate prediction of both WFT and CS is crucial for mixture design and performance optimization. Nevertheless, UHPC experiments are costly and time-consuming, leading to limited datasets. To overcome this small-sample challenge, this study proposes an integrated framework combining conditional tabular generative adversarial network (CTGAN) for data augmentation, Bayesian optimization (BO)-based eXtreme Gradient Boosting (XGBoost) for predictive modelling, and interpretable machine-learning methods (SHAP and PDPs) for mechanism analysis.

Methods

A total of 160 experimental data points were obtained from binary, ternary, and quaternary cementitious systems. The input variables included cement content, PCE dosage, SF, GGBS, sGGBS, FA, sFA, and CA, while the output variables were WFT and CS. P·I 42.5 cement was used as the reference cement, and quartz sand with particle sizes of 0.84-2.00 mm was used as the aggregate. In the binary systems, one mineral admixture was used to replace cement by volume at replacement levels of 0%, 5%, 10%, 15%, and 20%. In the ternary systems, the SF content was fixed at 10%. In the quaternary systems, the GGBS content was fixed at 20% based on the ternary system. WFT was calculated using the wet packing method, and CS was measured after curing for 3, 7, 28 d, and 90 d.

CTGAN was used for data augmentation to expand the original small dataset. The model learned the joint distribution of the real experimental data through adversarial training between the generator and discriminator. After optimization, 1000 synthetic samples were generated for WFT prediction, and 900 synthetic samples were generated for CS prediction. The quality of the synthetic data was evaluated by comparing histograms, cumulative distribution curves, and correlation matrices between real and synthetic datasets.

The synthetic datasets were used to train XGBoost regression models. BO was applied to optimize hyperparameters, including tree depth, learning rate, number of estimators, subsample ratio, column-sampling ratio, minimum child weight, and regularization. Model performance was evaluated by R², MAPE, RMSE, and MAE. SHAP and PDPs were employed to interpret feature contributions and nonlinear effects.

Results and discussion

The comparison between real and synthetic datasets showed that CTGAN effectively preserved the main distributional characteristics of the original data. The generated samples had value ranges and distribution patterns similar to those of the real samples. The cumulative distribution curves also showed good consistency between real and synthetic datasets. In addition, the correlation matrices indicated that CTGAN captured not only the marginal distribution of individual variables but also the correlation structure among key variables. Therefore, the generated data could provide a reliable data basis for BO-XGBoost model training under small-sample conditions.

The BO-XGBoost model exhibited excellent prediction performance. For WFT, test R² reached 0.986, with training and test RMSE values of 0.909 3 nm and 0.387 6 nm, respectively. For CS, test R² was 0.976, with RMSE values of 5.244 8 MPa and 3.837 9 MPa. Predicted values clustered around the 45° ideal line, with most samples within ±10% error. Residuals were concentrated and symmetric, indicating no systematic bias, and loss curves showed stable convergence without overfitting. These results demonstrate that CTGAN augmentation effectively improved the ability of XGBoost to capture complex nonlinear relationships under small-sample conditions.

The SHAP and PDP analyses further revealed the influence mechanisms of different variables. For WFT, sFA showed a strong positive contribution, mainly because its fine and spherical particles improved particle dispersion and increased effective water-film formation. SF had a positive effect at low-to-medium dosages, but excessive SF could reduce WFT because of its high specific surface area. GGBS and sGGBS tended to show negative effects at high dosages, which may be related to denser particle packing and compression of effective water-film space. PCE showed a nonmonotonic effect: moderate PCE improved WFT, while excessive PCE could weaken the stability of the water-film structure.

For CS, CA was the most important factor and showed a strong positive contribution, reflecting continuous hydration and microstructure densification with curing time. sGGBS, SF, and sFA generally promoted strength development. sGGBS contributed to strength through microfilling and potential pozzolanic activity, while SF improved strength through pore refinement and matrix densification. In contrast, GGBS and FA showed negative effects at high dosages because excessive cement replacement reduced clinker content and limited the availability of calcium hydroxide (Ca(OH)₂) for secondary reactions. These results demonstrate that different mineral admixtures have dosage-dependent and material-specific effects on UHPC performance.

Conclusions

This study establishes a multi-response prediction and interpretation framework integrating CTGAN, BO-XGBoost, SHAP, and PDPs for UHPC under small-sample conditions. CTGAN effectively generated high-quality synthetic data while preserving original distribution and correlation characteristics. The optimized BO-XGBoost model achieved outstanding predictive accuracy and generalization, with test R² of 0.986 for WFT and 0.976 for CS. The interpretability analysis demonstrated that the effects of mineral admixtures on WFT and CS are highly dependent on material type and dosage: sFA and SF are beneficial for WFT within suitable ranges, whereas sGGBS, SF, and sFA promote CS; excessive GGBS and FA impair CS due to limited reaction activity. The proposed framework offers an effective data-driven approach for UHPC mixture design, performance prediction, and mechanism interpretation when only limited experimental data are available.

Key words

conditional generative adversarial network / ultra-high-performance concrete / data augmentation / water film thickness / compressive strength

引用本文

张增起, 杜伟杰, 刘宇, 马善亮, 刘晓明, 荆锐. 基于机器学习的小样本超高性能混凝土性能预测及可解释性分析. 硅酸盐学报, 2026 , 54 (10) : 3436 -3452 . DOI: 10.14062/j.issn.0454-5648.20260296

Zengqi ZHANG, Weijie DU, Yu LIU, Shanliang MA, Xiaoming LIU, Rui JING. Performance Prediction and Interpretability Analysis of Ultra-High-Performance Concrete with Small Sample Size Based on Machine Learning[J]. Journal of the Chinese Ceramic Society, 2026 , 54 (10) : 3436 -3452 . DOI: 10.14062/j.issn.0454-5648.20260296

基金

收起

国家自然科学基金面上项目(52478232)

参考文献

收起

文献

收起

参考文献引证文献

排序方式：

[1]

刘一帆, 吴泽媚, 张轩翰, 等. 超高性能混凝土流变特性及调控研究进展[J]. 硅酸盐学报, 2023, 51(11): 3025-3038.

LIU Yifan, WU Zemei, ZHANG Xuanhan, et al. J Chin Ceram Soc, 2023, 51(11): 3025-3038.

[2]

樊健生, 丁然. 超高性能混凝土在房屋建筑结构中的研究与应用进展[J]. 硅酸盐学报, 2023, 51(5): 1246-1258.

FAN Jiansheng, DING Ran. J Chin Ceram Soc, 2023, 51(5): 1246-1258.

[3]

FAN D Q, ZHU J Y, FAN M X, et al. Intelligent design and manufacturing of ultra-high performance concrete (UHPC)-A review[J]. Constr Build Mater, 2023, 385: 131495.

[4]

张云升, 张文华, 陈振宇. 综论超高性能混凝土: 设计制备·微观结构·力学与耐久性·工程应用[J]. 材料导报, 2017, 31(23): 1-16.

ZHANG Yunsheng, ZHANG Wenhua, CHEN Zhenyu. Mater Rev, 2017, 31(23): 1-16.

[5]

SCHRÖFL C, GRUBER M, PLANK J. Preferential adsorption of polycarboxylate superplasticizers on cement and silica fume in ultra-high performance concrete (UHPC)[J]. Cem Concr Res, 2012, 42(11): 1401-1408.

[6]

KHAYAT K H, MENG W N, VALLURUPALLI K, et al. Rheological properties of ultra-high-performance concrete: An overview[J]. Cem Concr Res, 2019, 124: 105828.

[7]

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 Rui, CHEN Feixiang, FAN Dingqiang, et al. J Chin Ceram Soc, 2025, 53(5): 1165-1176.

[8]

LIU H R, SUN X, DU H, et al. Effects and threshold of water film thickness on multi-mineral cement paste[J]. Cem Concr Compos, 2020, 112: 103677.

[9]

王圣杰, 李兵, 李传习, 等. 超高性能混凝土流动性与流变性关系[J]. 硅酸盐学报, 2023, 51(8): 1962-1970.

WANG Shengjie, LI Bing, LI Chuanxi, et al. J Chin Ceram Soc, 2023, 51(8): 1962-1970.

[10]

YE H, GAO X J, WANG R, et al. Relationship among particle characteristic, water film thickness and flowability of fresh paste containing different mineral admixtures[J]. Constr Build Mater, 2017, 153: 193-201.

[11]

FAN D Q, YU R, FU S Y, et al. Precise design and characteristics prediction of Ultra-High Performance Concrete (UHPC) based on artificial intelligence techniques[J]. Cem Concr Compos, 2021, 122: 104171.

[12]

AYLAS-PAREDES B K, HAN T H, NEITHALATH A, et al. Data driven design of ultra high performance concrete prospects and application[J]. Sci Rep, 2025, 15: 9248.

[13]

HE Y Z, GAO S J, LI Y, et al. Adaptive machine learning framework: Predicting UHPC performance from data to modelling[J]. Results Eng, 2025, 27: 106724.

[14]

周祥胥, 段锋, 朱博. 基于机器学习的SHAP和PDP分析对UHPC流变性能的研究[J]. 材料导报, 2026, 40(2): 86-93.

ZHOU Xiangxu, DUAN Feng, ZHU Bo. Mater Rep, 2026, 40(2): 86-93.

[15]

余睿, 李旺, 范定强. 基于条件生成对抗网络与马尔可夫链蒙特卡洛双驱动的超高性能混凝土数据增强及智能预测方法[J]. 硅酸盐学报, 2026, 54(3): 842-856.

YU Rui, LI Wang, FAN Dingqiang. J Chin Ceram Soc, 2026, 54(3): 842-856.

[16]

孙浚博, 王雨飞, 赵宏宇, 等. 基于深度学习的3D打印混凝土蒸汽养护力学性能研究和抗压强度预测[J]. 硅酸盐通报, 2024, 43(5): 1723-1738.

SUN Junbo, WANG Yufei, ZHAO Hongyu, et al. Bull Chin Ceram Soc, 2024, 43(5): 1723-1738.

[17]

WANG Y F, SUN J B, WANG X Y, et al. Multi-objective optimization of engineered cementitious composite based on machine learning and generative adversarial network[J]. J Build Eng, 2024, 96: 110471.

[18]

WANG S Q, XIA P, GONG F Y, et al. A Bayesian-physical informed conditional tabular generative adversarial network framework for low-carbon concrete data augmentation and hyperparameter optimization[J]. Eng Appl Artif Intell, 2025, 152: 110811.

[19]

HUU NGUYEN M, NGUYEN T A, LY H B. Ensemble XGBoost schemes for improved compressive strength prediction of UHPC[J]. Structures, 2023, 57: 105062.

[20]

HARUNA S I, IBRAHIM Y E, ABBA S I. Machine learning algorithm for predicting the compressive strength of sustainable ultra-high-performance concrete (UHPC) containing supplementary cementitious materials (SCMs)[J]. J Asian Archit Build Eng, 2025: 1-17.

[21]

KASHEM A, KARIM R, MALO S C, et al. Hybrid data-driven approaches to predicting the compressive strength of ultra-high-performance concrete using SHAP and PDP analyses[J]. Case Stud Constr Mater, 2024, 20: e02991.

[22]

郭孟环, 谢咏菲, 周英武, 等. 基于多目标靶向性能的超高韧性地聚物复合材料智能化设计[J]. 硅酸盐学报, 2026, 54(3): 823-841.

GUO Menghuan, XIE Yongfei, ZHOU Yingwu, et al. J Chin Ceram Soc, 2026, 54(3): 823-841.

[23]

NGUYEN N H, ABELLÁN-GARCÍA J, LEE S, et al. Efficient estimating compressive strength of ultra-high performance concrete using XGBoost model[J]. J Build Eng, 2022, 52: 104302.

[24]

JAISWAL R, BHATTA N, MUHIT I B. An interpretable machine learning approach for shear capacity prediction of prestressed UHPC bridge girders with GUI deployment[J]. Structures, 2026, 88: 111793.

[25]

XU W Y, ZHANG L Y, FAN D Q, et al. AI-infused characteristics prediction and multi-objective design of ultra-high performance concrete (UHPC): From pore structures to macro-performance[J]. J Build Eng, 2024, 98: 111170.

[26]

CHEN J J, KWAN A K H. Triple blending with fly ash microsphere and condensed silica fume to improve performance of cement paste[J]. J Mater Civ Eng, 2013, 25(5): 618-626.

[27]

MARCHETTI G, RAHHAL V F, IRASSAR E F. Influence of packing density and water film thickness on early-age properties of cement pasteswith limestone filler and metakaolin[J]. Mater Struct, 2017, 50(2): 111.

[28]

LI L G, KWAN A K H. Effects of superplasticizer type on packing density, water film thickness and flowability of cementitious paste[J]. Constr Build Mater, 2015, 86: 113-119.

[29]

PHILIP S, MARAKKATH N. Compressive strength prediction and feature analysis for GGBS-Based geopolymer concrete using optimized XGBoost and SHAP: A comparative study of optimization algorithms and experimental validation[J]. J Build Eng, 2025, 108: 112879.

[30]

KWAN A K H, LI Y. Effects of fly ash microsphere on rheology, adhesiveness and strength of mortar[J]. Constr Build Mater, 2013, 42: 137-145.

[31]

RANDL N, STEINER T, OFNER S, et al. Development of UHPC mixtures from an ecological point of view[J]. Constr Build Mater, 2014, 67: 373-378.

[32]

ZOU S, SUN Z P, LU Z C, et al. Fresh and hardened properties of cement paste under the synergistic influence of polycarboxylate superplasticizer and anionic polyacrylamide[J]. Case Stud Constr Mater, 2024, 21: e03895.

[33]

MOULA S, BEN FRAJ A, WATTEZ T, et al. The very early-age behaviour of Ultra-High Performance Concrete containing ground granulated blast furnace slag[J]. Constr Build Mater, 2023, 400: 132630.

[34]

MOULA S, BEN FRAJ A, WATTEZ T, et al. Mechanical properties, carbon footprint and cost of ultra-high performance concrete containing ground granulated blast furnace slag[J]. J Build Eng, 2023, 79: 107796.

[35]

HU Q H, ZHANG L, LUO Q, et al. Impact of polycarboxylate superplasticizer dosage on controlled low strength material flowability and bleeding: Insights from water film thickness[J]. Constr Build Mater, 2024, 447: 138145.

[36]

XI J Y, LIU J Z, YANG K, et al. Role of silica fume on hydration and strength development of ultra-high performance concrete[J]. Constr Build Mater, 2022, 338: 127600.

2026年第54卷第10期

PDF下载

104

引用本文

BibTeX

文章信息

doi: 10.14062/j.issn.0454-5648.20260296

接收时间：2026-05-01
首发时间：2026-10-22
出版时间：2026-09-01

补充材料

相关文章

文章信息

作者

出版历史

收稿日期：2026-05-01
修回日期：2026-06-13

基金

国家自然科学基金面上项目(52478232)

作者信息

^1.北京科技大学绿色低碳钢铁冶金全国重点实验室，北京　100083

^2.北京科技大学冶金与生态工程学院，北京　100083

^3.中建一局集团建设发展有限公司，北京　100102

^4.唐山工业职业技术大学，河北　唐山　063299

参考文献

分享链接

https://castjournals.cast.org.cn/joweb/gsyxb/CN/10.14062/j.issn.0454-5648.20260296

分享至

全文二维码

扫描看全文

引用本文

BibTeX

本文的引用情况

2种不同金属材料的力学参数

科 Family	属数 Number of genus	种数 Number of species	占总种数比例 Percentage of total species (%)	属 Genus	种数 Number of species	占总种数比例 Percentage of total species (%)
鹅膏菌科Amanitaceae	2	11	5.26	鹅膏菌属 Amanita	10	4.78
小菇科 Mycenaceae	2	12	5.74	丝盖伞属 Inocybe	5	2.39
多孔菌科 Polyporaceae	8	14	6.70	蜡蘑属 Laccaria	5	2.39
红菇科 Russulaceae	3	23	11.00	小皮伞属 Marasmius	6	2.87
				小菇属 Mycena	11	5.26
				光柄菇属 Pluteus	5	2.39
				红菇属 Russula	17	8.13
				栓菌属 Trametes	5	2.39

关闭全屏

BibTeX
EndNote
RefWorks
TxT