结构工程师2025，Vol.41Issue(6)：22-30,9.DOI:10.15935/j.cnki.jggcs.202506.0003
基于不同机器学习方法的UHPC抗压强度预测
Prediction of UHPC Compressive Strength Based on Different Machine Learning Methods
摘要
Abstract
Ultra-High Performance Concrete(UHPC)is a high-performance cementitious composite material characterized by ultra-high compressive strength,superior toughness,enhanced ductility,and excellent durability.However,a systematic and standardized framework for UHPC mix design remains lacking.In both engineering applications and scientific research,mix proportioning still heavily relies on empirical approaches.To address this gap,this study compiled a comprehensive database consisting of 300 mix designs and corresponding experimental material properties from a thorough review of the existing UHPC literature.The database incorporates seven key influencing factors:cement content,fly ash content,silica fume content,quartz sand content,superplasticizer dosage,steel fiber content,and water content,along with their experimentally measured compressive strengths.Based on this dataset,several machine learning algorithms—including Random Forest(RF),Extreme Learning Machine(ELM),Support Vector Regression(SVR),and Backpropagation Neural Networks(BPNN)—were employed to develop predictive models for the compressive strength of UHPC.Furthermore,a machine learning-driven mix design methodology was proposed to achieve targeted compressive strength.The results demonstrate that all adopted machine learning models can provide reasonably accurate preliminary predictions of compressive strength based on the constituent proportions,with the coefficient of determination(R²)of the training set exceeding 0.8,indicating satisfactory predictive performance.The proposed method effectively generates UHPC mix proportions that meet specified strength requirements,offering a data-driven alternative to conventional empirical design.关键词
超高性能混凝土/机器学习/配比/抗压强度预测Key words
ultra-high performance concrete/machine learning/mix design/compressive strength prediction分类
建筑与水利引用本文复制引用
杨毅超,张金帆,潘钻峰,顾征宇,程欣玥..基于不同机器学习方法的UHPC抗压强度预测[J].结构工程师,2025,41(6):22-30,9.基金项目
国家自然科学基金(52378186) （52378186）
