冻融循环下玄武岩纤维混凝土冲击力学性能预测模型  

A predictive model of the impact mechanical properties of basalt fiber reinforced concrete under freeze-thaw cycling

作　　者：李艳[1] 何峻宇 翟越[1] 李昌昊 贾宇 谢梓涵 殷溥隆 梁文彪 LI Yan;HE Junyu;ZHAI Yue;LI Changhao;JIA Yu;XIE Zihan;YIN Pulong;LIANG Wenbiao(School of Geological Engineering and Geomatics,Chang’an University,Xi’an 710061,China;Northwest Branch of China Construction Eighth Engineering Bureau Co.,Ltd.,Xi’an 710065,China;School of Sciences,Chang’an University,Xi’an 710061,China)

机构地区：[1]长安大学地质工程与测绘学院,陕西西安710061 [2]中国建筑第八工程局有限公司西北公司,陕西西安710065 [3]长安大学理学院,陕西西安710061

出　　处：《西安科技大学学报》2024年第4期768-777,共10页Journal of Xi’an University of Science and Technology

基　　金：国家自然科学基金项目(42202307);咸阳市科技创新团队项目(L2023CXNLCXTD005);陕西省博士后自然科学基金项目(2023BSHEDZZ211);中央高校创新团队项目(300102262402);陕西省联合基金项目(2022JC-LHJJ-16)。

摘　　要：冻融循环下纤维混凝土的劣化规律是寒区服役混凝土工程安全性和耐久性评价的重要依据,现有服役混凝土工程的安全性和耐久性评价的研究具有工作量大、成本高、周期长等特点,构建基于机器学习的高精度力学性能预测模型已成为本领域研究热点。为探究冻融循环后玄武岩纤维混凝土冲击力学性能的高精度预测模型,采用SHPB装置对冻融循环后BFRC开展动态冲击压缩力学性能试验,并构建机器学习-Optuna混合预测模型,对60组以玄武岩纤维体积掺量、冻融循环次数、动荷载冲击速度为影响因素建立的动态峰值应力样本数据集进行预测。结果表明:k近邻、Lasso、多层感知机、极度梯度提升树和随机森林5种经典机器学习模型的预测准确度均较高,说明机器学习算法对于冻融循环后BFRC动态力学性能预测具有良好的预测效果,其中随机森林算法为最优预测算法;RF-Optuna混合预测模型显示出0.9754的拟合优度,具有良好的预测精度;非数据集工况预测表明,该混合模型对于各影响因素均具有良好泛化能力。研究成果可为冻融循环条件下BFRC动态力学性能的快捷精准预测提供参考。The deterioration law of fiber concrete under freeze-thaw cycling is an important basis for the evaluation of safety and durability in in-service concrete projects in cold regions.The research in this area is characterized by its extensive workload,high costs,and lengthy cycles.Consequently,the development of high-precision mechanical property prediction model based on machine learning has emerged as a focal point of research within this field.In order to investigate a high-precision prediction model for the impact mechanical properties of basalt fiber reinforced concrete under freeze-thaw cycling,the SHPB device was used to conduct dynamic impact compression mechanical property tests on BFRC that had undergone freeze-thaw cycling.Additionally,a hybrid predictive model combining machine learning with-Optuna was constructed to forecast the dynamic peak stress sample dataset of 60 groups,which were established based on such factors as the volume fraction of basalt fibers,the number of freeze-thaw cycles,and the velocity of dynamic impact loads.The results show that:the predictive performance of classic machine learning models,including k-Nearest Neighbor,Lasso,Multi-layer Perceptron,eXtreme Gradient Boosting,and Random Forest,is notably high,indicating that the machine learning algorithm has a good predictive effect on the dynamic mechanical performance prediction of BFRC after freeze-thaw cycles.Among them,the random forest algorithm is the optimal prediction algorithm;The RF-Optuna hybrid prediction model constructed shows a goodness of fit of 0.9754,with good prediction accuracy;non dataset operating condition prediction of the hybrid model shows that the hybrid model has good generalization ability for various influencing factors.The research results can serve as a reference for rapid and precise forecasting of the dynamic mechanical properties of BFRC under freeze-thaw cycling conditions.

关 键 词：动态力学性能预测 玄武岩纤维混凝土 冻融循环 随机森林模型 Optuna框架优化 

分 类 号：TU528[建筑科学—建筑技术科学]