基于BOA-GRU网络的混凝土抗压强度预测方法  

Concrete Compressive Strength Prediction Based on GRU Improved by Bayesian Optimization Algorithm

作　　者：刘立伟 邵海波 崔凤笠 侯中伟 LIU Liwei;SHAO Haibo;CUI Fengli;HOU Zhongwei(Qingdao Ruiyuan Engineering Group Co.,Ltd.Qingdao 266555,China)

机构地区：[1]青岛瑞源工程集团有限公司,青岛266555

出　　处：《广东土木与建筑》2025年第2期104-107,共4页Guangdong Architecture Civil Engineering

基　　金：青岛市企业技术创新重点项目(2320002000045,2320002005333)。

摘　　要：抗压强度是混凝土材料设计中的关键力学性能参数,可靠的强度预测可以减少设计成本和时间,并防止因大量配比实验而导致的材料浪费。为此,提出了一种基于门控循环单元(GRU)的混凝土抗压强度预测模型,并采用贝叶斯优化算法(BOA)对GRU的关键超参数进行优化。收集了1030组混凝土抗压强度数据对BOA-GRU模型进行验证,并与常规的BP神经网络、支持向量回归(SVR)模型的结果进行对比,结果显示BOA-GRU模型的预测精度和可靠性最高,其在均方根误差(RMSE)、平均绝对误差(MAE)、平均绝对百分比误差(MAPE)等3个指标上均优于对比模型。Compressive strength is a key mechanical property parameter in the design of concrete materials.Reliable strength prediction can reduce design costs and time,and prevent material waste due to extensive mixture experiments.To this end,proposes a concrete compressive strength prediction model based on Gated Recurrent Unit(GRU),and uses Bayesian Optimization Algorithm(BOA)to optimize the key hyperparameters of GRU.A total of 1030 sets of concrete compressive strength data were collected to verify the BOA-GRU model,and we compare its results with those of conventional BP neural networks and Support Vector Regression(SVR)models.The results show that the BOA-GRU model has the highest prediction accuracy and reliability,outperforming the comparison models in terms of Root Mean Square Error(RMSE),Mean Absolute Error(MAE),and Mean Absolute Percentage Error(MAPE).

关 键 词：混凝土 抗压强度 门控循环单元 贝叶斯优化算法 

分 类 号：TU528.1[建筑科学—建筑技术科学]