基于BP神经网络的寒区再生微粉工程水泥基复合材料力学性能研究  

Mechanical properties of recycled micropowder engineered cementitiouscomposites in cold region based on BP neural network

作　　者：纪泳丞[1] 季文昊 贾艳敏[1] 李泽闯 李艺铭 王锐[1] JI Yongcheng;JI Wenhao;JIA Yanmin;LI Zechuang;LI Yiming;WANG Rui(College of Civil Engineering and Transportation,Northeast Forestry University,Harbin 150040,China)

机构地区：[1]东北林业大学土木与交通学院,黑龙江哈尔滨150040

出　　处：《冰川冻土》2024年第1期126-136,共11页Journal of Glaciology and Geocryology

基　　金：国家自然科学基金项目(52370128)资助。

摘　　要：为了探究不同低温不利条件下再生微粉ECC材料力学性能的影响,本文采用控制变量法,以再生微粉种类和取代率为研究变量,探究了再生微粉ECC在冻融循环和恒低温两种低温不利条件下的抗压和抗折强度试验。分析了再生微粉种类、再生微粉取代率、冻融循环次数、恒低温温度对再生微粉ECC力学性能的影响。最后基于BP神经网络,建立了3-6-1的冻融循环和3-3-1的恒低温BP神经网络结构抗压强度预测模型。研究结果表明:在相同的冻融循环条件下,再生混凝土微粉ECC的力学性能要高于再生砖粉ECC,且均随再生微粉取代率的增加先小幅度下降后剧烈下降,在经历150冻融循环后力学性能损失20%左右。而经历恒低温保温后的再生微粉ECC力学性能呈现出相反的变化趋势,随着低温保温温度的降低再生微粉ECC的力学强度反而呈现上升趋势,从常温到-40℃恒低温状态下力学性能提高22%左右。建立的两个低温不利条件下BP神经网络预测模型,平均相对误差分别为1.43%、1.28%,并以质量损失率和相对动弹模量为评判标准,预测试验范围内不同配合比的再生微粉ECC可经受的最大冻融循环次数。The mechanical properties of recycled micro powder engineered cementitious composites(ECCs)un-der various low-temperature conditions are investigated.The control variable method explores the ECC’s com-pressive and flexural strength concerning two low-temperature conditions:various freeze-thaw cycles and con-stant negative temperatures,with different types of recycled micro powders and replacement rates as variables.The effects of recycled micro powder type,replacement rate,number of freeze-thaw cycles,and negative tem-perature on the mechanical properties of ECC were analyzed.Moreover,two prediction models were presented to calculate their compressive strengths based on 3-6-1 and 3-3-1 BP neural network structures for two conditions of freeze-thaw cycle and constant negative temperature.Results show that the mechanical strengths of recycled concrete micropowder ECC are higher than those of recycled brick micro powder ECC under the same freeze-thaw conditions.With the increase in the replacement rate of recycled micro powder,a tiny decline and a sharp decline were observed,as well as a 20%loss of mechanical performance for 150 freeze-thaw cycles.However,the mechanical properties of the recycled micro powder ECC showed the opposite trend after experiencing con-stant low-temperature insulation.The recycled micro powder ECC mechanical strength showed an upward trend of 22%,from room temperature to-40℃constant low-temperature state.The two BP neural network prediction models were established under low-temperature conditions with an average relative error of 1.43%and 1.28%,respectively.The maximum freeze-thaw cycles of recycled micro powder ECC with different mix ratios were predicted based on the mass loss rate and relative dynamic modulus.

关 键 词：再生微粉ECC 冻融循环 低温保温 BP神经网络 力学性能 

分 类 号：TU528[建筑科学—建筑技术科学]