基于MLP神经网络的水泥分别粉磨强度预测模型与分析  

Prediction model and analysis of cement grinding strength based on MLP neural network

作　　者：黄雄 姚丕强[1] 杜鑫[1] 聂文海[1] 刘迪 HUANG Xiong(Tianjin Cement Industry Design&Research Institute Co.LTD.,Tianjin 300400,China)

机构地区：[1]天津水泥工业设计研究院有限公司,天津300400

出　　处：《水泥》2023年第5期33-37,共5页Cement

基　　金：中国建材集团攻关专项(2021HX0405);全国建材行业重大科技攻关“揭榜挂帅”项目(20221JBGS02-05)。

摘　　要：为研究分别粉磨配制水泥中多组分下各原料对强度的影响,提出利用多层感知器(multi-layer perceptron,MLP)神经网络建立模型的方法。选用矿粉、粉煤灰和石灰石粉作为混合材,进行单因素试验得到活性数据,设计分别粉磨配制水泥配比组成试验样本,通过训练数据构建MLP神经网络模型,归纳出相关因素与强度之间的表达式。结果表明:活性越高的混合材多掺或适当磨细更有助于提高强度,增加石灰石粉含量有助于提升3 d抗压强度;所建立模型对3 d和28 d龄期抗压强度的预测结果精准,平均相对误差小于2%。In order to study the influence of each raw materials on the strength of the multi-component cement prepared by grinding separately,a model based on multi-layer perceptron(MLP)neural network was proposed.Ground granulated blastfurnace slag,fly ash and limestone powder were selected as admixture,and the activity data were obtained by single factor experiment.The test samples of cement prepared by grinding separately were designed.The MLP neural network model was constructed through the training data,and the expressions between the related factors and strength were summarized.The results show that more mixing or grinding of admixture with higher activity is more conducive to improving the strength,and increasing the content of limestone powder is conducive to improving the 3 d compressive strength.The prediction results of the model for 3 d and 28 d compressive strength were accurate,and the average relative error is less than 2%.

关 键 词：分别粉磨 MLP神经网络 活性 抗压强度预测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TQ172.63[自动化与计算机技术—控制科学与工程]