神经网络在无机保温砂浆收缩预测中的应用

Application of Neural Networks on Shrinkage Prediction of Inorganic Thermal Insulation Mortar

以水灰比、相对湿度、水泥、粉煤灰以及玻化微珠含量为输入变量,基于径向基概率神经网络对无机保温砂浆的收缩进行预测。与多项式回归模型相比,RBPNN模型的预测精度、平衡性以及泛化性都显著优于前者。此外,通过反演的方法该模型还可以用于无机保温砂浆配比的优化。

In this work,shrinkage of inorganic thermal insulation mortar were predicted by RBPNN with five input variables covering water-cement ratio,relative humidity,content of cement,fly ash and aggregate.The simulation results showed that the RBPNN model exhibited promising precision,equilibrium and generalization ability for predicting shrinkage of mortar compared with polynomial regression model.Furthermore,the RBPNN model was employed for optimizing mixtures of mortar with satisfactory shrinkage by means of refutations.