铺面水泥混凝土抗冻标号预估模型

Prediction model of frost-resistance grade for pavement cement concrete

在充分利用国内外已有的混凝土抗冻试验数据基础上,用人工神经网络与多元非线性回归两种建模方法相结合的途径,建立了能够反映铺面混凝土配合比参数与抗冻标号之间综合非线性关系的铺面水泥混凝土抗冻标号预估模型。通过BP神经网络虚拟试验分析,得到水灰比对混凝土抗冻标号的影响最大,含气量其次,水泥用量对抗冻标号的影响最小,含气量与混凝土抗冻标号呈幂级数关系,水灰比、水泥用量与抗冻标号之间均呈指数函数关系。在这些基本定性关系分析基础上,用多元非线性回归方法得到的铺面混凝土抗冻标号预估模型。该模型具有较好的预估精度,能够反映各配合比参数对铺面混凝土抗冻标号影响的内在规律。

Combining BP neural network with multi-nonlinear regression method, a frost-resistance grade prediction model, which depends on the main mix parameters of air-entrained concrete, was developed. Through BP neural network virtual tests, cement water ratio has a deepest influence on the concrete frost durability with powered relation, and air content exponentially does it less, then cement content performs least in the same form as cement water ratio. Based on these basic relational analysis, the multi-variables nonlinear model shows good prediction precision and reflects the influences of main mix parameters on concrete frost durability rationally.