基于BP神经网络算法的高模量混合料HMM-13级配优化设计研究

Research on the Optimization Design of HMM-13 Grading forHigh Modulus Mixtures Based on BP Neural Network Algorithm

空隙率是高模量混合料HMM-13配合比设计中重要指标。通过成型不同配合比条件下的高模量混合料旋转压实试件,测试高模量混合料的空隙率,以关键筛孔尺寸为13.2 mm(K_(1))、4.75 mm(K_(2))、2.36 mm(K_(3))、0.075 mm(K_(4))的通过率及综合油石比(O/S)作为输入参数,空隙率(V)作为输出参数,采用双隐含层BP神经网络算法,建立了高模量混合料HMM-13体积指标预测模型,进行了关键筛孔通过率对旋转压实试件空隙率的敏感性研究。研究结果表明,建立的模型误差小于1%,泛化能力强,可用于高模量混合料体积指标的预测;当90%<K_(1)<93%,45%<K_(2)<50%,25%<K_(3)<34%,6.6%<K_(4)<6.9%,O/S≥5.2%时,可满足混合料空隙率1.5%~2.5%要求。

Voids Content is an important indicator in the mix design of high modulus mixture HMM-13.This article tests the porosity of high modulus mixtures by forming rotating compaction specimens under different mix ratio conditions.The key mesh sizes of 13.2 mm(K_(1)),4.75 mm(K_(2)),2.36 mm(K_(3)),0.075 mm(K_(4))and the comprehensive oil stone ratio(O/S)are used as input parameters,and the voids(V)is used as output parameter.The double hidden layer BP neural network algorithm is used.A volume index prediction model for high modulus mixture HMM-13 was established,and the sensitivity of key sieve pass rate to the porosity of rotary compacted specimens was studied.The research results indicate that the established model has an error of less than 1%and strong generalization ability,which can be used for predicting the volume index of high modulus mixtures;When 90%<K_(1)<93%,45%<K_(2)<50%,25%<K_(3)<34%,6.6%<K_(4)<6.9%,and O/S≥5.2%,the requirement of mixture porosity of 1.5%~2.5%can be met.