基于改进的径向基函数神经网络水泥强度预测方法

The Cement Strength Prediction Method Based on Improved Radial Basis Function Neural Network

通过将遗传算法应用于径向基函数神经网络参数设计中,提出一种基于遗传算法优化的径向基函数神经网络水泥强度值预测模型,实现径向基函数神经网络隐层节点函数的中心矢量、基宽向量和隐层与输出层之间权值的优化设计.以经归一化处理后的输入样本数据为模型输入,以水泥28 d强度值为模型输出,建立经遗传算法优化后的径向基函数神经网络预测模型.仿真结果表明,优化后的径向基函数神经网络能达到较高的预测精度,可用于水泥强度的预测.

A prediction model of cement strength based on genetic optimized radical basis function neural net- work was proposed through the application of genetic algorithm to the parameters design of RBF（radial basis function） neural network, which realized the optimized design of RBF neural network hidden layer nodes function center vector, widths and weights between the hidden layer and output layer. The prediction model based on genetic optimization RBF neural network was designed with the normalized sample data as input and 28 days strength of cement as output. The simulation results show that the optimized RBF neural net- work has the advantage of high forecast accuracy and can be used to predict cement strength.