基于RBF神经网络的混杂纤维混凝土强度预测

Predict ion of the st rength of hybrid fiber reinforced concrete based on RBF neural network

针对混杂纤维混凝土强度受多种因素影响,强度与各影响因素之间关系为复杂的非线性问题,通过人工神经网络的自适应、自学习和非线性映射,可以找到以影响因素为输入变量、以混杂纤维混凝土强度为输出变量之间的非线性关系,在文献试验实测值的基础上采用MATLAB神经网络工具箱建立了四个三层RBF和BP神经网络模型,采用所建立的RBF和BP神经网络对混杂纤维混凝土的抗拉强度和抗折强度分别进行预测,并将各自的预测值和实测值进行了对比分析。结果表明:RBF神经网络预测值与试验实测值吻合良好,较之BP神经网络有更高的强度预测能力,该方法可行且预测精度满足工程需要,为工程上研究混杂纤维混凝土强度提供了新方法。

The strength of hybrid fiber reinforced concrete is influenced by many factors, and the relationship between them are complex nonlinear problem, but the nonlinear relationship between input variables like some of the factors and output variables like the strength of hybrid fiber reinforced concrete can be obtained by self-adapting, self-studying and nonlinear mapping of artificial neural network.Based on experimental values, four RBF and BP neural network models were established in MATLAB neural network toolbox, compressive strength and flexural strength of hybrid fiber reinforced concrete were predicted respectively by using RBF and BP neural network model. The predicted values and measured values were analyzed in comparison.The results showed that the predicted values of RBF neural network was in good agreement with the experimental values, and compared with the BP neural network had a higher strength prediction ability, the method was feasible and the prediction accuracy can meet the needs of engineering, providing a new method for the research on strength of hybrid fiber reinforced concrete in engineering field.