Improved BP Neural Network to Predict the Comprehensive Performance of Concrete

Abstract

In order to improve the feasibility and accuracy of the concrete comprehensive performance prediction model, the factor analysis method and the BP neural network method are combined to propose an improved BP neural network concrete comprehensive performance prediction model. Example data was selected to test the improved BP neural network concrete comprehensive performance prediction model. The test results: the relative average errors of the 28d strength and slump expansion of the 14 groups of training sample predicted values and actual values were 3.398% and 1.712%, respectively. The average relative errors of 28d strength and slump expansion of each predicted sample are 4.33%, 3.80%, 5.77%, 2.07%, 1.02% and 3.55%, 0.94%, 1.48%, 0.59%, 2.00%, all of which are less than 10%, Which proves that the improved BP neural network prediction model has better prediction accuracy.