Research on a concrete compressive strength prediction method based on the random forest and LCSSA-improved BP neural network

• Published in: Journal of Building Engineering Vol. 76; p. 107150
• Main Authors: Wang, Kewei, Ren, Jie, Yan, Jianwen, Wu, Xiangnan, Dang, Faning
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.10.2023
• Subjects: HPC compressive strength prediction; Logistic chaotic mapping; Random forest; Sparrow search algorithm; Random forest; Sparrow search algorithm; Logistic chaotic mapping; HPC compressive strength prediction
• ISSN: 2352-7102; 2352-7102
• DOI: 10.1016/j.jobe.2023.107150

The compressive strength of high-performance concrete (HPC) determines the safety of the structural engineering in modern construction projects. The compressive strength of high-performance concrete (HPC) is a highly nonlinear function of its components. To better predict the compressive strength of HPC, the initial population of the Sparrow Search Algorithm was improved based on Logistic Chaos Mapping and Nonlinear Decreasing Inertia Weight Method, and the initial weights and thresholds of the BP neural network are optimized using this algorithm to establish the RF-LCSSA-BP model for predicting the mechanical properties of HPC. Finally, the RF-LCSSA-BP model, classic algorithms, and improved SSA were used to predict the compressive strength of HPC under the influence of six factors, and the prediction results are compared and analyzed. The results show that the RF-LCSSA-BP model can predict the compressive strength of HPC better and has more advantages than the traditional methods in terms of goodness of fit and prediction accuracy. Its training error and prediction error are less than 5%, and its R2 value is close to 1. It can significantly reduce the test requirements and time costs and has important engineering significance for predicting the strength of concrete and concrete mix design. •A reliable model for the prediction of the compressive strength of concretes was proposed.•Artificial Neural Network (ANN) and the Sparrow Search Algorithm (SSA) techniques were used.•Hybridization of both ANN and the logistic chaos mapping improved sparrow search algorithm (LCSSA) improved the performance of the models.•Feature selection using Random Forest.