A BFRC compressive strength prediction method via kernel extreme learning machine-genetic algorithm. (15th August 2022)

• Record Type: Journal Article
• Authors: Li, Hong; Lin, Jiajian; Zhao, Dawei; Shi, Guodong; Wu, Haibo; Wei, Tianxia; Li, Dailin; Zhang, Junliang
• Is Part Of: Construction & building materials. Volume 344(2022)
• Publication Date: 2022-08-15
• Subjects: Basalt fiber -- Compressive strength -- Extreme learning machine -- Genetic algorithm; Building materials -- Periodicals; 624.18
• DOI: 10.1016/j.conbuildmat.2022.128076
• Languages: English
• ISSNs: 0950-0618

Abstract

Abstract: The compressive strength (CS) of basalt fiber reinforced concrete (BFRC) is usually determined by uniaxial compression or triaxial compression of BFRC. However, the test method is not only time-consuming but also expensive. Based on certain test samples, the machine learning (ML) technique is used to predict the CS of BFRC, which can save some test costs and improve efficiency. However, the problem of improving the prediction accuracy enhancement of BFRC compressive strength is still challenging. This paper proposes a novel hybrid KELM-GA model for predicting the CS of BFRC by combining the kernel extreme learning machine (KELM) with a genetic algorithm (GA). To evaluate the performance of the KELM-GA model, four ML models, backpropagation neural network (BPNN), support vector regression (SVR), Gaussian process regression(GPR), and kernel extreme learning machine (KELM), were used to compare with the KELM-GA model. The results show that combining the KELM model with the GA model can improve the performance of the model. Compared with other models, the KELM-GA model can achieve better performance indexes, and the method has high engineering application value in predicting the CS of BFRC. Highlights: The compressive strength of basalt fiber reinforced concrete is predicted by using the genetic algorithm to optimize the algorithm of the kernel extreme learning machine. The performance of the KELM-GA algorithm is the highest among several classical machine algorithms (such as BP, SVM, GPR and KELM), and the experimental analysis further verifies the robustness and effectiveness of the algorithm.