Prediction of constitutive model for basalt fiber reinforced concrete based on PSO-KNN

Abstract

In order to find a simple method to study the effect of basalt fibers on the mechanical properties of concrete when incorporated into concrete, machine learning is introduced in this work on an experimental basis. The basalt fiber-reinforced concrete (BFRC) specimens were fabricated through independent processing, and the compression tests under different stress states were conducted on the BFRC specimens with different fiber compositions using the MTS816 rock testing system. After obtaining the experimental dataset with the four influencing factors of fiber volume fraction, fiber length, circumferential pressure and strain as input variables and stress as output variable, the BFRC prediction model was established based on extreme gradient boosting, support vector machine, K-nearest neighbor, and Particle Swarm Optimization K-Nearest Neighbor (PSO-KNN) algorithms; Then the predicted fitting results of the training set and test set are analyzed according to the relevant evaluation indexes, and the data indexes indicate that the PSO-KNN model has the best prediction performance, and the PSO-KNN model is used to predict the stress-strain fitting curves of BFRC, and finally the parameter contribution is analyzed based on the information of the curves. This is the first time that PSO-KNN is used in the study of BFRC eigenmodel, and the prediction effect is good, which not only overcomes the drawbacks of time-consuming and expensive experimental research, but also provides a basis and reference for engineering applications and later scholars' research on BFRC eigenmodel.

Keywords

• Basalt fiber concrete
• Particle swarm
• K-nearest neighbor
• Constitutive model