Prediction of tensile strength of basalt continuous fiber from chemical composition using machine learning models

Abstract Tensile strength is the major mechanical property of basalt continuous fiber and is closely related to their chemical composition. This study constructed a machine learning model framework to predict the tensile strength of basalt continuous fiber. The database included the characteristic variables of oxides and their derived parameters, as well as the target variables of tensile strength. The mean squared error (MSE) were calculated to evaluate the performance of six machine learning models of decision tree, kernel ridge regression, multivariable linear regression, support vector regression, random forest, and k‐nearest neighbors. The k‐nearest neighbors model had a minimum MSE and was tuned using the grid search method with cross‐validation. The optimal hyper‐parameters of the k‐nearest neighbors model were K = 6, p = 10, and the determination coefficient of the model reached the maximum of 0.7110 on the test data. The limitations and implications of the present study were also demonstrated. We finally expected that the constructed framework could achieve higher prediction accuracy for tensile strength at a low cost in the future. Highlights A machine learning framework was built to predict the tensile strength of BCF. Oxides and their derived parameters were used for modeling. K‐nearest neighbors performed best in predicting the tensile strength. The framework was robust, cost‐saving, and efficient.