Prediction of masonry prism strength using machine learning technique: Effect of dimension and strength parameters

Abstract

The compressive strength of masonry must be determined to design masonry constructions. Although the compressive strength of masonry mostly depends on the compressive strength of the block/brick and mortar, there are several factors, such as joint mortar thickness, prism height, and masonry unit aspect ratio, that also impact the strength of masonry. To predict the compressive strength of masonry prisms, the present study proposes different models of machine learning techniques, including linear regression, decision tree, ridge regression, random forest regression, artificial neural network, and XG Boost. Based on 540 experimental datasets collected from published literature, the performance of the models was compared and the best model was chosen for the forecast of compressive strength of masonry prism. Among the several machine learning models assessed in this work, the ANN model performed best for the prediction of the compressive strength of masonry prisms (R2 0.95, RMSE 1.83 MPa). The sensitivity analysis of the ANN model shows that the compressive strength of the masonry prism is dominantly influenced by the strength of the masonry unit, followed by the thickness of the mortar, and least by the height-to-width ratio of the masonry prism. Therefore, the present study offers a methodical assessment of the compressive strength prediction of masonry prism, which may add to the knowledge base and practical application of this field.