Development of novel design strength model for sustainable concrete columns: A new machine learning-based approach

Abstract

Billions of tons of construction and demolition (C&D) waste generation is causing global environmental crises. The application of C&D waste in concrete columns is a sustainable avenue but hindered due to a lack of comprehensive design guidelines. Currently, no work is available in the open literature regarding the machine-learning-based comprehensive design strength models of spiral steel confined natural aggregate concrete (SSCNAC) and recycled aggregate concrete (SSCRAC) columns. This study comprehensively evaluates ten machine-learning techniques, including random forest, gradient boost, Ada boost, k-nearest neighbor, bagging regressor, support vector, XG boost, decision tree, artificial neural network, and gene expression programming for design strength modeling of SSCNAC and SSCRAC columns. A test database comprising 290 experiment results is developed and used for modeling. The accuracy of models is improved using hyper-parameter tuning and cross-validation functions. The performance of machine-learning techniques is evaluated through various statistical parameters. Results indicate that the spiral steel strength and the unconfined concrete compressive strength are the most critical inputs to predict the design strength of SSCNAC and SSCRAC columns. All the machine-learning models can accurately predict the design strength of SSCNAC and SSCRAC better than the existing models and can be practical tools for the accurate design strength modeling of the SSCNAC and SSCRAC. A one of its kind of a comprehensive machine learning-based design equation is proposed in this study, which can be used to predict the design strength of SSCNAC and SSCRAC accurately, leading towards the sustainable design of eco-friendly concrete columns.