Artificial neural network and associated computer tool for rapid shear strength prediction of two-span continuous deep beams

Continuous deep beams exhibit complex shear behavior due to non-linear strain distribution and coexistence of high shear and high moment within the same regions. This behavior makes it challenging to predict their shear strength accurately. The objective of this study is to develop a pioneering artificial neural network and associated computer tool to accurately and rapidly predict the shear strength of two-span continuous deep beams. Literature lacks machine learning algorithms specifically developed for continuous deep beams. To achieve this objective, a comprehensive database of 150 experimental specimens is compiled from eight previously published research studies. The 140 data points, which passed the exploratory data analysis, are used to train, validate, and test the proposed network for applicability to a wide range of input parameters. The accuracy and reliability of the proposed network are evaluated using a comprehensive set of evaluation metrics including the correlation coefficient (R), determination coefficient (R 2 ), mean absolute error (MAE), mean squared error (MSE), root mean squared error (RMSE), and the coefficient of variation (CV). The importance of the input parameters on the predicted shear strength is evaluated using the SHapley Additive exPlanations analysis. Another important objective is to develop an open-access computer tool to execute network formulations and rapidly calculate the shear strengths of continuous deep beams. The computer tool aims to bridge the gap between research and practice and enable the use of the developed network for continuous beam when there are no experimental testing results available.