• 2025-04-25
• O.E. Osanyinlokun
• Volume 1, Issue 1

Abstract

An artificial neural network (ANN) was used to predict sawdust ash (SDA)-reinforced concrete beams’ (RCBs) bending moment (BM) and shear force (SF). With compressive strength-test results data using an ANN, a numerical maximum SF and BM-predicting concrete model for various mixtures of RC containing different percentages of SDA as a partial replacement for cement was developed and validated. The ANN used in this study was a multi-layer perceptron (MLP) with three hidden layers. The MLP was trained on the data using the back-propagation algorithm. A cross-validation technique that repeatedly splits the original dataset into training and validation sets across multiple iterations was utilized. The cross-val-score function from sci-kit-learn evaluated the model’s performance using 5-fold cross-validation. The original dataset is split into five equal-sized subsets (folds). The model is then trained on 4 of these folds and evaluated on the remaining fold. This process is repeated five times, with each fold used as the validation set once. The final cross-validation score is an average of the scores from each fold. The ANN could predict the BM and SF of the RCBs with an average error of less than 5% and identify the most critical factors that influence their properties.

Download PDF