Abstract
The present study focuses on proposing various statistical models, such as linear regression (LR), nonlinear regression (NLR), and artificial neural network (ANN), to forecast the compressive strength of environmentally friendly high-strength concrete, incorporating waste agricultural material like palm oil fuel ash (POFA). A dataset of 105 experimental observations was compiled from existing literature to achieve this goal, which was subsequently partitioned into training and testing subsets. Each model was developed based on the training data and evaluated using the testing data. The performance of each proposed model was gauged using diverse statistical metrics like the coefficient of determination, mean absolute error, root mean square error, and scatter index to identify the most effective model. The findings indicate that using POFA with a finer particle size exerts a greater influence on the concrete's properties. The replacement was done using the weight method, and the predicted equation worked with the variation of the used rate of POFA from 0 to 60% of total binder weight. Substituting a portion of cement with POFA leads to a reduction in the heat of hydration and an extension of the setting time. The optimal percentage of POFA is 30%, yielding mechanical properties superior to those of the control mixture, particularly in the later stages of development. Among the models considered, the ANN demonstrates superior efficiency and accuracy in predicting the compressive strength of conventional concrete modified with POFA compared to LR and NLR models. This is evident in the ANN's higher R2 values of 52% and 16%, respectively, and a lower scatter index below 0.1%.