Optimizing concrete block properties through the use of coconut shell and coconut shell ash: a multilayer perceptron approach

This research utilized a multilayer perceptron model based on an artificial neural network (ANN) to improve concrete blocks by partially replacing stone dust and cement with coconut shell and coconut shell ash. From 35 experimental data points, 68.6 % were used for training and 31.4 % for testing. Chemical analyses showed that coconut shell ash, rich in silicon and other oxides, enhanced pozzolanic reactions as a cement substitute. Results indicated that higher proportions of coconut shell ash and shells reduced density, making the mixture lighter but potentially affecting structural integrity. However, increased replacement levels improved workability, shown by higher slump values. The curing period significantly impacted strength, with longer times leading to increased strength due to continued hydration. An optimal compressive strength of 14.17 N/mm 2 was achieved with a 5 % replacement after 70 days. The ANN model demonstrated a sum square error of 0.198 and a high correlation coefficient of 0.983 when predicting strength. Optimal conditions for achieving a compressive strength of 33.4 N/mm 2 were identified, underscoring the cost-effectiveness of using coconut byproducts instead of traditional materials.