Neural Network Prediction of Compressive Strength of Lightweight Coconut Shell Concrete

Abstract

This work is a part of research investigation into the effective use of crushed coconut shells in the production of lightweight concrete. The natural coarse aggregate of this concrete was fully replaced with coconut shell aggregate and partial cement replacement with 0%, 5%, 10% and 15% of fly ash and silica fume. Based on trial mix method a suitable mix proportion was arrived. Long term compressive strength up to 365 days was studied. The optimum compressive strength was obtained for 10% silica fume mix. 28 days compressive strength of coconut shell concrete with partial replacement of silica fume and fly ash satisfies the minimum requirement of structural lightweight concrete. Hence it can be useful for structural purposes. For the prediction of compressive strength an Artificial Neural Network Model was developed using MATLAB and it includes seven inputs and one output. One hundred and sixty eight experimental data were used for developing the multilayer feed forward and back propagation neural network model. The mean square error for the predicted values with respect to the experimental value is 0.0348. The predicted compressive strength was compared with the experimental strength and found remarkably close to each other.