Strength Prediction of Agro Waste Mixed Composites Using a Neural Network Regression Model

Mixing of agricultural wastes with alloys to produce composite materials is a well-known waste management practice, and consequently, the compressive strength of the mixed composite materials changes accordingly. In such a scenario, an artificial neural network (ANN) model is presented to predict the compressive strength of composite materials. This specific ANN model is the Neural Network Regression Model (NNRM). In NNRM, the goal is to predict the compressive strength based on the constituent inputs. In the proposed NNRM, the number of hidden layers and their sizes are optimally determined using the Genetic Algorithm (GA), which uses a non-gradient approach for optimization. Optimal determination of hidden sizes avoids the tedious trial and error process of selecting the hidden layer parameters for the best performance. In this work, GA in combination with NNRM finds the optimal number of hidden layers as six with neuron sizes as [37, 38, 35, 28, 30, 34], without any trial and error approach. The optimal training and test mean square errors are found to be 8.51 E−07 and 0.03. Additionally, the proposed work can decide the best combination of %BLA and %BA ratios required for minimum prediction error.