Image Compression Using Different Optimization Algorithms: A Review Artificial Neural Network Modeling of the Compressive Strength of Concrete with Polyethylene Terephthalate (PET) Waste as Fine Aggregate Replacement

Abstract

This study modelled the compressive strength of concrete with polyethylene terephthalate (PET) waste as fine aggregate replacement. Artificial neural network (ANN) was used to model and predict the compressive strength of PET concrete at various percentage replacements (2 to 50% at a step of 2% by weight), with the multilayer feedforward neural network and the radial basis function methodologies compared to see which is more accurate. The multilayer feedforward neural network modelling results showed a predictive accuracy of 95.364% with root mean square error value of 3.6621 × 10-15 while, the radial basis function neural network modeling results showed a higher predictive accuracy of 99.812% with root mean square error value of 3.7748 × 10-15. The results of this study demonstrated that computer-generated models such as the radial basis function may accurately predict the compressive strength of PET concrete, as the results of the experimental and predicted tests were similar. Additionally, it was discovered that the radial basis function method takes less time to create the model because there is no repetition required to get at the model's favorable parameters. Furthermore, radial basis function networks train more quickly than multilayer perceptrons, but classification is slower since each hidden layer node must calculate the radial basis function for the input sample vector during classification.