Shape of aggregates classification by using MLP network based training algorithm and activation function

Mechanical sifting and hand grading have long been used to assess the quality of aggregates. It must pass a range of mechanical, chemical, and physical testing to produce superior aggregates; these tests are typically performed manually and are sluggish, arbitrary, and time-consuming. This work aims to develop an image-based classification system that can categorise aggregates. An artificial neural network was used to reprocess the image after it had been taken in order to classify its shapes. In comparison to Backpropagation (BP) training techniques, the Bayesian Regularization (BR) methodology offers better performance with reduced mean square error (MSE) and higher regression. The LM training approach using the Multilayer Perceptron (MLP) network-based MSE offers the maximum regression and the lowest mean square error (MSE). The BR-trained network has 1.4235 MSE and 0.9760 regression capabilities.