Classification of shape aggregate using MLP network based training algorithm

Historically, mechanical sifting and hand grading havebeen used to assess the quality of aggregates. It must pass a number of tedious, slow, subjective, time-consuming mechanical, chemical, and physical testing in order to produce better aggragates. This study's objective is to develop a classification scheme that can categorise aggregates based on pictures of aggregate morphologies. Using an artificial neural network, the image was reprocessed for shape categorization after it had been taken. The Levenberg Marquardt (LM) training algorithm technique outperforms Backpropagation (BP) training algorithms in terms of performance, mean square error (MSE), and regression. The LM training approach that uses Multilayer Perceptron (MLP) networks has the lowest mean square error (MSE) and maximum regression. With 1.6923 on MSE and 0.9572 on regression, the LM-trained network can successfuly classify.