This document presents a method to estimate the weights of main materials (copper, iron, oil) for transformers using an artificial neural network (ANN) with the Levenberg-Marquardt backpropagation algorithm. Training data consisting of 24 input/output pairs obtained from a transformer manufacturing company are used to train the ANN, with inputs like short circuit impedance, installation height, and temperature and outputs of material weights. The trained ANN is then able to accurately estimate the material weights of transformers, providing a method to forecast transformer costs for manufacturers.