Aplicação da rede neural MLP (Multilayer Perceptron) em indústria de pisos e revestimentos do pólo cerâmico de Santa Gertrudes - SP

Abstract

Neural Network MLP (Multilayer Perceptron) application in the flooring and ceramic tile industries in the Ceramic Pole of Santa Gertrudes-SP - Artificial Neural Networks constitute an alternative to traditional programmed computation and have been applied in almost all branches of human knowledge. However, they are rarely applied in Geotechnology, so this work aims to show that they can be applied in the flooring and ceramic tile industries in the Ceramic Pole of Santa Gertrudes, PCSG, São Paulo State. For this purpose, proof specimens elaborated, tested and analyzed in the industries of Triunfo Cerâmica and Rochaforte Cerâmica were used. These proof specimens were composed of well differentiated clays from nine mines in the PCSG (Limeira, Cordeirópolis, Rio Claro, Ipeúna, Piracicaba, Araras and Santa Gertrudes), and these mines are representative of all the stratigraphic column of the Corumbataí Formation. For analysis, the proof specimens were subjected to four firing temperatures 1000°C, 1020°C, 1040°C and 1060°C. Each one of these values gave rise to a neural network MLP (Multilayer Perceptron) of three tiers for which the Backpropagation rule of learning was used.