Aplicação de redes neurais para a determinação do coeficiente de atrito de filmes plásticos flexíveis

Abstract

Flexible plastic packaging is gradually being used, since it is the most versatile packaging, with various applications, and of great durability. In order to obtain these packages, a whole production strategy is required, which involves not only its production process but also the product quality control. Aiming the parameters prediction, the present work aims to determine the friction coefficient (COF) in flexible plastic films by using Perceptron neural networks. According to data obtained in experimental tests, the neural networks in both software (i.e., Python and Matlab) obtained satisfactory results for determining the COF of the films with an R2 = 95% and a mean error of about 0.001, therefore, an application was created for both neural networks so that material engineers could interact as users with these networks. Neural networks appear as a possible aid for quality control tests, aiming at improving the speed of achieving and analyzing, keeping accuracy of responses.