Aplicação de redes neuro fuzzy ao processamento de polímeros na indústria automotiva

Abstract

There is an increasing tendency in the worldwide automotive market to consume polymers (plastics materials), because of their processability and low cost in high volumes. This need motivates the search for technological improvements to the material performance, even at the product development stage. The cycle time is an important data point in product design, once it allows the choice of the product design and material with the most competitive final cost, especially in high-volume production. The injection mould of automotive parts is a complex process, due to the many non-linear and multivariable phenomena occurring simultaneously. There are several commercial software applications for modeling the parameters of the injection of polymers using finite element methods; however the chosen software could be expensive or even commercially unviable. It is possible to find these parameters analytically; however, solving this problem by applying classical theories of transport phenomena (Navier-Stokes equations) requires accurate information about the injection machine, product geometry, and process parameters. Considering the above points, the Artificial Intelligence approaches like Neuro Fuzzy Networks (NFN) had shown success, consists in the synergy achieved by combining capabilities to learning and generalization from Artificial Neural Network (ANN) with the Fuzzy sets inference mechanism. The purpose of this paper is to use a Multilayer Perceptrons Artificial Neural Network (MLP-ANN) and a Radial Basis Function Artificial Neural Network (RBFANN) combined with Fuzzy Sets to produce a Mamdani inference mechanism in order to predict the injection mould cycle time. There was used a datasets obtained from Automotive Industry plant try outs. Obtained results were satisfactory, confirming NFNs as a good alternative to solving such problems.