Use of neural networks to predict the morphology and mechanical properties of injection molded PPS.
Cybele Lotti, Rosario E.S. Bretas
Department of Materials Engineering - Federal University of Sao Carlos -UFSCar
Brazil

Keywords: neural networks, injection molding, mechanical properties

---

To predict the properties of an injection molded part, before processing, is the goal of many researchers and engineers. However, this is not an easy task, because the injection molding process has many variables that are related in a very complex way. Besides that, the mechanical behavior of the injection molded parts depends also of the intrinsic properties of the polymer and of the particular supra-structural morphology that results of a particular set of processing conditions. Therefore it is necessary to understand how these processing conditions influence the morphology, and how this processing dependent morphology determines the mechanical properties.
Thus, the objective of this work was to use artificial neural networks, as an alternative to empirical and computational methods, to predict the morphology and the mechanical properties of an injection molded polyphenylene sulphide, PPS, from the processing conditions.
First, an artificial neural network was built to predict morphological aspects of the PPS part, like crystallinity degree, thickness of the flow induced crystalline layer and the average spherulite size, from the mold temperature, flow rate and holding pressure. A second neural network was later built to predict the elastic modulus and the tensile and impact strengths from the morphology. Finally, another neural network was built to predict the mechanical properties from the processing conditions.