Optimization of tensile strain on injection molded polyamid-6 parts by neural networks and nonlinear programming techniques

Abstract

 The objective of this research is optimizate tensile stress of 
injection molded parts of polyamide-6 to establish process 
conditions that maximize tensile strength of parts in a real 
industrial process. The methodology consisted in development 
of essays based on I-optimal experimental design in order to 
get a data base. Four parameters were considered as inputs: 
injection holding pressure, injection time holding, % wt 
virgin material and % wt recycled material. Measurement of 
máximum tensile stress in parts was made according to ISO 
527-1 standard. 
Three models were developed by the techniques Response 
Surface Metodology, Back Propagation Neural Network and 
Generalized Regression Neural Network to predict parts 
máximum tensile stress. Finally, the best model (lowest 
forecasting error) was optimized by Trust Region Method Based 
on Interior Point Techniques for Nonlinear Programming to 
maximize tensile strength. 
This proposed methodology is capable for modeling the process 
with low error and for stablish process conditions to obtain the 
maximum tensile stress on molded parts