ABSTRACT
ABSTRACT The genetically optimized neural network system (GONNS) has been introduced for automated modeling and optimization of systems when only experimental data or observations are available. GONNS represents the system with multiple artificial neural networks after a training process. The optimum solution is calculated by using one or more genetic algorithms operating simultaneously. The GONNS does not require the user to provide any analytical or empirical models. GONNS was implemented for optimization of three systems in this paper. First, the estimation accuracy of the GONNS was evaluated by using the data of an analytical expression with three inputs. Second, the feasibility of the GONNS was studied for control of the impact resistance of composite materials by applying electricity. Third, selection of optimal operating conditions and material by using the Composite Material Selection Advisor package was demonstrated. Composite Material Selection Advisor uses the GONNS for modeling and optimization. It has additional programs for estimation of the complexity index of parts from the STL files of the commercial computer-aided design/ computer-aided manufacturing programs. If necessary, the complexity is corrected by removing the vertical holes that will be drilled after the manufacturing operation. All three implementations demonstrated the feasibility of the GONNS.
