@Article{cmes.2003.004.587, AUTHOR = {R. Mathur, S. G. Advani, B. K. Fink}, TITLE = {A Real-Coded Hybrid Genetic Algorithm to Determine Optimal Resin Injection Locations in the Resin Transfer Molding Process}, JOURNAL = {Computer Modeling in Engineering \& Sciences}, VOLUME = {4}, YEAR = {2003}, NUMBER = {5}, PAGES = {587--602}, URL = {http://www.techscience.com/CMES/v4n5/24828}, ISSN = {1526-1506}, ABSTRACT = {Real number-coded hybrid genetic algorithms for optimal design of resin injection locations for the resin transfer molding process are evaluated in this paper. Resin transfer molding (RTM) is widely used to manufacture composite parts with material and geometric complexities, especially in automotive and aerospace sectors. The sub-optimal location of the resin injection locations (gates) can leads to the formation of resin starved regions and require long mold fill times, thus affecting the part quality and increasing manufacturing costs. There is a need for automated design algorithms and software that can determine the best gate and vent locations for a composite part by using the current simulation capabilities. In the work presented here, the gates are encoded into real number strings for the GA. A sensitivity gradient-based fill time optimization algorithm was developed using the process physics, which can be used as a local optimization algorithm. The global search capabilities of the GA and the local search capabilities of the sensitivity gradient-based fill time optimization algorithm were combined in two separate hybrid optimization algorithms: a serial hybrid optimization algorithm and an interactive optimization algorithm. In addition, the sensitivity gradient-based algorithm involves the computation of the gradient of the fill time with respect to the gate location coordinates. This gradient information was included in the criteria for optimization to increase the capabilities of the hybrid GA. Several RTM molds with geometric and material complexities were selected and discretized. A number of studies were performed using the pure genetic algorithm, the gradient-based optimization algorithm and the two hybrid optimization algorithms using the mold fill time and it's gradient with respect to gate location coordinates as the cost criteria. These studies were performed for the cases of a single gate and two gates. The results were benchmarked against known best solutions in terms of quality of final solutions and the computational effort required.}, DOI = {10.3970/cmes.2003.004.587} }