Model-based design optimization of process parameters for composite manufacturing processes

The optimal design of the process inputs for composite manufacturing processes, using models and simulations, is critical for the design and manufacture of increasingly complex composite structures in the infrastructure, defense and automotive sectors. The application of model-based design optimization for composite manufacturing is investigated in this dissertation. The design optimization methodology is formulated and its' efficiency is explored in three important composite manufacturing processes. The first example problem examined is of optimal spacing of injectors in VARTM process to minimize mold filling time and associated costs. An analytical model for the resin flow is formulated and then used to determine the optimal inlet spacing as a function of the material and geometric process parameters. The second design optimization problem involves design of resistive susceptors for minimum temperature variation in field repair of composite structures by induction bonding. Stochastic methods such as genetic algorithms (GAs) are investigated to minimize non-uniform heating by selecting for the best possible cut mesh patterns. It is shown that the GA was able to reduce the variations in heat generation in the mesh for all cases and delivered significant improvements over the baseline case in reasonable computational time, evaluating less than 2% of the possible cut mesh patterns. The third design challenge is of optimal gate and vent location in the resin transfer molding process. A preliminary case study revealed that GAs were capable of addressing this problem and resulted in near optimal results. To improve their performance, a sensitivity gradient-based fill time minimization algorithm was derived for optimal gate location in the resin transfer molding process using the process physics and validated using three complex mold geometries. By formulating real-coded genetic algorithms and coupling them with the gradient-based gate location algorithm, two hybrid optimization algorithms were developed: (a) a serial hybrid optimization algorithm and (b) an interactive “memetic” optimization algorithm. A number of studies are performed for three representative complex mold geometries using the pure genetic algorithm, the gradient-based optimization algorithm and the two hybrid optimization algorithms using the mold fill time and its gradient with respect to gate location coordinates as cost criteria. The results are benchmarked against known best solutions in terms of quality of final solutions and the computational effort required. The results indicate that the memetic algorithm is the most consistent hybrid optimization algorithm and gives gate locations with the least mold fill times for all mold geometries.