Optimization of composite parts placement in autoclave

In the production of composite parts, usually multiple parts are loaded into the autoclave in one batch. A typical autoclave curing cycle has three main phases and proceeds as follows: all parts are heated until they reach the curing temperature during the heat-up phase, they are then held at this temperature for a certain period of time during the dwell period, and then they are cooled in the cooling phase. The main goal in a curing cycle is the complete curing of all parts. In an ideal curing cycle, all parts reach the curing temperature at the same time, avoiding any over-curing. However, due to some factors such as part positions, geometries, total mass in the load, it is often not possible to realize this requirement without any delays between parts. The parts that reach the curing temperature earlier than the others are over-cured as much as it takes for the last part (the lagging part) to reach the curing temperature. This time delay can be monitored via thermocouples on all parts during the curing cycle and it affects the product quality to a great extent. If these delays are minimized, higher quality products can be produced. Another consideration is minimization of the total process duration by shortening the heat-up phase. This way, time and energy savings can be obtained. In this work, we develop a two-stage approach to minimize both the time delay between parts and the duration of the heat-up phase. In the first stage, we determine the factors affecting the time to reach the curing temperature. We divide the autoclave charge floor to 18 regions, evaluate different parameters and their interactions that are assumed to affect the curing process. Then, regression models that relate the curing duration of each area with those parameters are developed. In the second stage, we determine the placement of the products in the autoclave using the regression models of the first stage. We develop a multi-objective nonlinear mixed integer programming model that minimizes the two objectives mentioned above. We linearize this model using additional variables and use ε-constraint method to generate the nondominated frontier. To obtain solutions in shorter durations, we employed one of the well-known multi-objective evolutionary algorithms, Nondominated Sorting Genetic Algorithm-II (NSGA-II). The validity of the practical use of the model is tested on real cases in a composite factory in Turkey.