Mathematic algorithm used in Stacking Sequence Optimization (Shuffling Phase)

Hello, good afternoon. I am conducting a three-phase optimization for composite material structures using Altair:
.According to OptiStruct, the default optimization methods for free-size and size parameters are the DUAL2 optimization method (based on the CONLIN algorithm) and the MFD method, respectively. However, despite my research, I haven't found which mathematical optimization method is used for the Shuffling phase (Stacking Sequence Optimization). I would venture to say that OptiStruct executes a branch and bound approach based on the optimization data.
More specifically, in the third phase, I am performing a MINMAX optimization, using Tsai-Hill both as the objective function to minimize and as a constraint function (not exceeding a Tsai-Hill value of 1).
What do you think? Is Branch & Bound the default method in the shuffling phase?
Answers
-
honestly i'm not sure what method exactly is used there. I see it as focused on shuffling the layers, while achieving the manufacturing constraints, such as max number of consequent plies and so on.
Did you achieve a good result using the same response as Objective and constraint?
0 -
Hello Adriano,
In fact, I have tried doing the shuffling without any constraints, and it always performs two or three iterations at most, without lowering the Tsai-Hill. However, using a MINMAX optimization, reducing the Tsai-Hill through shuffling, and introducing it as a constraint (for example, setting 0.9 as the maximum), it converges to very good laminate results. It is true that the optimization history shows many oscillation peaks, but in the end, my final laminate remains with a Tsai-Hill below 0.9. I’m attaching a picture of the history to make it easier to see.Maybe these peaks are related to the fact that the mathematical method behind the shuffling is backtracking (where not all laminate sequence combinations are explored, only the admissible ones), and the peaks could be due to the exploration of critical situations before evaluating the optimal sequence. I’m attaching a scientific article here that discusses a method where I find some similarities with shuffling. Do you think this could be the algorithm used by OptiStruct?
0