Lamination Parameter-Based Two-Dimension Sampling Optimization Method for Stacking Sequence Design of Composite Laminates

Abstract

A novel two-dimension sampling optimization method (2DSO) is proposed for the stacking sequence design of composite laminates. The concept of using lamination parameters (LPs) to characterize distances of different stacking sequences is introduced. The first step involves using the boundary vertices and boundary lines of the LPs to yield randomly and uniformly distributed continuous variables in the LP design space. In the second step, stacking sequences are generated randomly and evaluated with a dynamic distance constraint to ensure their LPs are close to previous continuous variables. Following this, objective values are computed for these feasible points; and several best point candidates are identified. Next, the local sampling optimization is performed by generating new stacking sequences in the vicinity of the aforementioned best point candidates. The preceding two steps are repeated until the best objective converges. Lastly, a sequential permutation search (SPS) serves as a local optimization solver to target the global optimum. Because design optimization is always conducted in the stacking sequence design domain and the LP is employed to measure the distances of distinct stacking sequences and ensure uniform distribution of initial points, no feasible constraint is needed for LP; therefore, the optimal vortices are captured efficiently. The 2DSO method is used to optimize the fundamental frequency and buckling load of composite plates and shells, and the results are compared to a layerwise optimization approach (LOA), the SPS, and a genetic algorithm (GA). The results demonstrate that 2DSO outperforms LOA and SPS, and it has similar robustness to GA while being significantly more efficient.