Data-driven inverse method with uncertainties for path parameters of variable stiffness composite laminates

The uncertainties quantification of fiber path parameters is essential for fully developing the variable stiffness composite laminates (VSCL) performance. How to design a reliable fiber placing scheme effectively with a given mechanical requirement is an important issue to be considered. To achieve a feasible solution, a data-driven-based approximate Bayesian computation (ABC) inverse method is suggested. The novelty of this paper lies in three aspects. We employ the auto-encoder to extract the feature vector of the high-dimensional physical field as the form of a summary statistical quantity of the ABC. Back propagation neural network is applied to construct the mapping between the fiber path parameters and the feature vectors to avoid the time-consuming physical field simulation. Finally, a hybrid adaptive nested sampling strategy is proposed by introducing the differential operation to accelerate the sampling process of posterior distribution of fiber path parameters. Two VSCL engineering examples are used to verify the feasibility of the suggested method. The experiment results indicate that the available fiber path parameters with uncertainties can be obtained based on the suggested inverse identification framework.