Digital Twin Data-Driven Multi-Disciplinary and Multi-Objective Optimization Framework for Automatic Design of Negative Stiffness Honeycomb

The honeycomb structures perform well in shock and vibration isolation with a light weight and small material. They cannot reuse due to plastic buckling after a deformation. To overcome this disadvantage, many researchers have actively studied the development of a reusable honeycomb structure using the elastic buckling mechanism. They proposed a new shape for honeycombs: one that can absorb and dissipate energy through a negative stiffness (NS) effect caused by the elastic buckling phenomenon. However, existing studies have focused on the concept design of the NS curved beam, and the research on how to design products in detail is insufficient. Therefore, this study proposes a digital twin data-driven design framework for the detailed design of the semi-symmetric NS curved beam. For this, multi-disciplinary and multi-objective optimization of the virtual NS beam model is performed by maximizing the energy absorption in the elastic buckling discipline through geometric design variables and analytical parameters, and enhancing its similarity to accurate theoretical models in the digital twin discipline through coupling variables. For the optimization, the specific energy absorption (SEA) and error rate are modeled as regression models, and convergence is modeled as a classification model using an artificial intelligence (AI) model. The optimum results improved the energy absorption and reduced the relative error rates while ensuring convergence of the virtual model.