Optimization of adhesively bonded joint under out of plane impact loading

Abstract

This thesis investigates the optimization of adhesively bonded metal and composite plates subjected to out of plane impact loading, with the aim of enhancing structural performance and reducing weight. Adhesively bonded joints are increasingly preferred in aerospace applications due to their advantageous mechanical performance and ease of manufacturing compared to conventional fasteners. However, they remain sensitive to impact loads, which can reduce the structural integrity of the assembly. The study is divided into three main parts. In the first part, the thickness of an aluminum alloy plate, AL7050, is optimized under impact loading. Finite element simulations are created using LS-DYNA to generate training data. Kriging surrogate models are built to estimate two objective functions: von Mises stress and total mass. The optimization process is carried out using the multiobjective genetic algorithm, gamultiob, in MATLAB, resulting in a Pareto front of optimal thickness values. This approach enables an effective trade-off between minimizing stress and minimizing mass while maintaining a specified safety factor. The second part of the study focuses on optimizing a woven composite plate, AS4 8552, subjected to impact loading. The primary goal is to maximize absorbed energy while minimizing mass. The stacking sequence and total number of plies are treated as design variables, selected from discrete values, the orientation angle is selected as (0o and 45o) and number of plies vary 8 from 16 for symmetric laminate. LS-DYNA simulations are used to impact model inputs and outputs, and Python scripts automate the generation of stacking sequence combinations. The Hashin failure index is used as a constraint to eliminate fail configurations. Metamodels are created using the Gaussian Process Regression (GPR) method to predict absorbed energy and mass for different laminate designs. These models help estimate the results without running a full simulation each time. Then, a genetic algorithm is used to find the best designs by trying many different combinations. The goal is to find laminate configurations that can absorb a high amount of impact energy while keeping the weight as low as possible. In addition, only the designs that meet the required Hashin failure index limit is accepted. In the third section, the optimization of the adhesively bonded joint is performed. Input and output data are obtained from LS-DYNA simulations. The design parameters include the aluminum plate thickness, the stacking sequence, and the number of plies. The objective functions are the total mass and the maximum von Mises stress on aluminum plate. The constraints are defined by the Hashin failure index for the composite plate. This thesis presents an efficient design approach for adhesively bonded joints by combining finite element analysis with LS-DYNA, metamodels and multiobjective optimization with genetic algorithms. The proposed method helps reduce simulation time while maintaining safety and performance, supporting the use of composite and metal bonding structures in aerospace and automotive applications.