Multiscale modelling of a novel wingbox structure with increased bend-twist coupling
Topic(s) : Material and Structural Behavior - Simulation & Testing
Co-authors :
Mario MIRANDA (UNITED KINGDOM), Runze LI (UNITED KINGDOM), Silvestre TAVEIRA PINHO (UNITED KINGDOM)Abstract :
Reducing the environmental impact of aircraft production and operations stands as a paramount goal for the aeronautical industry. The adoption of composite materials to minimise weight has already become pervasive, featuring prominently in the flagship models of civil airliner manufacturers.
Currently, high aspect ratio wings, which offer higher aerodynamic efficiency, have become a focal point of investigation among the scientific community as serious candidates for the aircraft of the future. However, long wings introduce some challenges, such as high deformations and complicated aeroelastic relations.
A novel wingbox structural (Z-Beam) concept [1], capable of intrinsically tailoring the aeroelastic properties of aircraft wings has shown great promise, especially when it comes to introducing desired bend-twist coupling with minimal weight. However, departing from traditional wing-structure configurations creates significant design challenges, as traditional design rules no longer apply. Additionally, the detailed failure analysis of such novel structure is not readily feasible with a full-scale model since smaller structural details such as bolt joints require relatively small mesh sizes to provide accurate results, which exponentially increases computational times.
In this work, a multi-scale approach is followed to predict failure of an identified hot-spot based on the global scale behaviour using finite element simulation and machine learning. This is done by using a global-local simulation wherein the boundary conditions of the local model are derived from a global model (Figure 1), allowing not only to reduce mesh size constraints but also to use different element types, capable of more accurately simulating composite failure. The global model, an elementary cell of the Z-Beam structure, is loaded in different ways using periodic boundary conditions. From the obtained displacement field, the displacements at the boundaries of the local model are retrieved, and a distinct analysis employing a model [2,3] evolved from LaRC05 [4] failure criterion is conducted to assess its behaviour (Figure 2). This procedure is repeated several times with different loadings cases, enabling to train a machine learning based surrogate model, which is ultimately able to predict the failure of a specific relatively small detail based on the loads applied on the global model, with reduced computational costs. Noteworthy benefits of this methodology include a reduction in computational times and the preservation of a high degree of accuracy in the failure analysis of smaller details.
Currently, high aspect ratio wings, which offer higher aerodynamic efficiency, have become a focal point of investigation among the scientific community as serious candidates for the aircraft of the future. However, long wings introduce some challenges, such as high deformations and complicated aeroelastic relations.
A novel wingbox structural (Z-Beam) concept [1], capable of intrinsically tailoring the aeroelastic properties of aircraft wings has shown great promise, especially when it comes to introducing desired bend-twist coupling with minimal weight. However, departing from traditional wing-structure configurations creates significant design challenges, as traditional design rules no longer apply. Additionally, the detailed failure analysis of such novel structure is not readily feasible with a full-scale model since smaller structural details such as bolt joints require relatively small mesh sizes to provide accurate results, which exponentially increases computational times.
In this work, a multi-scale approach is followed to predict failure of an identified hot-spot based on the global scale behaviour using finite element simulation and machine learning. This is done by using a global-local simulation wherein the boundary conditions of the local model are derived from a global model (Figure 1), allowing not only to reduce mesh size constraints but also to use different element types, capable of more accurately simulating composite failure. The global model, an elementary cell of the Z-Beam structure, is loaded in different ways using periodic boundary conditions. From the obtained displacement field, the displacements at the boundaries of the local model are retrieved, and a distinct analysis employing a model [2,3] evolved from LaRC05 [4] failure criterion is conducted to assess its behaviour (Figure 2). This procedure is repeated several times with different loadings cases, enabling to train a machine learning based surrogate model, which is ultimately able to predict the failure of a specific relatively small detail based on the loads applied on the global model, with reduced computational costs. Noteworthy benefits of this methodology include a reduction in computational times and the preservation of a high degree of accuracy in the failure analysis of smaller details.