This paper was produced for the 2019 NAFEMS World Congress in Quebec Canada
One of many critical analysis tasks in aircraft design is the evaluation of structural stability. This is frequently done using the finite element method to predict buckling behavior. For simple structures (like a column or pressure vessel), this is a simple and straightforward process. The FEA solution returns an eigenvalue that represents the proportional value of the buckling load, along with an eigenvector of the buckled shape. However, for complex structures and assemblies, interpreting the results from the Eigen solution can be much more complicated.
There are two challenges interpreting buckling results for complex structures. They are: 1) isolating global (or general) versus local (or panel) buckling modes, and 2) identifying and modeling the modes to represent the buckling response for a particular panel. Global buckling occurs when the entire structure (or a large portion of the structure) undergoes buckling. This is usually a catastrophic event. Conversely, local buckling is characterized by a small portion of the structure buckling (isolated to a skin or tertiary structure). Local buckling may, or may not, lead to catastrophic failure.
Over the years, engineers have used a variety of methods to identify buckling modes of interest. For global modes, an engineer typically has to review every buckled mode shape (eigenvector) of the buckling solution. They can generally isolate and identify which are “global” buckling modes vs. “local” panel modes. However, for complex structures, the process is subjective, time consuming, and inefficient, as it requires the engineer to manually examine hundreds of mode shapes. In addition, it cannot be automated (for processes like design optimization).
When local panel modes need to be evaluated, engineers employ a variety of modeling techniques to isolate the panel buckling solution from the global model. This limits the buckling response to the panel of interest. While these methods are convenient, they may introduce undetected errors.
To avoid these issues, a statistical method to evaluate modes was developed. The implementation was done with a MSC Nastran DMAP ALTER. While it is useful, it isn’t always convenient. For example, the user must include the DMAP ALTER with the buckling analysis. This limits its application when multiple evaluations are required.
To mitigate the DMAP limitations, a post processing procedure was developed that can be run after the analysis is completed. This procedure leverages the HDF5 results file from MSC Nastran. This paper provides a brief background on buckling analysis, the statistical methods used to identify buckling modes, the implementation using Python and PyTables to access HDF5 output, along with examples that demonstrate the use of the statistical evaluation tool.