FEA For Structural Engineers: Continuous Learning

During my studies, I learned a great deal about finite element modelling -- relative to what I knew before... My Master's research model incorporated a number of unique element types for capturing uplift-behaviour, non-linear elements to capture material crushing, 1D elements to capture post-tensioned cable behaviour, 2D beam elements representing wall panels acting in bending and shear… I also developed a rough understanding of the analysis settings used by the processor. But a year and a half on, working in industry, I’m still working on my finite element analysis (FEA) skill set, and in a very different, commercial software environment.

There are at least a couple motivations for this reflective article.

1. Like this blog from Autodesk, I believe structural engineering (SE) is trending towards increasingly complicated designs, and demanding more efficiency. If SE isn't replaced by computer automation in ten or twenty years, it's because those working in the industry are capable of addressing complicated structural engineering problems efficiently, and adapting our solutions to evolving designs - - more reliably than a computer program without a clearly laid out and verifiable thought process and calculation set.
2. A bit like this blogger, I think plenty of FEA models have been relied on, but built without appreciation of many of the input parameters (granted, there are many input parameters, and not all need to be understood for an adequate model, in some cases). It's a problem that can be addressed with training, and oversight from engineers with adequate experience to make a judgement on the outputs. But there is an onus on those who want to be part of the future SE industry, to understand the complexities of FEA for ourselves, to be able to push the boundaries of complex designs, and to become more efficient. I've put in a solid effort on my own, but I find that many online resources and blogs are written for the mechanical engineering industry, so this article includes links to a number of interesting resources I've found for the SE industry (aside from the FE software support forums).

Building on Finite Elements (Ben Curry, Being Brunel)

For those looking for a general appreciation of structural FE elements, a down-to-earth article is available on a blog called "Being Brunel", in honour of the great English engineers. The author generally outlines the differences between bars and beams, plates and shells, and more. It's important to understand the difference between these elements - - especially because you can build a converging model with either a bar or a beam (for example) but obtain very different results! Only one is (approximately) correct.

This same article gives some basics regarding meshing. It's by no means in-depth, but it's a first step, in a very readable format.

Adaptive Mesh Refinement for Civil and Structural Designs (“The Engineer” on engineering.com)

This article is software-specific, as it discusses a feature, adaptive meshing, available exclusively in one package. However, it shines a light on how automation may take over the manual (inefficient) modelling practice of refining the FE mesh. It's not a promo-only blog either: there are links to the academic papers behind the adaptive meshing concept. In short, (WARNING: this gets technical) adaptive meshing requires that the software has an idea of what the results should theoretically be, and with each iteration of an evolving FE mesh, it seeks to reduce the difference between the FEA results and the theoretical prediction. It does so by increasing the density of the finite element mesh in specific areas of the model.

Autodesk Robot Webinars: Instability warnings during model analysis

This webinar is also software-specific, to a degree. However, the concepts of instability warnings are all too important no matter the FE package. The webinar looks at warnings about model instability, how the instabilities can be resolved, or if they can be ignored (and how to tell for yourself).

Singularities in Finite Element Models (Henrik Sönnerlind, COMSOL Blog)

The author looks at meshing, and “Red Spots”, i.e. the extreme analysis results which are the product of sharp corners, point loads, constraints, welds, and cracks on a surface model (mesh singularities). These red spots usually mean extreme spikes in the results near these singularities, and the author points out that it can sometimes be difficult for a structural engineer to decide to ignore these issues. The author also discusses how to resolve these issues in many cases. I especially appreciate the practical structural engineering context on this matter.