Universals Civil Engineer.

With FEA, you can model material behavior and simulate conditions to develop a solution to a certain problem. However, without employing appropriate validation techniques, how can one trust the outputs from FEA?

Speaking at the British Universals Association Early Career Group event in May, Dr Andrew Lees of Tenser explained: “The more complex the analysis method, the more potential for error.”

Validation of FEA

Validation in terms of FEA means checking whether the FEA model and output represent reality with sufficient accuracy.

It’s important to realist that FEA will not model reality entirely, but has to simulate it sufficiently in order for the design to be appropriate.

However, the assumptions, decisions, and approximations identified throughout the design process are more influential on the FEA model and subsequent output compared to the output of more conventional methods, so there’ more potential for error.

Without validation, engineers risk adopting a solution or technique that doesn’t conform to reality, as evident in the Nichols Highway collapse in Singapore 2004, where four people lost their lives.

One factor in the collapse was the incorrect use of FEA in the design, particularly in modelling the ground conditions.

However, we shouldn’t be intimidated with the use of FEA.

As Dr Lees highlights, “there are benefits to be gained from more advanced analysis, but clearly you need to invest more time and resources into them to check they are correct and they are performed with proper validation practices”.

How is validation performed?

The first action to validate an FEA output is to see whether the output looks right and conforms to what’s expected from previous, ‘well-winnowed experience’. However, this expertise and experience isn’t always present in those in the early stages of their career.

Learn from the literature

An FEA user might also investigate how previous engineers have approached a similar problem they’re facing.

One can use this technique by reviewing published case studies. However, just because something is published doesn’t mean it’s correct.

Another issue then arises of how to determine whether a case study is correct or not. Typically, case studies from trusted journals are acceptable to reference, but sorting the reputable case studies and journals from the erroneous ones, once again comes with experience.

An engineer can also use previous case histories that didn’t utilize FEA to validate their output.

The engineer would need to set up another FEA model to mimic the same conditions from the case history. If the results from the case history and the FEA model glean similar results, the engineer has indirectly validated their original model. Depending on the discipline, one could also compare their FEA output with monitoring data via an observational approach.

However, knowing what validation is and how to execute it are two different subjects. There are challenges to overcome in executing data validation. One such challenge, is how to overcome the digital generation gap.

Overcoming the digital generation gap for FEA validation

There may be an in-house experienced engineer within one’s company, with knowledge of FEA who can therefore support the FEA user in their work. However, more commonly, it’s early career engineers that utilize FEA, rather than experienced engineers.

While having little knowledge of FEA, the experienced engineer has the experience and expertise of engineering behaviors and thus knows the anticipated outcome for the design.

Early career engineers must seek out this experience to validate their FEA model and outputs. To do this, Dr Lees recommends talking about the FEA outputs, but stripping the problem back to discuss the engineering behaviors behind the FEA model.

Language matters

To overcome the digital generation gap, a method is to phrase questions to the experienced engineer in such a way that defines the problem without delving into the mechanics of the FEA model.

For example, for a Universals FEA model, instead of asking “what constitutive model should be used?” the question can be rephrased as “which aspects of soil behavior are important to this problem?”

This enables the experienced engineer to engage with the problem, by recommending the appropriate soil behaviors critical to the design. The FEA user can then select the constitutive model that appropriately models the relevant soil behaviors.

Another example is to see whether outputs from FEA look correct. If the output is in a form that the experienced engineer understands, such as a graphical plot of stress vs strain. An early career engineer can seek help from the experienced engineer in validating their model.

The experienced engineer, with the help of the output, could recommend an alternative method to further validate the FEA output.

Ultimately, to bridge the digital generation gap, the solution lies in communication and collaboration. FEA validation is a team activity, and should not be undertaken in isolation.

Stripping the problem back to the engineering principles and behaviors allows early career and experienced engineers to discuss the problem with confidence. Working as a team, the challenges of FEA validation are easier to overcome, with the outputs proven to sufficiently simulate reality, via appropriate validation techniques.

As civil engineers, we must make sure we validate our FEA outputs appropriately. We must not simply take things for granted, so challenge the outputs from your FEA and start employing validation techniques to ensure you have sufficient confidence in your designs.