NAFEMS Americas recently presented a webinar with Linda Knudsen from Syncroness on the importance of simulation result credibility, titled ‘Define and Deliver Credible Results.’ We summarised some of the key points in this article, and you can view the full webinar in our resource centre.
Here are some of the questions (with the answers!) that the audience put to Linda at the end of the presentation.
How do different stakeholders participate in the credibility assessment process? For example, who creates the Model Risk? Manufacturers, certification bodies, or both?
Both. It’s a collaborative process, something you would put out there to all your stakeholders before you run the simulation. Effectively you would ask ’If I do these things, will the result be credible?’ and the stakeholder could reply ‘no, I need more of this,’ for example. Certainly, involvement from everyone is encouraged.
I am interested in how we should do VVUQ with FE material modelling, are there any resources you can recommend?
The Minerals, Metals & Materials Society (TMS) offer classes and they have done a lot of work on specific materials models.
In your experience, is working on a medical device which is quite similar for all patients but requires that certain individual aspects of each patient be taken into account particularly risky or not?
This is where the context of use (COU) and the credibility of the computational model for that COU come in. If a particular computational model is only meant to inform, say, fatigue performance of the device, then you would need another model for any other factor. Clarity in defining what you need a computational model to do is paramount here, and lack of it can lead to a lot of frustration. So, in terms of risk, the scenario described in the question depends on whether the computational model is credible for the ‘certain individual aspect’ you are varying and its range in variation.
How can we perform validation activities if we have no chance to execute real tests for comparison with the obtained simulation results? How is the credibility of the simulation results demonstrated then?
This is quite a common problem that we have because it's difficult to always measure what you are most interested in. And so, in this example, there is an abstraction relating fatigue performance to stiffness that is presented. This abstraction is handled by understanding that we can still determine Model Form and the other credibility factors for the validation to stiffness even without having the equivalent stiffness. Then, with the other credibility activities including mesh refinement and sensitivity, we can be sure that with small strain elastic conditions we really are representing the stress correctly. It is not straightforward or as easy as we want it to be, but with thoughtful engineering judgment, you can make those abstractions.