The path toward certification by simulation, Part 3: Simulation governance
Simulation governance is reportedly a prerequisite for certification by simulation, basically enabling V&V and UQ. And yet, it is claimed that the design of most simulation tools used today makes this very difficult, if not impossible. What is this hurdle and how do we overcome it?
I started this series of blogs with reference to Dr. Byron Pipes’ strategy to reduce the time and cost for composite structures certification by increasing the use of simulation and virtual testing. As he described in his March 2014 HPC column, “Accelerating the certification process for aerospace composites”:
“The most direct approach to reducing the cost and time of product certification is to first certify the simulation tools themselves. Certified simulation tools could be used by engineers with the expectation that, within the bounds of the certification, no further verification or validation would be necessary. This would result in enormous savings because the contemporary approach involves verification and validation testing by each user before the simulations can be trusted.”
This may appear pretty radical: certifying the simulation tool vs. the individual structural and/or manufacturing simulation. However, Pipes isn’t alone.
Dr. Barna Szabo, a retired professor of mechanics from Washington University (St. Louis, Mo., USA), asserts, “The only way to achieve certification by simulation is by instituting simulation governance.” He defines simulation governance as the procedures to ensure the reliability of numerical-based simulation predictions with the goal to reduce uncertainty and errors, including:
Formulating rules for collecting, verifying and archiving the experimental data input into simulations;
Establishing protocols for the revision of simulation tools and models in light of new information from UQ and physical testing.
Szabo is principal author of two textbooks on finite element analysis (FEA) simulation and is co-founder and president of Engineering Software Research and Development (ESRD, St.Louis).
One of Szabo’s more intriguing assertions is that the conventional software tools in use today were not designed to support simulation governance. He says this is because the infrastructure of these products evolved long before the more recent maturation in simulation technology and formulation of the technical requirements for V&V. He explains, “The majority of the FEA programs have large finite element libraries that combine the mathematical model with its numerical approximation, making it impossible to identify and control the errors of idealization separately from the errors of approximation.” (See the figure below).