Imagine the ground outside of Leonardo da Vinci’s window. I picture a heap of scrap wood and canvas, as he experimented with the idea of flying machines. His ink drawings and iterative physical models set the baseline for engineering. Early CAD systems simply automated the sketching process, adding modeling, dimensioning, and manufacturing plans along the way—eventually yielding the capability to develop 3-D solid models with a high-fidelity representation of the system.Designers, analysts, systems engineers, manufacturers, and sustainers now can work from the same high-fidelity model.
Now, with the application of high performance computing, we are gaining true digital twins: realistic instantiations of actual systems. We’ve moved from trial-and-error to a systematic, science-based optimization process to explore trade spaces, component performance and interaction, manufacturing processes, and life cycle operation. Think of the guy in a garage versus a systematic scientific exploration of the trade space.
Dominating with digital
The methods of the past had their flaws. The design and the as-built object were distinct and often exhibited differences in performance due to simplifications, manufacturing variances, and limited computational capabilities. Ideas, concepts, and models were only representative of the actual system.
The benefit of a full-fidelity digital model, coupled with a computational performance model, is that the digital system will perform in a manner that more accurately reflects the actual physical system. This vastly reduces unexpected events in tests and enables the exploration of a much greater design space.
It also enables the modeling of the system life cycle to support performance-based maintenance. The remarkable developments in computational and computer sciences have brought simulation theory and technology into a new era where digital models can realistically model the key physics driving system performance. Digital twin technology will allow for better designs, as it can consistently, reliably, and easily adapt to changing factors.
The implications don’t end with engineering—as you might recall from grade school, scientific theory attempts to explain some aspect of our world, based on observation. We then test and confirm to the best of our abilities. The concept is built on two pillars: theory and experimentation. Computational science offers a third pillar, as discussed in detail in a 2005 report from the President’s Information Technology Advisory Committee: “Together with theory and experimentation, computational science now constitutes the ‘third pillar’ of scientific inquiry, enabling researchers to build and test models of complex phenomena.”
The possibilities offered by an age of digital twins also come with a warning: know what you’re doing! Fidelity refers to your trust level in your digital models. SAIC works tirelessly to join scientific domain expertise with HPC expertise so that engineers know why a model is saying X instead of Y.
Trust in the models becomes even more important as HPC and simulation-based engineering and science move further into the mainstream. Manufacturers and users want higher fidelity design and decision-support tools. Additionally, the growing complexity of systems is requiring systems-based research and development, systems-based design, systems-based engineering, and physics-based technology assessment (trade space studies).
A host of critical technologies--bespoke manufacturing, self-driving cars, hypersonic transportation, space travel--is on the horizon that cannot be understood, developed, or utilized without simulation methods. SAIC and its team of computational scientists are on the front lines of those methods, partnering with our government customers to build and use the third pillar.
About the author: Hugh Thornburg is the Department of Defense High Performance Computing Modernization Program's Productivity Enhancement, Technology Transfer, and Training program technical lead. He has 28 years of experience in applied computational fluid dynamics, high performance computing, mesh generation, and multi-physics tool development and analysis. Hugh provides leadership for acquisition of state of the practice tools and methods and leads teams supporting the use of such tools and methods through application management, research, consulting, and collaborative work with both users and developers. He has a master's degree and Ph.D. from the University of Cincinnati and a bachelor's degree from Rose-Hulman Institute of Technology.