System & Balanced Complexity
Modeling any system — whether it’s an analysis of the 9/11 tragedy, study of the human body modeling (HBM) or aerodynamic behavior of a Formula 1 car — can take a great amount of time. Including geometrical details that may not affect results stretches computational time further, especially if performing parametric analysis and optimization. Alternatively, oversimplifying a model in terms of dimension, physics or scope may lead to misleading conclusions. Engineering and modeling engineering skills are still required to make proper assumptions regarding the target model, available computational power and goals.
A reduced-order model (ROM) locally reduces the size of a model and computational task, applicable to areas in which details of the result are less important. This technique is used increasingly for systems models, in which detailed models are required only in the regions of interest. Consider these examples: Walls of a skyscraper are modeled as 2-D shell elements or arteries of the cardiovascular system as 1-D element. This allows modeling of the entire building or full human body, respectively, enabling parametric simulation in a reasonable timeframe.
Alternatively, for some detailed analyses, in-depth troubleshooting investigations or marketing demonstration, very high-end simulations can be required of the entire system. Extreme software scalability combined with several thousand cores of supercomputing power provide the necessary infrastructure to support and balance the problem complexity.