(7c) Reduced Order Models (ROMs) for Simulation Democratization and Digital Twins

CFD analysis generally tends to take relatively long time and requires significant computational resources, which makes it difficult for real-time monitoring or predictive maintenance. In addition, prior knowledge/experience is required to perform CFD simulations and extract meaningful results. Reduced Order Models (ROMs) can significantly reduce these requirements. ROM is a simplification of a high fidelity computational model that preserves essential behavior and dominant effects, for the purpose of reducing solution time or storage capacity required for the more complex model. In this paper a series of use cases for ROMs will be covered. These use cases will highlight how ROMs can be used by operators /non-simulation experts to perform ‘what-if’ analysis within second without compromising on accuracy, to take engineering decisions. ROMs also act as virtual sensors for controllers as well for guiding the placement of physical monitoring systems. Thirdly, ROMs can also be integrated into system level analysis for operational optimization and to digitize the asset. The created ROMs are not just input-output signal type models but complete 3D field view models. These 3D-nonlinear ROMs can also be used separately for quick estimates of the outputs to create performance charts for the operators in the field. The results which took hours on multiple CPUs earlier can now be achieved within seconds/minute on single CPU. This capability is extremely helpful for using simulation model for predictive maintenance and prognostic health monitoring where real time results are useful.