Abstract
Abstract Multi-scale dynamic capture models are often too computationally expensive for use in real-time applications, such as operator training and online process control. As a result, innovative methods are required to reduce the complexity of PDE-based and rate-based dynamic models while preserving input-output behavior. In this study, dynamic reduced models (D-RMs) approximate the high-fidelity capture models, thereby offering a trade-off between accuracy, range of applicability, and computational cost. In this paper, we highlight a D-RM builder tool that automatically generates fast D-RMs using pre-computed results from repeated simulation of high-fidelity dynamic process models. Dynamic uncertainty quantification (UQ) analysis is provided that quantifies confidence in the generated D-RM when used as a surrogate ‘plant’ model. In addition, an Advanced Process Control (APC) framework tool is presented, which uses these reduced-order surrogate models as fast predictive-model(s) for implementing various nonlinear model-predictive control (NMPC) strategies. The developed framework is targeted towards superior disturbance rejection and better responsiveness to demand-changes for otherwise slow capture processes.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have