Reduced order modeling for optimal aerodynamic design and operation of the industrial air-jet ejector.

A reduced-order model was generated from aerodynamic simulation data for the optimal design and operation management of an air-jet ejector to remove contaminated materials in manufacturing plants. Three significant design parameters of the nozzle structure after studying the computational fluid dynamics (CFD) of the preliminary design cases were determined. Polynomial type and deep neural feedforward network-based meta-models were established based on parametric CFD simulations considering three design variables, which were used to predict the suction performance; the models were successfully validated by comparing the results with those of the complete CFD. Finally, a proper orthogonal decomposition was adopted to reconstruct an internal flow field within a few seconds to facilitate real-time monitoring of the air flow without complete CFD simulations, which was realized as a digital twin for the air-jet ejector system. [ABSTRACT FROM AUTHOR]