Session
2023 session 6
Location
Weber State University
Start Date
5-8-2023 11:20 AM
Description
Centrifugal flow impellers are commonly used in a wide variety of industrial machines. Predicting the performance of these impellers over the entire operating range is key to the system development during the early design stages. The two element in series (TEIS) and two-zone model can be used to accurately predict impeller performance based on flow physics and empirical correlations. Correlations were made with linear regression on a database of 50 pumps and 75 impellers. These correlations were later found to only apply to designs that are similar to those in the database. This paper proposes a new method to generate correlations for the TEIS and two-zone model using reduced order modeling and machine learning. The models will be trained using the same database used in developing the previous correlations, but will also include additional data and computational fluid dynamics (CFD) results. The models will be used to improve the known correlations and to discover new correlations between the machine performance and design variables.
One-Dimensional Radial Turbomachinery Modeling
Weber State University
Centrifugal flow impellers are commonly used in a wide variety of industrial machines. Predicting the performance of these impellers over the entire operating range is key to the system development during the early design stages. The two element in series (TEIS) and two-zone model can be used to accurately predict impeller performance based on flow physics and empirical correlations. Correlations were made with linear regression on a database of 50 pumps and 75 impellers. These correlations were later found to only apply to designs that are similar to those in the database. This paper proposes a new method to generate correlations for the TEIS and two-zone model using reduced order modeling and machine learning. The models will be trained using the same database used in developing the previous correlations, but will also include additional data and computational fluid dynamics (CFD) results. The models will be used to improve the known correlations and to discover new correlations between the machine performance and design variables.