THermal management

Back to all research

Thermal management of aircraft or ground vehicle

Modern aircraft and ground vehicle contain highly complex and integrated systems, especially recently with More Electric Aircraft becoming the trend of the next generation aircraft. Industries face substantial challenges in developing new thermal management architectures and methodologies to maintain electrical power requirements for each system and to meet overall aircraft or ground vehicle performance and safety requirements. One type of thermal management technology being investigated is the vapor cycle system, which is composed of compressor, condenser, expansion valve, and evaporator.

Centrifugal compressor high fidelity simulation


AIAA SciTech


Centrifugal compressor

Centrifugal compressors have widely applications such as in turboshaft engines, refrigeration systems, fuel cell electric vehicles. The design and off-design performance can be estimated from map-based calculations or physics-based one-dimensional models. In the physics-based one-dimensional models, individual losses are computed using analytical equations and empirical correlations which relate loss levels to velocity diagram characteristics and overall geometry. The calculated losses are then used to estimate overall changes in enthalpy loses which in turn used to estimate the outlet pressure and temperature rise. One-dimensional models provide good guidance on preliminary design of centrifugal compressors. However, it is found that the predicted total pressure ratio and efficiency were higher than the experimental measurements, when the compressor mass flow is limited by the impeller. In the present study, CFD simulations are used to understand the flow characteristics in the centrifugal compressors, and to further improve first principle modeling. Using ANSYS CFX to understand flow phenomena and predict the off-design performance of centrifugal compressors. We first start with air as the fluid medium, then extend to refrigerants. In our study, we begin with fundamental input parameters and geometry data, generate a preliminary design of the impeller using Vista Centrifugal Compressor Design (CCD). Then the resulting geometry is passed to BladeGen to generate impeller blades. The obtained centrifugal compressor is then meshed and CFD simulations are conducted using ANSYS CFX. In this work, we conduct both 1D modeling and CFD simulations to study the sizing and performance of air-based and refrigerant-based centrifugal compressors.

Centrifugal compressor modeling using machine learning techniques




Gaussian Process Regression

Due to the robustness and flexibility in model development, machine learning techniques are also used to modelling the performance of centrifugal compressors. In modeling, the operating conditions of a centrifugal compressor can be specified by two input parameters, i.e., the impeller rotational speed, and the mass flow rate. The corresponding performance of the centrifugal compressor can be described by pressure ratio and efficiency. We conducted CCD simulations to generate compressor data, based on which typical machine learning techniques are used to model the centrifugal compressor performance. We first use various regression models to predict the compressor performance map, focusing on the interpolation of pressure ratio. Then we study the extrapolation performance of GPR and ANN. We would like to extend the current framework to modelling the performance of other type of centrifugal compressors with different sizes and working fluid medium.

OCSE A446, UCCS, CO 80918-1740
PHONE: (719)-255-3135