Session: 10-08-01: Multiphase Flows and Applications

Paper Number: 150071

150071 - Real Time Refrigeration Simulation: Using Simscape to Make Data Driven Models 

Thermofluid systems, such as refrigeration, are growing in importance due to electrification and climate change.  Efficient thermal management is needed for electrical systems and human comfort.  Models of these systems offer more value to the development process if they can be used for HIL testing. 

Refrigeration systems often involve multiple working fluids, some in a closed system.  The systems commonly include actuators, pumps, compressors, fans, valves, and controllers.   The refrigeration cycle is often the limiting factor in getting a model to be real-time capable.  We will see how to use data-driven approaches to simplify that part of the system so that it will run in real time.  This capability will give us an important advantage over other simulation tools.

Models using the two-phase domain are typically the hardest to make real-time capable.  The physics are more complex and they tend to interact with other fluid systems, resulting in a wider range of time constants. 

 

The workflow for optimizing a refrigeration network model involves a series of detailed steps to ensure accuracy and efficiency. Here is an expanded summary:

Generate Data: Begin by creating a comprehensive dataset using a detailed model that captures all relevant conditions and scenarios of the refrigeration network. This ensures that the data covers the entire operational envelope and all possible variations in system behavior.

Build and Validate ROM: Utilize tools such as the Regression Learner App in MATLAB to build a Reduced Order Model (ROM). This involves training the ROM on the generated data to accurately replicate the behavior of the original detailed model. Validation is a critical step here, where the ROM's predictions are compared against the actual data to ensure high fidelity and reliability.

Replace Simscape Refrigeration Network with ROM: Once validated, replace the detailed Simscape refrigeration network within your simulation environment with the ROM. This step is crucial for simplifying the model, reducing computational load, and speeding up simulations without significantly compromising accuracy.

Verify Performance: Conduct thorough performance verification of the ROM by comparing its output against the original detailed model. This involves running simulations in both variable-step and fixed-step modes to ensure that the ROM performs consistently and accurately under different simulation conditions. This step helps identify any discrepancies and ensures the ROM's robustness and reliability.

By following these steps, the workflow aims to create an efficient, accurate, and computationally less intensive model of the refrigeration network, facilitating faster simulations and easier analysis while maintaining the integrity of the original system's behavior

Presenting Author: Andrew Greff MathWorks

Presenting Author Biography: Andrew Greff is a senior application engineer at MathWorks. He specializes in physical modeling using Simscape and focuses on thermal, fluid, and multibody systems. Before joining Mathworks, Andrew worked for GM and Stellantis developing advanced hardware and controls for engines. He obtained his PhD in mechanical engineering from the University of Alabama.

Authors:

Andrew Greff MathWorks