Session: K6-08: HEAT TRANSFER IN ENERGY SYSTEMS - WASTE HEAT I

Paper Number: 122272

122272 - Digital Twin Development of a R134a Plate-Tube Evaporator 

Abstract: 

The increasing adoption of heat pumps for waste heat recovery heralds a promising shift away from fossil fuel-operated boilers which cater to the diverse demands of both brownfield and greenfield industrial applications. However, an essential challenge in the design of heat pumps lies in accurately predicting the system's transient and dynamic behaviour, particularly when faced with sudden changes in demand (sink) and source sides. The conventional approach of conducting high-fidelity numerical simulations or experiments at each possible off-design operational conditions is (computationally) expensive. To address these challenges, low-fidelity system-level models offer efficient solutions for developing a digital twin of a heat pump, allowing performance assessments for various scenarios. Yet, suchmodels for often rely on general empirical formulations specific to particular cases and/or solving one-dimensional governing equations across the component domain. However, the components of heat pumps are considerably more complex than what can be described by such models. This frequently leads to disparities between experimental data and the low-fidelity model, thereby limiting their ability to accurately estimate performance, especially under transient and dynamic conditions. This work proposes a machine-learning framework for reliable surrogate modelling of evaporators, crucial components within heat pumps, by training nonlinear regression models with high-fidelity unsteady CFD data. This is applied to a plate-tube type heat evaporator within a single-stage heat pump utilizing R134a refrigerant to recover heat from waste water. The design parameters are chosen as the refrigerant and water mass flow rates, inlet pressure, and enthalpy values varies under transient conditions. The Latin hypercube data-sampling method is employed to generate a data space along the selected design variables. Reynolds-averaged Navier-Stokes (RANS) filtered flow equations, coupled with the Volume of Fluid (VOF) method for modeling phase change in the 3D heat exchanger domain, are solved at each design point to generate the training data. Gaussian process regression, trained using 80% of the generated data and validated by the remaining set, is utilized to identify the surrogate model at each time interval during operation. This surrogate model represents the digital twin of the evaporator and can subsequently be implemented in system-level models of the heat pump. This framework enables efficient performance predictions of the considered heat pump yet also represents a generic methodology for digital twining of heat exchangers with different configurations.

Presenting Author: Ali Can Ispir Eindhoven University of Technology

Presenting Author Biography: Dr. Ali Can Ispir is a Postdoctoral Researcher at Eindhoven University of Technology. He completed his Bachelor's degree in Mechanical Engineering in 2014 and went on to obtain his Master's degree in Mechanical Engineering, specializing in heat-fluid dynamics, in 2017. In 2018, he successfully completed the Research Master's program at the Von Karman Institute for Fluid Dynamics. He furthered his academic journey at the same institute and obtained his Ph.D. degree in 2023. Currently, he is focusing on the development of digital twins for industrial heat pumps in his postdoctoral research.

Authors:

Ali Can Ispir Eindhoven University of Technology
Michel Speetjens Eindhoven University of Technology