Session: 07-02: Fluidized Bed Heat Exchangers

Paper Number: 164856

164856 - Predicting Local Heat Transfer in Solid-Particle Gravity-Driven Fluidized Heat Exchangers Using Machine Learning 

Abstract: 

The growing global energy demand underscores the critical need for sustainable energy sources. Solar energy, particularly through Concentrating Solar Power (CSP) systems, presents a promising solution. Generation 3 CSP systems employ high-temperature heat exchangers (HX) to transfer thermal energy from a ceramic-based working fluid, such as sand, to tube walls containing a secondary fluid, such as sCO₂, that drives the power cycle. These HXs – designed to operate at temperatures exceeding 700°C, must efficiently extract heat from solid media while integrating seamlessly into the CSP system – represent a significant engineering challenge. Current HX configurations investigated for application in CSP systems include gravity-driven designs (e.g., shell-and-tube and parallel plate) – offering lower cost and easier integration – and forced (e.g., fluidized bed) designs – offering better thermal performance.

We propose a vertical fluidized HX that merges the cost-effectiveness and simplicity of gravity-driven configurations with the superior heat transfer capabilities of fluidized designs. Although fluidization has been studied for many years, characterizing the local particle-to-wall-contact interactions and their influence on heat transfer coefficient (HTC) remains challenging due to the dynamic motion of ceramic particles and limitations of accurate heat transfer measurements.  The absence of detailed insights into the local HTC at the particle-wall interface constrains efforts to design high-performance HXs.

This research advances the analysis of heat transfer in fluidized systems by using Machine Learning (ML) algorithms to predict local heat transfer performance from solid media to adjacent tube walls. To achieve this, we collected experimental data on granular interactions with a circular tube using a high-speed camera, to capture various fluidized velocities in a lab-scale test rig. We then used a CFD package to simulate the sand and air motion within the test rig, creating a digital counterpart of the experiment. Furthermore, we modeled energy transport in the system by assigning appropriate temperature boundary conditions to the tube wall, granular media, and air to replicate operating conditions.

The CFD results served as the training data for the ML algorithm, which include particle velocity, granular concentration and local HTC around the circular tube. We validated the ML model by comparing experimental velocity and granular concentration. Finally, we applied the trained model to predict the transient local HTC. At 30 psi air pressure, the local HTC ranged from 160 to 250 W/m²K. We identified four distinct zones around the circular tube: a stagnant zone at the top, a periodic stagnation zone with large voids at the bottom, and high-velocity, high-mixing zones along the sides.

These findings offer valuable insights into the fluidization process by quantifying local heat transfer and predicting HTC across various operating conditions. This approach aims to optimize fluidized HX performance by pinpointing fluidization regimes that most effectively transfer heat to HX walls, providing a framework for enhancing CSP systems and other related applications.

Presenting Author: Julio Izquierdo university of louisville

Presenting Author Biography: Julio Izquierdo is a Ph.D. student of Mechanical Engineer at the University of Louisville. His research focuses on improving heat exchanger performance in industrial applications. He is a member of the American Society of Mechanical Engineers (ASME) . Outside of work, he enjoys traveling and scuba diving.