Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Innovative Tubular Corrugation Pattern for Heat Exchangers Using CFD and Machine Learning

Liao, Liangdou LU (2026) MVKM05 20261
Department of Energy Sciences
Abstract
This thesis aims to develop and evaluate a fast design and optimisation workflow for corrugated tubes used in tubular heat exchangers by combining computational fluid dynamics (CFD) simulations and machine learning. A steady-state Reynolds-averaged Navier-Stokes (RANS) model was established to simplify the simulation process and reduce computational time. Based on the selected CFD setup, a CFD database including 45 corrugation geometries and 135 simulation cases was generated using discrete Latin hypercube sampling (LHS). Gaussian process regression (GPR), feed-forward neural network (FFNN), and symbolic regression (SR) models were developed and evaluated for pressure drop and heat transfer coefficient (HTC) prediction. The FFNN model was... (More)
This thesis aims to develop and evaluate a fast design and optimisation workflow for corrugated tubes used in tubular heat exchangers by combining computational fluid dynamics (CFD) simulations and machine learning. A steady-state Reynolds-averaged Navier-Stokes (RANS) model was established to simplify the simulation process and reduce computational time. Based on the selected CFD setup, a CFD database including 45 corrugation geometries and 135 simulation cases was generated using discrete Latin hypercube sampling (LHS). Gaussian process regression (GPR), feed-forward neural network (FFNN), and symbolic regression (SR) models were developed and evaluated for pressure drop and heat transfer coefficient (HTC) prediction. The FFNN model was selected for the design space sweep and Pareto front analysis. The Pareto fronts showed the trade-off between pressure drop and HTC and were used to select candidate geometries under the given pressure drop constraints. The SR model provided explicit mathematical expressions for understanding the main parameter trends. These selected geometries were then revalidated using CFD simulations to evaluate the prediction accuracy of the FFNN model. The selected designs can be regarded as candidate optimal designs within the investigated design space under the given pressure drop constraints, and can provide useful design guidance for further improvement of tubular heat exchangers. (Less)
Popular Abstract
Designing Heat Exchanger Tubes with Higher Efficiency:

Heat exchangers are used to transfer heat from one medium to another. They are important in many industrial processes, such as heating, cooling, and food processing. In the food industry, tubular heat exchangers are often used because they are simple and hygienic to process food products such as milk and juice.

This thesis focuses on a special type of tube called a corrugated tube. A corrugated tube has repeated patterns on its wall. These patterns can make the fluid mix better when it flows through the tube. Better mixing can help heat transfer faster from the wall to the fluid. This means the heat exchanger can work more efficiently.

However, there is also a problem. When... (More)
Designing Heat Exchanger Tubes with Higher Efficiency:

Heat exchangers are used to transfer heat from one medium to another. They are important in many industrial processes, such as heating, cooling, and food processing. In the food industry, tubular heat exchangers are often used because they are simple and hygienic to process food products such as milk and juice.

This thesis focuses on a special type of tube called a corrugated tube. A corrugated tube has repeated patterns on its wall. These patterns can make the fluid mix better when it flows through the tube. Better mixing can help heat transfer faster from the wall to the fluid. This means the heat exchanger can work more efficiently.

However, there is also a problem. When the tube wall has stronger patterns, the fluid becomes harder to pump through the tube. This causes a higher pressure drop, so more pumping energy is needed. Therefore, a good tube design should improve heat transfer, while keeping the pressure drop at an acceptable level.

In this thesis, computer simulations were used to study different corrugated tube designs. These simulations can show how the fluid moves inside the tube and how heat is transferred. Instead of testing every design in a real experiment, the simulations make it possible to compare many designs in a faster and more systematic way.

Machine learning was also used in this work. After several tube designs were simulated, the results were used to train machine learning models. These models can then predict the performance of new tube designs much faster than running a new simulation every time.

The results showed that the tube shape has a clear influence on both heat transfer and pressure drop. Deeper and denser corrugation patterns can improve heat transfer, but they also increase the pressure drop. By combining simulations and machine learning, promising tube designs were selected under different pressure drop limits.

Overall, this thesis shows that computer simulations and machine learning can help engineers design better corrugated tubes for heat exchangers. This method can save time during the design process and provide useful guidance for developing more efficient heat exchangers in industrial applications. (Less)
Please use this url to cite or link to this publication:
author
Liao, Liangdou LU
supervisor
organization
course
MVKM05 20261
year
type
H2 - Master's Degree (Two Years)
subject
report number
ISRN LUTMDN/TMHP-26/5672-SE
ISSN
0282-1990
language
English
id
9237346
date added to LUP
2026-06-17 14:22:23
date last changed
2026-06-17 14:22:23
@misc{9237346,
  abstract     = {{This thesis aims to develop and evaluate a fast design and optimisation workflow for corrugated tubes used in tubular heat exchangers by combining computational fluid dynamics (CFD) simulations and machine learning. A steady-state Reynolds-averaged Navier-Stokes (RANS) model was established to simplify the simulation process and reduce computational time. Based on the selected CFD setup, a CFD database including 45 corrugation geometries and 135 simulation cases was generated using discrete Latin hypercube sampling (LHS). Gaussian process regression (GPR), feed-forward neural network (FFNN), and symbolic regression (SR) models were developed and evaluated for pressure drop and heat transfer coefficient (HTC) prediction. The FFNN model was selected for the design space sweep and Pareto front analysis. The Pareto fronts showed the trade-off between pressure drop and HTC and were used to select candidate geometries under the given pressure drop constraints. The SR model provided explicit mathematical expressions for understanding the main parameter trends. These selected geometries were then revalidated using CFD simulations to evaluate the prediction accuracy of the FFNN model. The selected designs can be regarded as candidate optimal designs within the investigated design space under the given pressure drop constraints, and can provide useful design guidance for further improvement of tubular heat exchangers.}},
  author       = {{Liao, Liangdou}},
  issn         = {{0282-1990}},
  language     = {{eng}},
  note         = {{Student Paper}},
  title        = {{Innovative Tubular Corrugation Pattern for Heat Exchangers Using CFD and Machine Learning}},
  year         = {{2026}},
}