Abstract:Numerical simulation of fluids plays an essential role in modeling many physical phenomena, which enables technological advancements, contributes to sustainable practices, and expands our understanding of various natural and engineered systems. The calculation of heat transfer in fluid flow in simple flat channels is a relatively easy task for various simulation methods. However, once the channel geometry becomes more complex, numerical simulations become a bottleneck in optimizing wall geometries. We present a combination of accurate numerical simulations of arbitrary, flat, and non-flat channels and machine learning models predicting drag coefficient and Stanton number. We show that convolutional neural networks (CNN) can accurately predict the target properties at a fraction of the time of numerical simulations. We use the CNN models in a virtual high-throughput screening approach to explore a large number of possible, randomly generated wall architectures. Data Augmentation was applied to existing geometries data to add generated new training data which have the same number of parameters of heat transfer to improve the model's generalization. The general approach is not only applicable to simple flow setups as presented here but can be extended to more complex tasks, such as multiphase or even reactive unit operations in chemical engineering.

What problem does this paper attempt to address?

This paper aims to solve the problem of heat transfer enhancement in fluid flow, especially to improve the heat transfer efficiency by modifying the channel wall structure without increasing the pressure drop. Specifically, the paper explores whether heat transfer can be increased by making structural changes to the channel wall without a corresponding increase in pressure drop. This problem is of great significance in multiple application fields, such as microfluidic devices in chemical engineering, the design of biomedical implants, and fields such as high - temperature physics and cosmology. ### Core issues of the paper 1. **Relationship between heat transfer and pressure drop**: Traditional research shows that increasing heat transfer is usually accompanied by a significant increase in pressure drop, which limits the improvement of heat transfer efficiency. The paper attempts to explore whether there are specific wall structures that can break this relationship and achieve a significant enhancement of heat transfer without a large increase in pressure drop. 2. **Optimization of complex geometries**: For simple straight channels, numerical simulation is relatively easy, but when the channel geometry becomes complex, numerical simulation becomes a bottleneck in optimizing the wall geometry. The paper proposes a method that combines accurate numerical simulation and machine - learning models to efficiently predict the drag coefficient and Stanton number under complex wall structures, thereby accelerating the optimization process. 3. **Application of machine - learning models**: The paper shows that convolutional neural networks (CNNs) can accurately predict target properties (drag coefficient and Stanton number) in a short time and are used for virtual high - throughput screening to explore a large number of randomly generated wall architectures. ### Specific methods - **Numerical simulation**: Use the direct numerical simulation (DNS) method, combined with the immersed boundary method (IBM) to generate arbitrary wall structures and study their effects on fluid flow and heat transfer. - **Machine - learning model**: Train a CNN model to predict the drag coefficient and Stanton number based on the generated data set. Improve the generalization ability of the model through data augmentation techniques. - **Performance evaluation**: Evaluate the accuracy and generalization ability of the model by comparing the prediction results of the CNN model with the real values of numerical simulation. ### Main findings - **Efficiency of the CNN model**: The trained CNN model can predict the drag coefficient and Stanton number in less than 100 milliseconds, which is much faster compared to numerical simulation (about 20 - 30 minutes per channel). - **Challenges of complex geometries**: Although the CNN model performs well in most cases, the prediction accuracy still needs to be improved for complex geometries. - **Potential applications**: This method is not only applicable to simple fluid setups but can also be extended to more complex tasks, such as multiphase flow or reaction unit operations in chemical engineering. ### Conclusion The paper proposes a method that combines numerical simulation and machine learning, which can efficiently explore and optimize the heat transfer characteristics under complex wall structures. Although the current method still has limitations in some complex geometries, it provides a new direction for future research, especially in the application in the field of chemical engineering, and is expected to be further expanded through active learning and generative models.

Machine learning for rapid discovery of laminar flow channel wall modifications that enhance heat transfer

Deep-Learning Strategy Based on Convolutional Neural Network for Wall Heat Flux Prediction

Heat Flux Prediction of Radiation Balance Wall by Deep Convolutional Neural Networks

Machine Learning Strategy for Wall Heat Flux Prediction in Aerodynamic Heating

Machine learning accelerated computational fluid dynamics

Prediction of turbulent heat transfer using convolutional neural networks

Computer Vision and Machine Learning Methods for Heat Transfer and Fluid Flow in Complex Structural Microchannels: A Review

Non-intrusive, transferable model for coupled turbulent channel-porous media flow based upon neural networks

A combined active control method of restricted nonlinear model and machine learning technology for drag reduction in turbulent channel flow

Data-driven turbulence modeling for fluid flow and heat transfer in peripheral subchannels of a rod bundle

Advancing heat transfer modeling through machine learning: A focus on forced convection with nanoparticles

Assessment of Machine Learning Wall Modeling Approaches for Large Eddy Simulation of Gas Turbine Film Cooling Flows: An a Priori Study

Transfer learning model to predict flow boiling heat transfer coefficient in mini channels with micro pin fins

Comparative study for flow condensation heat transfer in horizontal enhanced tubes based on machine learning

Machine Learning based algorithms for modeling natural convection fluid flow and heat and mass transfer in rectangular cavities filled with non-Newtonian fluids

Large eddy simulations of wall jets with coflow for the study of turbulent Prandtl number variations and data-driven modeling

Using deep learning to recognize liquid-liquid flow patterns in microchannels

ML-LBM: Predicting and Accelerating Steady State Flow Simulation in Porous Media with Convolutional Neural Networks

Machine learning-supported CFD optimization of heat transfer in a pipe with a corrugated wall shape

Innovative simulation of Al2O3 nanofluid heat transfer using advanced machine learning methods