A transfer learning method to assimilate numerical data with experimental data for effusion cooling

Article Thermodynamics

Optimization of the semi-sphere vortex generator for film cooling using generative adversarial network

Yaning Wang et al.

Summary: This study developed a new model and optimization framework using deep learning and machine learning techniques, which improved the film cooling effectiveness in protecting the high-pressure turbine from melting down.

INTERNATIONAL JOURNAL OF HEAT AND MASS TRANSFER (2022)

添加到收藏夹

Article Mechanics

A data assimilation model for wall pressure-driven mean flow reconstruction

Sen Li et al.

Summary: This study establishes a continuous adjoint data assimilation model to reproduce global turbulent mean flow using limited wall pressure measurements. The model corrects the error induced by the Boussinesq assumption and isolates crucial information from the Reynolds stress using the Stokes-Helmholtz decomposition. The CADA model minimizes discrepancies between wall pressure measurements and numerical predictions, revealing the optimal anisotropic contribution of the Reynolds force vector.

PHYSICS OF FLUIDS (2022)

添加到收藏夹

Article Thermodynamics

Transonic vane film cooling with crescent-shaped craters using an endoscopic pressure-sensitive paint technique

Kechen Wang et al.

Summary: This study examines the cooling performance of a crescent-shaped crater (CSC) hole design on the pressure side of a transonic vane. The results show that the CSC hole design is more effective in cooling than the circular hole design, and it also reduces aerodynamic losses.

APPLIED THERMAL ENGINEERING (2022)

添加到收藏夹

Article Energy & Fuels

Effect of Blockage Inside Holes on Film Cooling Performance on the Suction Side of a Turbine Guide Vane

Wei Zhang et al.

Summary: The effect of thermal barrier coating deposition inside the film holes of turbine guide vanes on film cooling performance has been studied. The film effectiveness on the suction side was measured using infrared thermal imaging technology. The results show that film effectiveness decreases and the spanwise inhomogeneity becomes evident with an increase of the blockage ratio. The adhesion of the film is improved near the hole exit at a small blockage ratio of 0.2 due to the rounded corners formed by blockage.

ENERGIES (2022)

添加到收藏夹

Article Thermodynamics

Study of an effusion-cooled plate with high level of upstream fluctuation

X. Chen et al.

Summary: The study investigated the flow field and surface adiabatic coolant-film effectiveness distribution of a combustor representative effusion cooling array with cylindrical cooling holes. The focus was on the influence of high inflow turbulence, which is often neglected in literature. Both experimental and numerical methods were used to examine the effects of inflow turbulence on the cooling performance.

APPLIED THERMAL ENGINEERING (2021)

添加到收藏夹

Article Thermodynamics

Establishment of a generalizable model on a small-scale dataset to predict the surface pressure distribution of gas turbine blades

Qi Wang et al.

Summary: By introducing the concept of transfer learning, this study successfully established a model with generalizability and accuracy on a small-scale dataset to predict the surface pressure distribution of a turbine rotor cascade with varying geometries and boundary conditions cost-effectively. The model transferred from a higher-fidelity dataset showed better generalization performance, reducing the root mean square error and modeling cost significantly.

ENERGY (2021)

添加到收藏夹

Article Thermodynamics

Two-dimensional film-cooling effectiveness prediction based on deconvolution neural network

Yaning Wang et al.

Summary: A theoretical model using Deconvolutional Neural Network (Deconv NN) was developed to establish the correlation between coolant jet parameters and surface temperature distribution, achieving better performance and robustness compared to existing methods.

INTERNATIONAL COMMUNICATIONS IN HEAT AND MASS TRANSFER (2021)

添加到收藏夹

Article Engineering, Civil

A transfer Learning-Based LSTM strategy for imputing Large-Scale consecutive missing data and its application in a water quality prediction system

Zeng Chen et al.

Summary: Water quality monitoring is critical for improving water resource protection and management, yet data missing is a common issue in this field. This paper introduces a novel algorithm named TrAdaBoost-LSTM to address large-scale consecutive missing data problems effectively, resulting in improved imputation accuracy.

JOURNAL OF HYDROLOGY (2021)

添加到收藏夹

Article Thermodynamics

Searching for irregular pin-fin shapes for high temperature applications using deep learning methods

Li Yang et al.

Summary: This study proposed a new optimization approach for pin-fins using deep learning techniques and Genetic Algorithms, generating optimized geometries and providing general suggestions for pin-fin designs through machine learning methods.

INTERNATIONAL JOURNAL OF THERMAL SCIENCES (2021)

添加到收藏夹

Article Engineering, Mechanical

Data assimilation for turbulent mean flow and scalar fields with anisotropic formulation

Chuangxin He et al.

Summary: This work presents an anisotropic formulation for data assimilation of turbulent flows using the continuous adjoint system, which allows for the reconstruction of flow fields and heat transfer characteristics from a limited number of observations through optimized forcing terms.

EXPERIMENTS IN FLUIDS (2021)

添加到收藏夹

Proceedings Paper Computer Science, Hardware & Architecture

SURFNet: Super-resolution of Turbulent Flows with Transfer Learning using Small Datasets

Octavi Obiols-Sales et al.

Summary: This paper introduces SURFNet, a transfer learning-based super-resolution flow network that accelerates the traditional numerical solver by training the DL model on low-resolution datasets and transferring the model to high-resolution flow problems. Empirical evaluations demonstrate a 2 - 2.1x speedup over the traditional physics solver across various resolutions, showcasing resolution-invariance and generalization capabilities. Compared to a baseline model, SURFNet achieves the same performance gain while significantly reducing data collection and training time.

30TH INTERNATIONAL CONFERENCE ON PARALLEL ARCHITECTURES AND COMPILATION TECHNIQUES (PACT 2021) (2021)

添加到收藏夹

Article Thermodynamics