Abstract:Reconstruction and fast prediction of flow fields are important for the improvement of data center operations and energy savings. In this study, an artificial neural network (ANN) and variational autoencoder (VAE) composite model is proposed for the reconstruction and prediction of 3D flowfields with high accuracy and efficiency. The VAE model is trained to extract features of the problem and to realize 3D physical field reconstruction. The ANN is employed to achieve the constructability of the extracted features. A dataset of steady temperature/velocity fields is acquired by computational fluid dynamics and heat transfer (CFD/HT) and fed to train the deep learning model. The proposed ANN-VAE model is experimentally proven to achieve promising field prediction accuracy with a significantly reduced computational cost. Compared to the CFD/HT method, the ANN-VAE method speeds up the physical field prediction by approximately 380,000 times, with mean accuracies of 97.3% for temperature field prediction and 97.9% for velocity field prediction, making it feasible for real-time physical field acquisition.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: how to improve the operating efficiency and achieve energy conservation in data centers by constructing and rapidly predicting three - dimensional flow fields. Specifically, the research aims to develop an efficient and accurate method to replace the traditional computational fluid dynamics and heat transfer (CFD/HT) simulation methods, thereby achieving real - time reconstruction and prediction of physical fields (such as temperature fields and velocity fields). ### Problem Background 1. **Energy Consumption Problem in Data Centers** - Data centers are increasing rapidly around the world, resulting in a significant increase in power consumption. - Environmental systems (such as computer room air - conditioning systems, CRAC) consume about 40% of the total power in data centers, so optimizing their operation is crucial. 2. **Limitations of Traditional Methods** - Although CFD/HT simulations are accurate, the computational cost is extremely high, making it difficult to achieve real - time prediction. - Other simplified methods (such as POD - based methods) improve the speed but have limited accuracy. ### Solution To overcome the above problems, the author proposes a composite model (ANN - VAE) that combines artificial neural networks (ANN) and variational auto - encoders (VAE) for efficiently and accurately reconstructing and predicting three - dimensional flow fields. The main objectives of this model include: 1. **High - Precision Reconstruction and Prediction** - VAE is responsible for extracting features from a large amount of data and achieving the reconstruction of 3D physical fields. - ANN is responsible for mapping the input operating parameters to the latent space recognizable by VAE, thereby achieving rapid prediction of the flow field. 2. **Significantly Reducing Computational Costs** - Compared with the traditional CFD/HT method, the ANN - VAE method can increase the physical field prediction speed by about 380,000 times while maintaining a relatively high prediction accuracy (the average prediction accuracy of the temperature field is 97.3%, and the average prediction accuracy of the velocity field is 97.9%). ### Method Overview - **Data Acquisition**: Generate a data set containing steady - state temperature and velocity fields through CFD/HT simulations. - **Model Training** - First, use AE for feature extraction to determine the optimal latent space dimension. - Then introduce the variational principle to construct VAE and optimize the latent space distribution. - Finally, train ANN to map operating parameters to the latent space and combine with the VAE decoder to achieve rapid prediction. ### Conclusion This research shows the application potential of the ANN - VAE model in the data center cooling cycle. It not only achieves efficient flow field reconstruction and prediction but also greatly reduces computational costs, providing a new approach for real - time monitoring and optimizing data centers.

Reconstruction and fast prediction of a 3D flow field based on a variational autoencoder

Reconstruction and Fast Prediction of 3D Heat and Mass Transfer Based on a Variational Autoencoder

Three-dimensional autoencoder for the flow field reconstruction of an inclined circular disk

Flow field prediction of supercritical airfoils via variational autoencoder based deep learning framework

Predicting Pressure Fields from Incomplete Velocity Fields Based on Deep Convolutional Neural Network

Flow Field Reconstruction of 2D Hypersonic Inlets Based on a Variational Autoencoder

Learning-Based High-Performance Algorithm for Long-Term Motion Prediction of Fluid Flows

Simulation and Prediction of Three-Dimensional Rotating Flows Based on Convolutional Neural Networks

Fast Flow Field Prediction Based on E(2)-equivariant Steerable Convolutional Neural Networks

Flowfield Reconstruction Method Using Artificial Neural Network

Flow Reconstruction and Prediction Based on Small Particle Image Velocimetry Experimental Datasets with Convolutional Neural Networks

Fast simulation of airfoil flow field via deep neural network

Real-time Physical Field Reconstruction for Nanofluids Convection Using Deep Learning with Auxiliary Tasks

Unsteady Flow-Field Forecasting Leveraging a Hybrid Deep-Learning Architecture

Flowfield Prediction of Airfoil Off-Design Conditions Based on a Modified Variational Autoencoder

Rapid prediction of indoor airflow field using operator neural network with small dataset

PFVAE: A Planar Flow-Based Variational Auto-Encoder Prediction Model for Time Series Data

Deep Learning-Based Rapid Flow Field Reconstruction Model with Limited Monitoring Point Information

Applications of autoencoder in reducedGorder modeling of flow field

Reconstruction of Two-Dimensional to Three-Dimensional Flow Transition Fields Using Neural Network-Based Generative Adversarial Networks