基于卷积神经网络的几何自适应稳态传热数据驱动建模:对流换热
Data-Driven Modeling of Geometry-Adaptive Steady Heat Transfer based on Convolutional Neural Networks: Heat Convection
摘要:
在本文中,基于神经网络,我们开发了一个数据驱动的模型,用于在二维空间中具有任意复杂几何形状的热物体的稳态热对流的快速预测。根据控制方程,稳态热对流是由对流和热扩散项,因此,物理场的分布将表现出相邻点之间的更强的相关性。因此,所提出的神经网络模型使用卷积神经网络(CNN)层作为编码器,并且使用解卷积神经网络(DCNN)层作为解码器。与完全连接(FC)网络模型相比,基于CNN的模型更有利于捕捉和重构低秩特征空间的空间关系,如边相交、平行性和对称性。此外,我们应用符号距离函数(SDF)作为网络输入来表示问题几何,与二进制图像相比,它包含更多的信息。为了显示所提出的网络模型的强大学习和泛化能力,训练数据集只包含具有简单几何形状的热对象:三角形,四边形,五边形,六边形和十二边形;而验证案例使用任意和复杂的几何形状。研究表明,训练后的网络模型能够准确预测网络模型在训练过程中从未见过的复杂几何问题的速度场和温度场,预测速度比CFD快四个数量级。该网络模型预测精度高、预测速度快,预示着这类模型在实时控制、优化设计等领域的应用前景。
英文摘要:
In this paper, based on neural networks, we develop a data-driven model for extremely fast prediction of steady-state heat convection of a hot object with arbitrary complex geometry in a two-dimensional space. According to the governing equations, the steady-state heat convection is dominated by convection and thermal diffusion terms, thus the distribution of the physical fields would exhibit stronger correlations between adjacent points. Therefore, the proposed neural network model uses Convolutional Neural Network (CNN) layers as the encoder and Deconvolutional Neural Network (DCNN) layers as the decoder. Compared with fully connected (FC) network model, the CNN based model is good for capturing and reconstructing the spatial relationships of low-rank feature spaces, such as edge intersections, parallelism and symmetry. Furthermore, we applied the signed distance function (SDF) as the network input for representing the problem geometry, which contains more information compared to a binary image. For displaying the strong learning and generalization ability of the proposed network model, the training dataset only contains hot objects with simple geometries: triangles, quadrilaterals, pentagons, hexagons and dodecagons; while the validating cases use arbitrary and complex geometries. According to the study, the trained network model can accurately predict the velocity and temperature field of the problems with complex geometries which has never been seen by the network model during the model training; and the prediction speed is four orders faster than the CFD. The ability of the accurate and extremely faster prediction of the network model suggests the potentials of applying such kind of models to the applications of real-time control, optimization, and design in future.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。