Heat transfer between two porous parallel plates of steady nano fludis with Brownian and Thermophoretic effects: A new stochastic numerical approach
Heat transfer between two porous parallel plates of steady nano fludis with Brownian and Thermophoretic effects: A new stochastic numerical approach
复制标题
DOI：
10.1016/j.icheatmasstransfer.2021.105436
复制
发表时间：
2021-06-25
影响因子：
7
通讯作者：
Shoaib, Muhammad
中科院分区：
文献类型：
--
作者：
Khan, Rafaqat Ali;Ullah, Hakeem;Shoaib, Muhammad
提出了通过使用 LevenbergMarquard 系统（BNN-LMS）的反向传播神经网络进行集成数值计算的设计，以探索在热泳和布朗运动的刺激下控制两个稳态纳米流体多孔平行板（HTPSNF）之间传热系统的流体力学问题。通过引入相似变换，将原始偏微分方程系统模型HTPSNF转换为非线性常微分方程。同伦强度分析法（HAM）利用原始模型HTPSNF的控制方程来获得数据集。建议的 BNNLMS 方案的参考集合源自与 HTPSNF 相关的各种场景，例如孔隙度参数、施密特数、布朗参数、粘度参数、普兰德数和热参数。为了维护建议的 BNN-LMS 的真实性，完成了 BNN-LMS 的验证、训练和测试过程，以控制各种情况下的 HTPSNF 估计解决方案并使用参考结果进行评估。基于 MSE 结果、误差直方图、相关性和回归的比较研究和性能分析表明了所设计的 LMBNN 技术的有效性和优点。 10-07 至 10-14 范围内的均方误差证实了所提出的方法的完美性，以确保建议结果和参考结果之间的紧密对应。
The design of integrated numerical computing through back-propagated neural networks with LevenbergMarquard system (BNN-LMS) is presented to explore the fluid mechanics problems governing the system of heat transfer between two porous parallel plates of steady nanofluids (HTPSNF) under the stimulus of thermophoretic and Brownian motion. By introducing the similarity transformations, the original system model HTPSNF in terms of PDEs is converted to nonlinear ODEs. Strength of Homotopy Analysis Method (HAM) is utilized the governing equations of original model HTPSNF to obtain the data set. Reference collection for the suggest BNNLMS scheme is originated in terms of various scenarios associated HTPSNF such as Porosity parameter, Schmidt number, Brownian parameter, viscosity parameter, Prandlt number and thermophoric parameter. To uphold the trueness of the suggest BNN-LMS, the validation, training and testing process of BNN-LMS are accomplished to govern the estimate solution of HTPSNF for various cases and evaluation with reference results. The comparative studies and performance analyses based on outcomes of MSE, error histograms, correlation and regression intimate the effectiveness and virtue of designed LMBNN technique. Mean Square Errors in the ranges of 10-07 to 10-14 confirm the perfection of the presented methodology for the closed correspondence between suggested and reference results.