Joint estimation of thermal and mass diffusivities of a solute-solvent system using ANN-GA based inverse framework
Joint estimation of thermal and mass diffusivities of a solute-solvent system using ANN-GA based inverse framework
复制标题
DOI：
10.1016/j.ijthermalsci.2017.09.008
复制
发表时间：
2018
影响因子：
4.5
通讯作者：
Samarjeet Chanda;K. Muralidhar;Yogesh M. Nimdeo
中科院分区：
文献类型：
--
作者：
Samarjeet Chanda;K. Muralidhar;Yogesh M. Nimdeo
这项工作发展了一种反方法来联合估计溶质-溶剂系统的热扩散率和质量扩散率。讨论了一种结合进化模拟和优化算法——人工神经网络（ANN）和遗传算法（GA）的逆参数估计框架。直接问题首先通过假设热扩散率和质量扩散率为常数，然后作为温度和浓度的函数来解决。利用规定的扩散系数值对直接数值模拟得到的数据进行高斯、均匀和Logistic三种概率分布采样，并将噪声撒在数据上，生成合成实验数据。然后将这些数据输入到逆框架中，以估计感兴趣的参数并系统地量化相关的不确定性。在扩散系数同时依赖于温度和浓度的情况下，还联合检索了高阶参数，并量化了相关的不确定性。进行了三组实验：外加温差引起的水中热扩散实验、盐溶液在淡水中的等温扩散实验和盐溶液在淡水中的扩散及同时逆流热扩散实验。在单色激光源（氦氖激光，λ = 632.8 nm）照射下，利用马赫-曾德尔干涉仪以干涉图的形式捕获了随时间变化的温度场和浓度场。由此获得的温度和浓度数据可用于利用逆框架估计溶质-溶剂体系的热扩散率和质量扩散率。将所得结果与文献进行比较，验证了所提出的参数估计方法。
This work develops an inverse methodology to jointly estimate the thermal and mass diffusivities of a solute-solvent system. An inverse parameter estimation framework using a combination of evolutionary simulation and optimization algorithms namely Artificial Neural Network (ANN) and Genetic Algorithm (GA) is discussed. The direct problem is first solved by assuming thermal and mass diffusivities as constants and then as functions of temperature and concentration. Synthetic experimental data is generated by sprinkling noise sampled from three probability distributions namely Gaussian, Uniform and Logistic on the data obtained from direct numerical simulations carried out using prescribed values of diffusivities. This data is then fed into the inverse framework to estimate the parameters of interest and systematically quantify the associated uncertainties. In case of diffusivities dependent on temperature and concentration both, the higher order parameters are also jointly retrieved and associated uncertainties are quantified. Three sets of experiments namely diffusion of heat in water due to applied temperature difference, isothermal diffusion of salt solution in fresh water and diffusion of salt solution in fresh water along with simultaneous counter current heat diffusion are performed. The time evolving temperature and concentration fields are captured in the form of an interferogram using a Mach-Zehnder interferometer illuminated by a monochromatic laser source (Helium-Neon laser, λ = 632.8 nm). The temperature and concentration data thus obtained is used to estimate the thermal and mass diffusivities of the solute-solvent system using the inverse framework. The results obtained are compared against literature to validate the proposed parameter estimation methodology.