Моделирование течения неньютоновских жидкостей в тороидальном канале инерционного вискозиметра с системой технического зрения

Purpose of research. Development of theoretical premises for the new inertial viscometer, as well as the development of an approximate model of the viscosity fluid flow using convolutional neural networks and laser speckle contrast imaging data.Methods. The study consists of two parts. The first is devoted to a theoretical study of viscosity fluid flow in the toroidal channel of еру new inertial viscometer. The mathematical model of the flow includes the dimensionless equations of Navier-Stokes and convective heat conduction, the analysis of which made it possible to estimate the conditions for the uniformity of pressure and temperature fields. The numerical solution of the simplified Navier-Stokes equation was obtained by the control volume method. The computational experiment made it possible to identify additional operating conditions for the viscometer. The second part of the research is aimed at solving the problem of predicting the values of the shear strain rate on the tour surface and the flow rate. The approximate flow model is based on an ensemble of convolutional neural networks trained on data from laser speckle-contrast visualization of a fluid flow in a transparent tube.Results. The recommendations on the operating parameters of the inertial viscometer for the studied types of liquids in a given viscosity range are obtained. An approximate model has been developed in the form of an ensemble of deep neural networks, which makes it possible to determine the volumetric flow rate and the shear strain rate on the flow surface based on fluid flow images.Conclusion. The approximate Navier-Stokes equation obtained as a result of theoretical analysis for the flow of a viscous fluid in a toroidal channel can be used to numerical determination the kinematic viscosity. So, the necessary flow characteristics, such as volumetric flow rate and shear strain rate on the flow surface, can be found using the developed and pretrained convolutional neural network based on laser speck contrast imaging data. The test fluid can be any non-Newtonian fluid capable of reflecting coherent radiation. In particular, it can be physiological fluids, including blood.