The study addresses the issue of improving the performance of fluid film bearing models, in particular hydrostatic and hybrid bearings, both conventional and adjustable. Data-driven models typically outperform numerical ones, but also require large amount of data and time for training. Applying physically-informed neural networks (PINN) for this purpose partly reduces the required amount of data and time costs when considering fairly simple bearing designs, mainly hydrodynamic ones. However, the problem is still acute for hydrostatic/hybrid bearings, especially having active design, due to a higher number of degrees of freedom. The study proposes an approach based on utilizing a combination of separate PINNs to reduce the costs of building surrogate dynamic nonlinear bearing models while maintaining the performance advantages. A relatively simple PINN-based model of a basic plain bearing is utilized at the first stage of the modeling process to calculate the hydrodynamic lubricant pressure distribution. At the subsequent stage, this distribution is modified by another PINN-based model taking into account the lubricant pressure in hydrostatic recesses. The proposed division of models reduces the dimensionality of the training dataset and shows an advantage over the numerical model in time required for solving a set of test rotor dynamic problems by approximately an order of magnitude with little difference in accuracy. The results are useful for design procedures for high-performance rotor supports, including active supports, reducing the time required for modeling, as well as for synthesizing such models.

Microscopic images of cells often contain multiple objects with similar properties. In this study, we collected videos of blood microcirculation in nailfold capillaries to identify correlations between video data and indicators of complete blood count tests. The specific features of these videos include a high number of capillaries and a high variance in their visual quality. Therefore, we proposed a novel loss function and a video processing framework that combines metric learning and classification with uncertainty estimation. We also applied test-time augmentation to enhance the confidence of the predictions. During simulation experiments, we used shallow and deep learning models to compare a set of classifiers for the collected metadata. Both types of models demonstrated good results, with an accuracy of up to 80% in gender classification. Deep learning models outperformed shallow models and also classified the age of the volunteers with an accuracy of 74%. Subsequently, we addressed a series of binary classification tasks to identify conditionally high or low levels of the main complete blood count indicators. The obtained accuracy of the models reached 63%. Although the accuracy of the blood test classifiers is not yet sufficient for clinical purposes, we have demonstrated that video data correlates with some of the indicators.

Artificial neural networks are a robust tool for approximating spatial and temporal functions. This study introduces a novel method for hydro- and hemodynamics based on the minimization of a power loss. This loss function corresponds to the variational formulation of the boundary value problem, generalizing the Helmholtz variational principle to accommodate various mechanical boundary conditions. The method employs a global approximation of velocity distribution within the flow domain, parameterized by flow rate, allowing a single inference model to address multiple problems. The efficacy of the method was validated using 2D and 3D images, including 3D vessel images, with results compared against analytical and numerical solutions. Potential applications span a range of medical tasks, such as drug delivery, pressure distribution calculation, and vessel strength assessment.

Conventional numerical models of fluid film bearings (FFB) typically require significant amounts of computations to achieve acceptable calculation accuracy. This may cause many inconveniences, especially for computationally expensive rotor dynamics tasks. The study addresses general issues in building synthetic data-driven dynamic FFB models using machine learning methods, focusing on artificial neural networks (ANNs). In the study, surrogate models predicting journal bearing forces were built using a variety of approaches, including single- and multi-component ANN-based models, as well as so-called physics-informed neural networks (PINN). The work presents a comparative analysis of the considered approaches, primarily in the context of rotor dynamics calculations. It includes an assessment of the accuracy and performance of considered models, as well as the time spent on their creation. The results show that all tested solutions outperform numerical methods by order of magnitude or more, but it is difficult to talk about the overwhelming advantage of any one approach over others. When choosing a methodology for building surrogate FFB models, one should proceed from the requirements for the model, the benefits and shortcomings of certain methods. The work gives an idea of their relationships, including the influence of key hyperparameters of the methods on the properties of the resulting models. The results can be useful both for traditional calculations in the area of rotor dynamics, and for relatively new problems in the rotor systems, such as predictive analytics and optimal design systems, predictive controllers of active components of rotary machines.

Artificial neural networks are a powerful tool for spatial and temporal functions approximation. This study introduces a novel approach for modeling non-Newtonian fluid flows by minimizing a proposed power loss metric, which aligns with the variational formulation of boundary value problems in hydrodynamics and extends the classical Lagrange variational principle. The method is distinguished by its data-free nature, enabling problem-solving through 2D or 3D images of the flow domain. Validation was performed using both multi-layer perceptrons and U-Net architectures, with results compared against analytical and numerical benchmarks. The method demonstrated good results with a relative error of 1.41% in comparison with the analytical solution for non-Newtonian fluids. The power loss formulation offers a clear advantage by simplifying the modeling process and enhancing interpretability. Notably, the proposed method demonstrates improvements over existing techniques by providing algorithmic simplicity and universality, with applications ranging from blood flow modeling in vessels and tissues to broader hydrodynamic scenarios.

RUL (remaining useful life) estimation is one of the main functions of the predictive analytics systems for rotary machines. Data-driven models based on large amounts of multisensory measurements data are usually utilized for this purpose. The use of adjustable bearings, on the one hand, improves a machine’s performance. On the other hand, it requires considering the additional variability in the bearing parameters in order to obtain adequate RUL estimates. The present study proposes a hybrid approach to such prediction models involving the joint use of physics-based models of adjustable bearings and data-driven models for fast on-line prediction of their parameters. The approach provides a rather simple way of considering the variability of the properties caused by the control systems. It has been tested on highly loaded locomotive traction motor axle bearings for consideration and prediction of their wear and RUL. The proposed adjustable design of the bearings includes temperature control, resulting in an increase in their expected service life. The initial study of the system was implemented with a physics-based model using Archard’s law and Reynolds equation and considering load and thermal factors for wear rate calculation. The dataset generated by this model is used to train an ANN for high-speed on-line bearing RUL and wear prediction. The results show good qualitative and quantitative agreement with the statistics of operation of traction motor axle bearings. A number of recommendations for further improving the quality of predicting the parameters of active bearings are also made as a summary of the work.

Hydrodynamics of viscous fluids deals with Navier-Stokes equation - a partial differential equation with unknown distributions for velocity and pressure in a flow domain. It is difficult to find its analytical solution, especially in cases of unsteady flows, flows of non-Newtonian or rheomagnetic fluids. It is usually solved numerically using finite difference, finite element, or control volume methods. The goal of this research is application of proposed physics-based loss to rheomagnetic fluids flows modeling. The basic network architecture is U-Net. The network receives an image of the flow domain and calculates the fluid velocity distribution in a form of an image of the stream function distribution. The network was tested for the asymptotic case, the results were compared with numerical solution and known analytical solution. Proposed tool allows modeling 2D flows of rheomagnetic fluids. The proposed method is general and allows modeling 3D flows.

Systems for online prediction of remaining useful life (RUL) of technological equipment, and, in particular, fluid film bearings, are usually based on the analysis of a large amount of data received from the operated objects. In practice, formation of a data set meeting the size and quality requirements often encounters a number of difficulties. The work presents the approach to the possible overcoming of such difficulties through the use of physics-based models of degradation of fluid film bearings. The most common reasons for replacing them in rotating machines are considered as the criteria for the end of the service life. They include achievement of the wear limit in accordance with the current standard, and disruption of the bearing surfaces after reaching the material’s fatigue strength limit. The work focuses mainly on the last factor and demonstrates mathematical and numerical simulation models of rotor-bearing systems considering this phenomenon and allowing generating data on the bearing degradation process. The generated data is used to train a predictive model that estimates online the current state and RUL of the bearing. In addition, the proposed physics-based models also allow to evaluate the impact of the adjustable design of the fluid film bearings on their expected service life. The variable parameters of the adjustable bearings are also taken into account by the proposed predictive model. The work shows the results of numerical studies demonstrating the change in the service life taking in account the adjustable bearing parameters.

The theory of rheology of non-Newtonian fluids is based on the generalized Newtonian hypothesis of viscosity. The viscometers for non-Newtonian fluids should implement fluid flows with the known stress and strain state parameters distributions. Ideally, the distributions should be homogeneous in the flow domain. The idea of the proposed method is based on a combination of a capillary and a rotational viscometers implemented in the torus-shaped capillary viscometer. Analysis of the mathematical model of the inertial non-Newtonian fluid flow in the torus allowed to determine the conditions of homogeneity of the mechanical and thermal parameters in the flow domain and to develop method of viscosity measurement. The measured values are the shear rate on the inner surface of the capillary and the flow rate. The measurements are implemented with the computer vision system that processes data obtained from the high speed CMOS camera that records inertial flow in the transparent capillary illuminated with laser. The computer vision system is based on the application of deep convolutional neural network for laser speckle contrast imaging processing. During the experiments, the proposed viscometer was compared with the Brookfield rotational viscometer. The relative error of the proposed viscometer and method is less than 2%. The inertial viscometer is compact, it allows to study the wide range of shear rates per one test in automatic mode, and it has low fluid capacity of approximately 1.87 ml. That makes it possible to use the viscometer as a point on care testing device in medicine to study the rheology of physiological fluids, in particular blood.

The paper considers the use of fully connected networks for classifying the states of a rotary machine based on a vibration signal. An experimental stand is proposed. We worked with three different states of the experimental setup. The new approach is to use generative adversarial networks to create artificial data and various architectures of fully connected neural networks. We also tested different combinations of training and validation datasets. As a result, the use of all these methods makes it possible to improve the accuracy of the network by about 6.5%.

Machine learning methods offer some alternatives to the conventional approaches to the development of passive and adjustable fluid film bearings. Data-based bearing models typically show an advantage over conventional numerical models in terms of computational speed, and can either replace or supplement them in certain applications. The most promising application of machine learning is to create high-performance models and optimal controllers for fluid film bearings. It covers a range of tasks connected with the rotor trajectory planning, like active vibration and friction reduction, that is the main scope of this work. On-line rotor position assessment considering the measured or estimated loads can also be implemented using fast data-driven models in diagnostics and predictive analytics systems. The work presents an analysis of this approach in terms of the accuracy of solutions, the time required for preparing data, and training the models. The results show that the calculation speed using data-driven models can be increased at least 10 times compared to the numerical models. Two ANN-based models with different structure were analyzed in accuracy and performance. A model consisting from three separate ANNs was introduced in addition to a single-ANN model based on the analysis of the bearing forces nonlinearities and demonstrated better accuracy and the training time reduced by 26

The paper deals with the application of deep learning methods to rotating machines fault diagnosis. The main challenge is to design a fault diagnosis system connected with multisensory measurement system that will be sensitive and accurate enough in detecting weak changes in rotating machines. The experimental part of the research presents the test rig and results of high-speed multisensory measurements. Six states of a rotating machine, including a normal one and five states with loosened mounting bolts and small unbalancing of the shaft, are under study. The application of deep network architectures including multilayer perceptron, convolutional neural networks, residual networks, autoencoders and their combination was estimated. The deep learning methods allowed to identify the most informative sensors, then solve the anomaly detection and the multiclass classification problems. An autoencoder based on ResNet architecture demonstrated the best result in anomaly detection. The accuracy of the proposed network is up to 100% while the accuracy of an expert is up to 65%. A one-dimensional convolutional neural network combined with a multilayer perceptron that contains a pretrained encoder demonstrated the best result in multiclass classification. The detailed fault detection accuracy with the determination of the specific fault is 83.3%. The combinations of known deep network architectures and application of the proposed approach of pretraining of the encoders together with using a block of inputs for one prediction demonstrated high efficiency.

The variational approach of finding the extremum of an objective functional is an alternative approach to the solution of partial differential equations in mechanics of continua. The great challenge in the calculus of variations direct methods is to find a set of functions that will be able to approximate the solution accurately enough. Artificial neural networks are a powerful tool for approximation, and the physics-based functional can be the natural loss for a machine learning method. In this paper, we focus on the loss that may take non-linear fluid properties and mass forces into account. We modified the energy-based variational principle and determined the constraints on its unknown functions that implement boundary conditions. We explored artificial neural networks as an option for loss minimization and the approximation of the unknown functions. We compared the obtained results with the known solutions. The proposed method allows modeling non-Newtonian fluids flow including blood, synthetic oils, paints, plastic, bulk materials, and even rheomagnetic fluids. The fluids flow velocity approximation error was up to 4% in comparison with the analytical and numerical solutions.

The aim of this work is to develop practical tools to recognize the average flow rate of physiological fluids in capillaries. This tool is represented by classification models in an artificial neural networks form. The flow rate data were obtained experimentally. Intralipid was used as the test liquid. Laser speckle contrast imaging was used to obtain images of liquid flow in a glass capillary. The experiment was carried out with an average flow rate of 0-2 mm/s with various concentrations of intralipid. The results of training of fully connected and convolutional neural networks for processing the received data are presented. The accuracy of determining the average flow rate of intralipid with different concentrations was comparable to the previously obtained results for a fixed concentration and amounted to approximately 97.5%.

The laser speckle contrast imaging allows the determination of the flow motion in a sequence of images. The aim of this study is to combine the speckle contrast imaging and machine learning methods to recognition of physiological fluids flow rate. Data on the flow of intralipid with average flow rate of 0-2 mm/s in a glass capillary were obtained using a developed experimental setup. These data were used to train a feed-forward artificial neural network. The accuracy of random image recognition was quite low due to pulsations and the uneven flow set by the pump. To increase the recognition accuracy, various methods for calculating speckle contrast were used. The best result was obtained when calculating the mean spatial speckle contrast. The application of the mean spatial speckle contrast imaging together with the proposed artificial neural network allowed to increase the fluid flow rate recognition accuracy from about 65 % to 89 % and make it possible to exclude an expert from the data processing.

The article considers general approaches and modern monitoring systems for rotary machines of electric generating equipment. The main characteristics of monitoring and diagnostics systems of Russian and foreign manufacturers are presented. Modern trends in the construction of intelligent systems for analyzing the performance of turbo generators and predicting possible failures in order to minimize the cost of repairs and forced shutdown of equipment are outlined. The concept of adaptive-predictive use of rotary machines, the difference from existing systems is the presence of adaptive module that allows to react to unwanted changes in real time and increase the predicted residual resource or eliminate the predicted probability of initially refusal.

The goal of this work is association of several machine learning methods in a study of rotating machines with fluid-film bearings. A fitting method is applied to fit a non-linear reaction force in a bearing and solve a rotor dynamics problem. The solution in the form of a simulation model of a rotor machine has become a part of a control system based on reinforcement learning and the policy gradient method. Experimental part of the paper deals with a pattern recognition and fault diagnosis problem. All the methods are effective and accurate enough.