Cyber-Physical Systems and Control II

The advantages of using neural network models as digital twins of objects with distributed parameters are shown. The analysis of using physically-informed neural networks to solve direct and inverse boundary value problems is presented. It is proposed to use radial basis function neural networks (RBFNNs) as physically-informed neural networks, which have a simple structure and the ability to adjust the non-linear parameters of the basis functions. The approach to the solving of the coefficient inverse boundary value problems, which allows determining the unknown function describing the physical environment, is proposed. The algorithm uses a unified approach to solve direct and inverse problems on RBFNN. RBFNN training is proposed to be performed using the fast algorithm of the Levenberg-Marquardt method developed by the authors. The analytical calculation of the Jacobi matrix in the Levenberg-Marquardt method is performed. An example of the solving of the inverse coefficient problem for a piecewise homogeneous medium is given. To solve the direct problem for a piecewise homogeneous medium we have used the algorithm developed by the authors which is based on solving separate problems for each region with different properties of the medium and on using the general error functional taking into account errors on the border of regions.

One of the methods of effective diagnosis of complex systems is considered in the article. The key feature of the suggested method is that the identification of the failure model is conducted in conditions of insufficient information using small ordered samples. When applying the methods of the theory of stochastic indication, the features of a limited volume and rapid aging of information are taken into account. Expansion of the range of possibilities of such a method is achieved by an operator algebraic or integro-differential transformation of the initial information, and the research is carried out on small samples presented in the form of a variational series. As a result, it is guaranteed to find a way to convert data to obtain statistics independent on the parameters of the parent distribution. If analytical construction is not possible, the distribution function of such statistics can be determined as a result of statistical modeling.

This work considers the problem of building decentralized resilient cyber-physical systems that are able to operate in open areas. Operating in an open dynamic environment causes increased risks of component failure and connectivity loss due to destructive impacts from this environment. This paper describes the known methods and tools that may be applied to build such systems and identifies their features and limitations. Classifications of threat sources, threats, incidents, methods and means are given, and an approach of integration of methods, and means to neutralize the threats is proposed. The presented requirements and models may be used to determine the probabilities of risk events for a certain set of implemented methods and means or to solve the problem of selecting methods and means that minimize the risk of certain incidents. For this purpose, various risk analysis models may be applied. In addition, general recommendations and suggestions that should be used to build cyber-physical systems that have the ability of dynamic connection and disconnection of their components are given.

An original approach to the development of mathematical tools of the reliability theory and expanding the range of its capabilities by using the properties of the stochastic theory of similarity are proposed and justified in the article. Such an approach to creating mathematical “symbiosis” of two large theories is aimed at improving the efficiency of solving the problem of identification of the quality indicators (reliability) of complex technical systems during its life cycles: from creation, active usage, and to subsequent disposal. The stochastic similarity is based on the elementary lemma and a metric in the form of a ratio of distribution functions of a random variable. The elementary lemma is also the basis of the probability integral transformation, which is used in the modeling of random variables. Useful properties of the lemma and the integral transformation are given in the study dedicated to the capabilities of the stochastic theory of similarity. In the course of assessing the applicability of the stochastic theory of similarity to the problems of comparing reliability indicators, two alternative methods of solution are presented: the first method is based on the construction of models using the principle of maximum entropy, and the second method is using stochastic similarity.

This paper examines the effectiveness of a recent method based on Kreiss’s constant minimization for designing output-feedback compensators to control instabilities in a laminar plane Poiseuille flow. The flow dynamics are highly non-normal, subsequently, small disturbances may be substantially amplified and reach large transient values that induce nonlinearities and lead to turbulence, even though such perturbations should eventually decay in a linear flow model. The maximum transient energy growth and the Kreiss constant both convey information regarding this transient behavior. A Kreiss-constant-minimizing controller is obtained and its effectiveness for the plane Poiseuille flow control problem is demonstrated.

Modifications of the well-known SHAP method for the local explanation of black-box classification models are proposed. It is supposed that the classifier prediction is represented in the form of a class probability distribution. A key idea behind the modifications is to replace the difference of characteristic functions in the framework of Shapley values with non-symmetric divergence measures between the predicted class probability distribution with a feature and without it. Such measures are used for estimating the feature contribution, for example, the Kullback-Leibler divergence and Chi-squared-divergence. For comparison purposes, we also study the symmetric Hellinger distance measure. A lot of numerical experiments on synthetic and real datasets illustrating the proposed modifications are provided and analyzed.

Two approaches for developing robust models of the distance metric learning under condition of interval-valued training data are proposed. Both approaches are based on a probabilistic representation of the feature intervals and on solving the minimax optimization problem which is stated for implementing robust decisions. The first approach is based on using the dual form representation of the linear maximization programming problem with probability distributions as optimization variables for computing optimal values of features within intervals. The second approach is based on considering extreme points of a polyhedron produced by the probability distribution constraints. The approaches can be regarded as a framework for developing a set of robust distance metric learning models which can be constructed on the basis of the Mahalanobis distance (linear transformation) as well as of Siamese neural networks (non-linear transformation).

This paper discusses joint measurement results reconciliation procedures of dependent quantities performed in cyber-physical systems. A method is proposed for approximate errors calculation for reconciliation procedures’ results and estimation of the potential accuracy increase due to the reconciliation procedure. Useful expressions were derived for error estimates of the reconciled results obtained within dependencies models formalized in the form of a linear or nonlinear equation or a system of such equations. The local linear approximation of the measurands dependencies was used to obtain these expressions. It was also analyzed how this nonlinearity neglecting affects the error estimation accuracy in such cases.

A cognitive system for measuring the magnetic characteristics of high-coercive permanent magnets made of rare-earth metals such as neodymium-iron-boron and samarium-cobalt is proposed. The cognitive system includes object-oriented programming, systems for simulating electrical circuits and multiphysical models, and allows you to restore the magnetic characteristic from the measured pulses of the magnetizing current in the magnetizing inductor and the voltage applied to the latter. The essence of the measurement method is as follows: the mathematical model is based on the data of the measured supply voltage and the characteristics of the magnetizing installation and inductor; the measured curve of the magnetizing current pulse is fed into the mathematical model and compared with the modeled current pulse in the comparison device, having previously passed the stage of the wavelet transform; further, if the difference in the current characteristics does not satisfy the accuracy of a measurement, then alterations are made to the mathematical model by changing the coefficients of the wavelet transform. Thus, the cognitive system determines the magnetic characteristics of a permanent high-coercive magnet. The method was developed based on full-scale model tests using a multiphysical computer model and wavelet transform.

Technological solutions used in the design of sensors to reduce the influence of uninformative external influences have led to the need to consider sensor parametric transducers as multidimensional measurement objects, and therefore, to create measurement circuits for sensors with a multicomponent equivalent circuit for their parametric transducers. The paper considers variants of multicomponent equivalent circuits of capacitive sensors parametric transducers and mentions methods for converting equivalent circuits parameters. It is proposed to use phase division multiplexing in measuring circuits of multidimensional objects to determine the parameters of the sensor equivalent circuit. This method eliminates the error of converting passive electrical quantities into active ones. Approaches to determining the parameters of multidimensional objects using the method of phase division multiplexing based on the allocation of the active transducer output voltage harmonic components are considered. The implementation of the phase division multiplexing method in the microprocessor measuring circuit of a capacitive pressure sensor with temperature error correction is given. It is proposed to use piecewise continuous test signals to correct the error caused by imperfect sampling. The research was carried out using the methods of the classical electrical circuits theory and the error theory.

The paper presents a digital twin of a local heating station with a heat consumption loop of a building. It contains a model of the object of control in a PLC which is utilized to stabilize the temperature inside the building. The work discusses the development of the application for process control based on the model loaded into a PLC. The distinguishing feature of the approach is based on code generation from a mathematical model of the control object created in MATLAB Simulink. An additional goal for the control system of the heating station deals with environmental conditions, namely, outdoor temperature. The model of a water-jet unit is built, simulated data are presented, and discussed, the control system is investigated, and the analysis of the data is carried out. The model is checked based on data for the city of St. Petersburg, Russia as an example. Implementation of the approach leads to overcoming risks and disadvantages of reactive organization of heating housing services, the main of which are downtime of the heating supply process and life quality reduction for energy end-users.

This paper shows a way to model moving swarms of fish in water bodies and detect them using virtual sensors. Real waters (e.g. lakes) can be explored and analyzed with sonar sensors based on sound waves. For automatic Artificial Intelligence (AI)-based detection of moving objects, such as fish, in sensor data, there is often a lack of training data. Through virtual sensors, automatic labeling, and a representation of fish swarms in 3D worlds by particle systems, we have developed a solution to generate a variety of synthetic data and enable AI training for this specific use case. Both the virtual environments and the representation and animation of the fish swarms can be freely parameterized.

In this paper, pelage pattern matching is considered to solve the individual re-identification of the Saimaa ringed seals. Animal reidentification, together with the access to a large amount of image material through camera traps and crowd-sourcing, provides novel possibilities for animal monitoring and conservation. Image retrieval techniques, such as global pooling, can be used to solve the individual re-identification. However, current global pooling methods incorporate only value distribution of features, losing spatial information. To overcome the problem, we propose a novel pooling approach that allows aggregating the local pattern features to get a fixed size embedding vector that incorporates global features by taking into account their spatial distribution. This is obtained by eigen decomposition of covariances computed for probability mass functions representing feature maps. Embedding vectors can then be used to find the best match in the database of known individuals allowing animal re-identification. The results show that the proposed pooling technique outperforms the existing methods on the challenging Saimaa ringed seal image data.

An artificial intelligence system is primarily a system that intensively uses various software. At the same time, the presence of machine learning libraries and ready-made open solutions sometimes gives the impression that it is very easy to implement software systems with artificial intelligence support. However, the development of viable and reliable artificial intelligence systems that are deployed in the field and can expand and develop over many years requires both serious planning and the constant allocation of significant costly resources. On the other hand, the threshold for entering this industry increases significantly, because, in addition to the highly qualified team members, the cost of the final solution and the resources that you need for AI will also rapidly increase. We believe that the principles of software engineering are applicable to artificial intelligence. In this paper, we consider various practices in the field of software development and related disciplines, which, we believe, should be used in the construction of artificial intelligence systems.

The paper is devoted to the problem of vehicles license plates detection and recognition in real time based on machine learning methods. Proposed by the authors is the software system which combines a pseudo-labeling method for automated preparation of the dataset and two separate convolutional neural networks for license plates detection and symbols recognition, where the YoloV3 convolutional neural network is used for detection and a convolutional recurrent neural network is used for symbols recognition. The main feature of the proposed system is the semi-supervised learning method which allows obtaining a huge dataset for neural networks training containing 95,000 images with marked coordinates of license plates. Integration of the mentioned approaches into a single working flow proves its efficiency due to significant reduction of efforts on dataset preparation and high accuracy of the detection and recognition: the implemented system recognizes license plates with 94,8% accuracy on a test sample of 20,000 images.

Smoke detectors are used to warn people of hazards due to fire and smoke. When smoke is detected, an acoustic alarm signal is emitted. False alarms, which can be caused by component failures, are extremely undesirable, depending on the field of application. Therefore, this paper presents a concept developed according to the requirements of the generic safety standard IEC 61508:2010. The aim of this development is to avoid false alarms due to component failures and additionally to increase the availability of the safety function (smoke detection). This concept includes a fault detection system that can detect all random hardware failures and assign them to the channel in which they occur. This channel can then be deactivated and the remaining channel is able to keep the safety function alive until the system is repaired or replaced. Besides an introduction to the normative basics, the advantages of the newly developed safety architecture are described. Based on a comparison of the safety characteristics Average Probability of Dangerous Failure per Hour and Average Probability of Dangerous Failure on Demand, it is argued why the newly developed circuit has an advantage compared to classic structures. The long-term goal is to develop a smoke detector that meets all the requirements of IEC 61508:2010 (including the necessary measures for fault avoidance for hardware and software development). The results may be transferred to other control systems in general, which require a combination of high reliability and high availability at the same time.

Creating a database of high quality is a considerable challenge when building Predictive Quality models for multi-stage production processes. These models aim to make predictions regarding the occurrence of defects based on acquired production data, therefore enabling the user to reduce scrap at an early stage and optimize process parameters. The performance of these models strongly depends on the quality and thus the completeness of the data acquired, which is often not known to the user. Since the completeness depends very much on the expert knowledge of the respective application, a context-specific methodology is needed for its evaluation and visualization. This work proposes such a methodology, which is consisting of four major steps. The first step includes process modeling and collection of data sources, taking into account information classes relevant for data-driven defect prediction. Then, based on the knowledge of process experts, potential quality-influencing variables are identified and weighted. In the third step, a metric is defined based on which the completeness of existing data for individual process steps and information classes is quantified. The fourth step includes the user-friendly visualization of the assessment. The transparency thus created with regard to the completeness of data in the context of Predictive Quality applications allows the identification of data gaps and the derivation of needs for action. This work, therefore, starts even before the actual modeling process and provides a tool for improving the underlying database.

The aim of the Industry 4.0 research initiative is to enable globally networked, flexible and adaptable production. This results in the vision of the Cluster of Excellence Internet of Production (IoP): Creation of a new level of cross-domain collaboration between all production process participants, so that semantically adequate and context-aware data is available along the value chain. The main challenges hereby are the large amount and heterogeneity of generated production data. Both challenges can be handled by combining the advantages of existing object relational databases (ORDBMS) for storing production data and ontologies for data management and integration within the ontology based data access (OBDA) approach. This paper provides a concept of mapping an application ontology as a gateway into an OBDA system for management of production data. For that, transformation rules of an ORDBMS into an application ontology are defined and implemented. The proposed system can enable a first step into the vision of a global Internet of Production, consisting of different levels of semantically enriched data. The aim of this paper is thereby to highlight potentials of semantic data modeling for a global sustainable production network.

The Integration of Process Planning (IPPS) and Job-Shop Scheduling Problem (JSSP) based problems is brought into a new scale through the concept of smart manufacturing. In that, the advent of industry 4.0 elements allows the manufacturing dynamic environment to be involved through the life-cycle data. In that case, seeking optimization is thus a function in the machines and other resources’ instant states. The new perspective is an awareness-oriented, where optimization is a function in the available sub-system data. Choosing suitable data is a part of the cloud design that attempted to bring manufacturing cloud into reality. The IPPS problem requires a more objective-oriented manner to be better analyzed and to obtain maximum benefits as possible from the life-cycle information. With the aid of the commonly used meta-heuristic techniques in that approach, the current study modifies a genetic-based algorithm. The algorithm is executed in two stages, whereas the second stage adopts a neighborhood searching algorithm to avoid local optima.

The quality of education directly affects the level of qualifications of graduates of higher education institutions. One of the most important factors is the practical experience gained during training. Unfortunately, most of the students of technical directions in the field of automation do not have the opportunity to gain practical experience of working with new equipment and, as a result, they do not have the skills to develop and manage modern automated control systems. This is influenced by several factors - the lack of Russian solutions in this area and the high cost of analogues offered by foreign companies. In this paper, the authors demonstrate their experience in developing highly specialized training stands and general-purpose stands. Step by step this process is described from design to ready-made smart automated control systems with accompanying documentation. The designed stands have sufficient flexibility and modifiability for use in a wide class of educational tasks.

In the context of the transformation of the electric power industry, leading to an increase in the efficiency of electricity production and the creation of new relationships between the areas of production and consumption (for example, the Demand Response mechanism, DR), ever deeper penetration of information technologies into the production technological processes of electric power systems (EPS) is arising. Systems, which include both technological and informational components, become cyber-physical ones. A new type of such system’s control appears—it is cyber-physical management (CPM). Certain intelligent nodes with special loops are created at the lowest level of the CPM hierarchic structure. Malicious intrusions into cyber-space of cyber-physical systems form a class of cyber-physical attacks. In the event of cybersecurity threats, those loops will carry out a defensive mission, blocking the CPM upper levels and the main control center from cyberattacks. The article proposes to discuss the features of cyber-physical EPS and consider the advantages of cyber-physical management using the example of the structure of the DR Aggregator.

The article presents results of analysis of the current state and perspectives of development of charging infrastructure for electric vehicles (EV). In Russia, there is an increase in the number of charging stations, but this growth is quite random and is largely determined by marketing considerations, which are based on the desire of individual organizations, corporations or regions to be in the trend of innovative development and demonstrate their competitiveness in the innovation sphere. The aim of the study is the development of a system’s criteria for the selection of concrete places for investment projects on creation of EV charging stations. Proposed criteria were piloted in the development of a preliminary plan for allocation of EV charging stations along the E-18 highway from Saint-Petersburg to border crossing Torphyanovka at the Finnish-Russian border. Vertically integrated refineries such as Gazprom, Lukoil, Rosneft, Tatneft, Shell, etc. represented in the automotive fuel retail market, have the opportunity to profit from both the wholesale and retail stages.

Two control algorithms are proposed for the conditions with an unknown part of parameters in a classical description of the model of a three-link robot-manipulator. The first algorithm for deriving the controlled motion equations is based on the motion kinematic analysis in an assumption of a possibility of determining (measuring) all variables and parameters except for the third-link mass. The second algorithm of control design relies on the principle of nonlinear adaptation, which implements the main concepts of the synergetic control theory and admits limited arbitrary noise (nonrandom/random) with respect to the controlled variable. The proposed controls synthesis techniques ensure the robust properties of both regulators. The algorithm based on non-linear adaptation has a clear priority (in terms of accuracy) over the first algorithm under conditions of physically acceptable values of the unknown third-link mass. But the control by the first algorithm is an order of magnitude more energy efficient than adaptive control. The results of numerical comparative modeling are presented for the case of an unknown mass of the third link (load), and the properties and bounds of preference are indicated for each of the algorithms depending on the level of uncertainty.

In this work we describe a cyber-physical decision support system integrated into the process of researching and developing novel antifungal antibiotics, particularly polyene macrolide antibiotic derivatives. The methods and technologies used to develop the decision support system include modern methods of web-application development (client-server architecture utilizing microservices, asynchronous queues for processing longer tasks), modern methods for developing predictive models (recurrent neural networks), specialized deterministic algorithms, and thin interfaces. The cyber-physical system enables chemists-researchers to make better decisions when selecting potential antifungal drug candidates by predicting the properties of antifungal antibiotics. The models predict toxicity, reaching an average AUC of 0.86 across relevant assays as well as 0.02 mg/kg mean squared error for oral toxicity (LD50, rats). Antifungal activity is predicted using a deterministic algorithm, which was able to correctly separate a set of antifungal and non-antifungal drugs into their respective categories. The mathematical models were trained, tested, and validated on a set of antifungal antibiotic data. Testing showed the models’ accuracy and viability for predicting antifungal antibiotics’ properties.

In this paper, the following issues are described: implementation of an adaptive regulator with a reference model of an energy turbine in the presence of parametric and signal disturbances with a lack of a priori information about the control object; identification of the control object for the expansion of the transient response in the Fourier series; approximation of the control object by a second-order aperiodic link; implementation of an adaptive power turbine regulator in a programmable logic controller; adjustment of the adaptive regulator of the power turbine, the removal of transients and characteristics; adaptive regulation of an energy turbine in a programmable logic controller; and implementation of identification and adaptive control algorithms on the OSRT using the C++ programming language. An important conclusion is that adaptability under parametric and signal disturbances in the presence of information about only one output of the object is possible. In this case, the question of separating parametric and signal disturbances is not worth it, since both parametric and signal disturbances are processed by the discrepancy.

Collecting and analyzing user experience with a digital product allows the most accurate determination of the decision efficiency made in the solution design and assessment of the degree of user satisfaction. The comfort of working with digital content is a key parameter that largely determines work efficiency and financial success. Surveys, questionnaires, and evaluation systems require the user to spend additional time and break away from the main activity. In this article, the authors propose a system for collecting and analyzing user experience based on recognizing the emotions displayed by the user during interaction with the graphical interface and the content provided by the digital product. The paper presents a convolutional neural network of its own architecture, with the test accuracy FER+ 71.1% for eight base classes. The method of fixing the emotional assessment of the user is considered. Approaches for evaluating interaction with the interface and video order are proposed. The results of the software performance with a PC test application are presented. The software results coincide by 67.2% with the results of the user survey. The 32.8% error mostly is of no more than one gradation of the rating scale. The analysis of the obtained results is carried out, and approaches for further system improvement are proposed.

The paper presents several recurrent algorithms that provide the implementation of fractional proportional-integral-differential controllers in digital control systems. The main attention is given to the study of the algorithm parameters’ role and influence. To illustrate the algorithms’ performance there are presented the dependencies of accuracy from the parameters for various methods of approximation. The benefits of the proposed algorithms include improving the operational and technical characteristics of a control system being applied in various problem domains, including production and transport industries. In general, the proposed approach allows improving the stability and quality of transient processes. Further development consists in combining different numerical methods to implement a fractional proportional-integral-differential controller. An example of the use of the proposed algorithms in practice is given for a control system of a gas-diesel locomotive internal combustion engine, which illustrates the benefits of the proposed approach. Modernization of the control system allows reducing fuel consumption and increasing the safety of the locomotive gas-diesel internal combustion engine.

The interaction of gamma rays with matter is the crucial factor in the detection of these rays. Scintillation detectors respond to gamma rays by producing a small flash of light, or scintillation. The detector crystal is mounted on a photomultiplier tube that converts the scintillation into an electrical pulse which is processed to generate a signal. A class of multiphase flowmeters is based on this principle to measure the different fractions contained in oil well flow. In this paper, different modes of gamma-ray interactions were described, the statistical distribution for gamma-ray generation and detection was presented, and a simulation of detection of gamma-ray passing through a tube containing oil and gas was performed. The simulation led to the generation of data which was analyzed and a conclusion was reached that the detected gamma rays have a different statistical distribution than the generated rays due to the gas bubbles’ presence and absence in the flow.

The nonlinear correction method is widely used in automatic control systems to increase stability margins and quality rating. This method has also found application in piloted aircraft control tasks, where the loss of stability and the occurrence of oscillations are unacceptable. This paper discusses new nonlinear correcting devices. Their structure is based on separate channels of the desired control signal amplitude and phase formation. Their application is demonstrated by the example of a piloted aircraft, the control systems of which have actuator rate limits. The illustrations of frequency characteristics, spectrograms of nonlinear correcting devices, time processes of the input, and actual signals of corrected and nonlinear uncorrected systems, from which the efficiency of the serial correcting devices are observable, are presented.

This paper focuses on the development of a non-intrusive driver warning system as part of ADAS (Advanced Driver Assistance Systems) to help improve the safety of all road users, when driving, on the road. The proposed algorithm uses computer vision, implemented based on facial landmark detection, to detect driver drowsiness based on the driver’s eye condition. This algorithm has shown good results with HOG + Linear SVM for searching and locating faces in the image, as well as determining the eye condition of the driver with and without glasses. If the eyes remain closed longer than expected or if the driver is not looking straight ahead, it is an indication that the driver is drowsy or tired, the system then sends a warning signal to the driver.

Quite often, developers face low performance, hanging, and other problems when they’re developing sites. To solve such problems, we need to trace site requests. Existing tracing methods do not allow tracing the progress of requests from a client’s web browser to a server or group of servers. In this paper, we propose distributed tracing mechanism that allows tracking requests starting from the browser. For generating complete client-to-server tracing, the client application must be able to initiate the appropriate request. For the execution of these actions, we need to use a unique library. In the paper, we consider the algorithm of such a library. A popular tracer (OpenTracing) is used on the serverside. Based on the proposed methodology, a library was developed. The library's work was tested. Testing has shown that using the library, and we can track the complete chain of requests from a browser to the server. Trace result is presented in graphical view. This allows analyzing received data and finding bottlenecks when queries are passing. The novelty of the proposed solution is that the request is traced from the client application and to the client application. That is, the full path of the request is shown. The result is presented in a graphical form that is convenient for analysis. The library is designed primarily for the development of client-server applications and for support services.

The presentation of cyber-physical systems and devices as components of a multi-agent model is considered. The main method of interaction between agents of the cyber-physical type is investigated - the exchange of messages in the multi-agent model. It is proposed to consider the processing of messages from other agents as an indicator of the agent's performance. The dependence of the efficiency indicator of the agent-based model as the probability of timely message processing on the message flow density is analyzed. The technique of numerical modeling of systems with continuous time is considered. For this, the representation of such systems in the form of equivalent signal graphs is used. A technique for constructing such graphs based on a system of differential equations is proposed. A method for joint modeling of dynamic systems and systems with random flows is proposed. A graphic-analytical technique for numerical modeling of a queuing system in the form of a system of inhomogeneous linear differential equations of the first order is considered. An algorithm for the numerical modeling of queuing systems proposed by the authors, based on the previously considered technique, is presented. An example of modeling message processing by the simplest agent in the form of a queuing system is considered. An equivalent numerical simulation graph is constructed, on the basis of which a matrix simulation scheme is obtained. Examples of a simulation experiment and comparison with analytical results are given.

The first open-source and low-cost fused filament fabrication 3D printer of 2004 represented a new opportunity to manufacture mechanical components easily and affordably. 3D printers have been continuously developed since the first one entered the market and are equipped with various sensors that can monitor the printing process. Despite these improvements, a probability of 41.1% remains that the printed part will have errors. This can lead to an irreparable print that has to be canceled, but is often not noticed by common sensors, while costly time, electricity and filament continue to be consumed. This investigation provides an account of a camera based monitoring system developed to detect complex problems that are not easily recognizable with sensors commonly used for fused filament fabrication. Image segmentation was used to remove the background of the printed part and the result was compared to a visualization of the G-Code. By using an exclusive-or method it was possible to determine differences, which can indicate defects. Depending on the similarity, the printing process can be canceled promptly. Tests have demonstrated that this method works reliably even under changing lighting conditions in most cases but can lead to poor segmentation due to shadows being cast in the infill. The application is also able to recognize differences when printed parts detach or layers have shifted if they are not covered by lower layers. The use of a light source on top of the 3D printer and additional cameras, beside the build plate, could solve both problems in the future.

The article presents a new approach to developing web audio services. The approach is based on the use of Progressive Web Apps (PWA) technologies and data caching. Data caching using special caching strategies allows for efficient use of network resources and makes the web application capable of working offline. PWA also provides the ability for a web application to be installed on a smartphone directly from the browser. The use of this approach allows you to reduce the cost of developing audio services for various operating systems. As part of our work, we have developed a client application for the audio service. The article presents the structure of the application and the technologies that we used. The results of the research are recommendations for configuring resource caching and a description of a set of tools for developing a client application of a web audio service that can work in online and offline modes.

This paper introduces the application of some principles of the actor-network theory (ANT) to understand the sociotechnical effects of the implementation of intelligent information systems in logistics, i.e. the effects on the organizational structure, business processes and decision-making policies. It is shown how intelligent agents (IA) playing a core innovative role in modelling an integrated information system for logistics participate in the processes of formation and stabilization of alliances with human actors (developers, users), the material infrastructure of a logistics system and information/knowledge resulting in a new quality of design of control systems for logistics, standardization of protocols and invariance of server-client relations in the space of networks supporting stability of the information infrastructure of a geographically distributed organization. The paper aims at bringing the attention of the AI researchers, MAS theorists, human-machine systems engineers, ergonomists, knowledge engineers, logistics specialists and broader research community to the actor-network paradigm and its applied potential in sociotechnical systems research.

Estimation of attention characteristics is used to analyze and improve processes of intellectual activity (e.g., in learning), as well as a way to obtain alternative feedback and create intelligent affective systems. This paper describes the creation of a tool designed to determine the correlation between levels of attention characteristics and electroencephalogram parameters using the Muse-Headband brain-computer interface (BCI), as well as the results of experiments carried out with our tool we designed and developed to find the correlation between levels of attention characteristics and blink attributes. The experiments have shown that there is an inverse linear relationship between measures such as blink rate and concentration level. The presence of this dependence gives us the opportunity to create more complex and in-depth systems for monitoring the level of attention properties, which can help to analyze and improve processes of intellectual activity. Such systems can operate independently, receiving data from a simple BCI via Bluetooth technology, which will have little or no distraction to users from their intellectual activities.

In this study, we consider the author verification problem as a binary classification problem where the aim is to identify whether the two inputs belong to the same author or not. For this task, we have used several Siamese convolutional neural networks-based models. The first model employs a Siamese network which is trained using binary cross-entropy loss after the absolute distance computation. In addition to this baseline model, we have implemented another model which employs a concatenation operation. Moreover, two contrastive loss-based models have also been implemented for the same task. Two publicly available benchmark datasets, IAM and CVL, have been used in the study. Training, validation and test datasets that have been used to train and test each of those models have been generated from those datasets. Experimental results show that the performances of the two cross-entropy loss-based models are similar and much better than that of the contrastive loss-based models.

The paper deals with the practical issues of the formation and analysis of a tool suite designed for the development of agent-oriented software operating under resource constraints. The suggested tools are based on the expansion of the concept of commitment by adding a time parameter as the condition of commitment feasibility for the interaction of agents (inter-agent communications). Thus the concept of a temporary commitment is introduced. In addition, an event-driven consensus for a multi-agent system (MAC) is considered, acknowledgment protocols are used to ensure predictability in real-time. The architecture of the instrumental platform is proposed for the development of multi-agent software to operate on a target real-time system. The proposed development tools comply with the specifications: FIPA, RTSJ. Certain aspects of dynamic analysis are considered, metrics such as the resource usage are determined - various time indicators of the designed MAC, the memory usage; the influence of the characteristics of the underlying software layers; the dependence of the performance and resource consumption of the agent-oriented application on the build mode of the executable code.

The article is devoted to the technology of the Internet of Things, namely, the weaknesses of the specification of the LoRaWAN protocol. The presented theoretical and practical material characterizes replay attacks as the most relevant. Different types of replay attacks are presented: ABP-based, OTAA-based and OTAA join. The article presents the results of the implementation of attacks, confirming the sensitivity to them of devices with both the ABP activation scheme and the OTAA. Based on the presented data, a conclusion is made about the need to use means to prevent or mitigate actions from attacks. It is proposed to use intrusion detection tools such as LAF to form a secure infrastructure of the Internet of Things network. A practical confirmation of the effectiveness of the use of such protection means is presented. Also, in the course of the research, the shortcomings of the LoRaWAN protocol itself were revealed, requiring immediate practical improvement.

This paper provides a detailed overview with an in-depth comparison of existing methods for solving problems arising in the analysis of mathematical models of Cyber-Physical Systems (CPSs) in the field of the Industrial Internet of Things. Alternative approaches to building models and solutions in this area are proposed. Both practical and theoretical components are considered. As an example, we consider the detection of anomalies in the operation of the FSC of the Internet of Things (IoT). System state forecasting is performed based on the analysis of multivariate time series. The data in question is obtained from the physical (lower) layer and the logical (high) layer. Predictions are made using neuroevolutionary algorithms, and analysis is carried out by comparing the obtained normalized data with the predicted indicators of the system. The research is conducted on a validated dataset. The results of the study demonstrate the effectiveness of the method used with indicators of its correctness and accuracy.

The possibility of applying the new intention-based networking (IBN) paradigm to the efficient functioning of information processes in cyber-physical systems (CPS) leads to the need to consider new security threats. This article discusses the purpose and main characteristics of IBN. IBN provides a transition from a device-centric network model to a business-centric model operating at a higher level of abstraction. The main differences between IBN and the traditional approach to the construction of network architectures are highlighted, the main advantages of the use of IBN are stated. They consist in providing a more flexible operation of network technologies and optimization of their performance, as well as ensuring compatibility of network technologies through algorithmic management of the configuration of the network. The actual researches in the field of security of such networks are analyzed, the main security problems, arising during the transition to the construction of network infrastructures on IBN paradigm are marked.

In the modern world, there are many examples of cyber-physical systems (CPS) that are characterized by the integration of physical and informational processes. Some of the CPSs refer to critical infrastructure, so, security breach in such systems can lead to disastrous consequences. The paper proposes an approach to cyber-physical systems security based on graph theory due to CPS distributivity and heterogeneity. The problems of CPS security were considered. The basic concepts of graph theory and its possible application to the tasks of modeling modern world objects are given. The existing approaches to CPS security were considered. The main disadvantage of existing approaches considered is that they do not take into account the structural properties of the CPS and the properties of the data circulating in the system. In the paper, it is proposed to model CPS as a graph. That approach allows considering structural features of CPS and also it can be used to take into account properties of data circulated in the system.

This paper presents the results of developing smart technologies and computer systems for data and knowledge gathering, integration, interpretation, and visualization, and the implementation of the results as an integrated information-analytic platform (IAP) of complex objects (CO) life cycle (LC) proactive control. The implementation employs an original approach to operation and design of IAP of CO LC proactive control based on the use of cyber-physical systems (CPS) and intellectual interfaces (InIf). This combination ensures effective cooperation between structural or/and functional elements, and facilitates proactive control of those elements’ LCs based on continuous processing of information flow generated by sensors and devices. The created InIfs implementing means of visual programming provide a user with an adjustable environment making the interaction between the user and IAP of CO LC proactive control seamless. The simplicity of interaction is increased even further through the use of user’s subject area terminology. The practical value of the developed IAP is its replicability and multifunctionality that made it possible to implement a broad spectrum of custom-built information systems one of which is the unified spacecraft component electronic passport containing the actual information on the state of the component throughout the entirety of its life cycle, thus eliminating ambiguity and reducing information incompleteness. Another project an IAP was applied in was the production of mobile RFID-based civil airlines service CPSs. The primary advantage of those systems is that they are autonomous and have a capacity for real time system supervision and control. Such IAPs have already been developed and deployed for 17 Roscosmos subsidiary engineering and service companies, as well as for Moscow, Saint Petersburg, and Vladivostok airports having 75 of those in total.

The interconnection between information technologies and enterprise functioning is increasing every day (“servitization”) [15]. The transition to a new management paradigm based on information services (Industry 4.0), assumes that the enterprise information system will be able to withstand the new increased loads. Otherwise, it is necessary to modernize the corresponding hardware and software resources. Such changes are impossible today without careful integrated planning and scheduling, because these are large-scale problems. However, these are the most common in practice. In general, the main goal of modern research is to develop models, methods and algorithms that increase the validity of management decisions, but the task of correct integration of investment projects and EIS modernization programs development is not considered. This study is devoted to the analysis of these issues.

Today, due to digitalization, it is important to extend the functional capabilities of state-of-the-art robot control systems. In the paper, we consider digital semiautomatic control systems of robot manipulators with tool position and orientation set on a tool’s trajectory by a human operator using multi-degree-of-freedom master handles. The programs for preliminary planning of the robot links’ trajectories in offline mode were developed with the use of iterative numerical methods. Known theoretical methods of tool orientation are discussed. The problems of planning with different ways of tool orientation setting are formulated. We analyzed iterative algorithms with and without calculating the robot’s Jacobi matrix. The algorithms based on solving the nonlinear least squares problems in suggested formulations were recommended. Examples of planning link trajectories for a 6-link robot are presented when a tool moves along a square piecewise 3D trajectory. Different techniques of tool orientation setting on the trajectory pieces were applied. Testing of planned trajectories was carried out using the robot animated model. The results of the work can be implemented in computer simulators for study and design of modern digital semiautomatic control systems of robot manipulators.

The problem of assessing the state of complex objects is solved by various methods. A new approach to solving this problem by constructing an analytical model based on expert knowledge is proposed. It is shown that the algorithm for formalizing expert knowledge is a synthesis of the elements of set theory and the theory of planning experiments. In this case, the expert fills in a special questionnaire, the rows of which are various combinations of variable values in the form of production rules of the implicative type “if… Then…”. The implementation of the proposed approach is shown by the example of assessing the state of the chimney of a metallurgical plant, for which expert knowledge is formalized in the form of a polynomial five-factor model. An example is illustrated by a figure of monitoring the state of a collapsing pipe if it is necessary to maintain its operability over many years.

The research is devoted to solving the problem of conjugation of the virtual information space and the physical world in terms of data retrieval. Wherein the methods for solving the problem of extracting data from the virtual information space are determined by these data themselves (data-driven). The paper discusses ways to solve the problem of thematic content obtaining (data retrieval) from an unstructured set of information resources or news feeds. The problems of the “growing bubble” of unprocessed documents that arise during the “blind” collection of documents are discussed and ways to solve these problems are proposed in the paper. To reduce the resource consumption of the problem of forming a periodically updated search base, three approaches to the automatic collection of “raw data” are proposed. The proposed approaches to developing the sub-search systems are the part of a large class of modern methods and algorithms of adaptive heterogeneous data filtering for content retrieval and aggregation in subject oriented knowledge bases formation. A possible field of application and relevance are determined by the fact that for the closed cyber-physical systems the use of sub-search systems is proposed with unlimited and unstructured information space as an input that is being processed in real time. One of the possible implementations of proposed methods is in the development of the knowledge bases for science-technical documentation.

Communication of two independently functioning mobile robots (MR) under the operation of ionizing radiation source (IRS) search and removal/elimination is discussed. Communication algorithm for two MR is proposed. Transition conditions for Petri nets, as responsible for the MR group control, are described. To study the communication of two ground-based MR equipped with gamma-detectors, a model was created and robots’ behavior simulated. The authors of this paper stand for discrete event-driven model to be implemented as the MR group operation control mechanism, if group members are to perform one and the same task of IRS search with further removal/elimination.

The paper presents a new approach to building a software system for diagnosing plant diseases from leaf images based on three-dimensional digital descriptions of leaf images and a modified convolutional neural network. Digital descriptions are presented as sets of Haralick parameters in the form of png images for each color component of the leaf images. The preprocessing task was to prepare a dataset (leaf images) by resizing and extracting 3D functions performed before accessing the neural network. This is achieved using the Gray-Level Co-occurrence Matrix (GLCM) and calculating a set of Haralick features that reflect the properties of the image texture and also have a significantly lower dimensionality. In this regard, convolution operations in the structure of the neural network are excluded. Combined with the reduction in dimensionality of digital descriptions of target images compared to real target images, it significantly reduces the time spent on neural network training and diagnostics. It is also proposed to form a dataset for training the neural network relying on a small number (about 20) real images of leaves with subsequent averaging the obtained Haralick features and randomization. This is done by adding random normally distributed values to the estimates of the mathematical expectations of the Haralick features. The confidence interval of the scatter is determined during the simulation or testing of the neural network, provided a given level of correct diagnosis is achieved.

The problem of creating datasets for training and testing neural networks is described in the example of the task of social network management. A method of expert dataset synthesis based on experts’ knowledge of the subject area is proposed. The essence of the method lies in the fact that sets are generated randomly within the range of values and regularities given by the expert. The results of using synthetic datasets are presented. The effectiveness of the expert method of dataset synthesis in cases where it is difficult or impossible to obtain real data is shown.

A method of structural training of a compartmental spike model of a neuron is proposed. The training task was to form a response to a pattern represented by a vector of single spikes with time coding, where each component of the vector enters a separate dendrite of the neuron. Depending on the given pattern, the algorithm automatically selects such parameters of the neuron model as: the size of the soma, the length of the dendrites, and the number of synapses on each of the dendrites. A sign of pattern recognition by the trained model is the generation of an output spike by the model. The quality of pattern recognition has been studied. Experimentally, the universal value of the neuron activation threshold is determined, which does not depend on the size of the vector describing the input pattern.

This article considers the problem of neural network approximating the solution of the Cauchy problem for a higher order ordinary differential equation. Adaptive approximations of this kind are useful for using as mathematical models of dynamic objects in the creation and adaptation of cyber-physical systems. The approximation scheme is based on the Taylor expansion of the solution with a remainder in the Lagrange form. Furtheron, the second derivative of the solution in the remainder expression is sought in the form of the output of a neural network with a radial basis function (RBF). An algorithm this RBF network training for (selecting the optimal parameters) based on solution data at selected points is presented herein. These data simulate measurements that can be obtained by observing the real object. The algorithm’s main idea involves utilizing the clusterisation for the adjustment of nonlinear input parameters (centres and Gaussian rates of decrease) and pseudoinverse matrix applying for linear input parameters selection. On the example of solving the Cauchy problem for a second order non-linear differential equation, i.e., the Duffing oscillator, a comparison of various approximation schemes is carried out.

The digitalization of the modern economy requires taking into account the properties of poorly formalized phenomena and situations in the financial sector, which affect the decision-making process. The decision-making process is influenced by under formalization of certain phenomena and problem situations, as well as various kinds of interactions in finance. Analysis and assessment of semi-structured problem situations and hybrid influences on the decision-making process dictate the need to study various kinds of relationships in finance, which necessitates the creation of a decision support system in finance in the form of a digital twin−end-to-end management technologies. The use of these technologies using a systematic approach based on logical-linguistic modeling of the decision-making process through cyber-physical technologies allows us to consider the decision-making process in an integrated, structural and systematic way within the subject area, ensuring the maintenance of homeokinetic equilibrium in the system. Digital twin (neural network cyber-physical technology) as part of the decision support system will collect and analyze in real time data about the current state of the facility for the subsequent and immediate recognition of risks in regard to problem situations.

The effectiveness of the functioning of cyber-physical systems is based primarily on the use of powerful methods of obtaining and processing information. The complexity of the structures and properties of cybernetic systems, as well as the conditions of their functioning, determines special requirements for the methods of measurements and calculations performed in such systems. As a rule, the uncertainty of CPS models, as well as the uncertainty of the influence of external environment factors and their interrelations with the properties of systems, primarily determine the requirements for the intellectualization of measurements and computational information processing. This article offers methods and tools of Bayesian intelligent measurements (BIM) to ensure the effectiveness of managing cyber-physical systems under conditions of uncertainty. The concept and methodology of creating an intelligent industrial Internet of Things (IIoT) is proposed, the distinctive feature of which is the intellectualization of measurement methods and predictive data processing. For this purpose, the IIoT includes an intelligent DATALAKE, which is built on the basis of a Bayesian intelligent measurement system that implements not only the functions of measurement and data integration, but also support for management decision-making. Examples of real cyber-physical systems with control based on Bayesian intelligent measurement tools are given. The prospects of using the proposed solutions based on BIM in various modern technologies based on the principles of BIG DATA, DATA SCIENCE, neural networks, IIoT, DATA MINING and others are considered.

The article deals with the challenge of understanding blockchain technology in the perspective of the implementation of blockchain based solutions in the industry. The authors present a new methodological approach to the study of blockchain technologies as complex systems, with emphasis on the business ecosystems including the analytical framework for the research of the blockchain technology as a complex productive phenomenon and the application of graphical model comparing visibility and production usability of new technology to the industrial blockchain. The paper discusses to what extent the introduction of distributed registry technologies can improve the efficiency and effectiveness of management methods in industry. In this regard, the most relevant avenues for blockchain technology implementation are outlined. The proposed methodological approach is used to consider the perspectives of blockchain adoption in the industry and an ability of its particular applications to pass the hype stages in the near future. The benefits of large-scale blockchain implementation are discussed for each of the three main areas: tax, financial, management accounting and audit; the support of the cyber-physical systems within business ecosystems with the facilitating of the working capital flows and decentralized finance for production companies.

The article describes a new approach to creating a hybrid online course “Building an intelligent enterprise based on SAP S/4HANA”, which formed the basis of the program “SAP S/4HANA Academy”, designed to train the personnel reserve for the SAP partners and clients−SAP ecosystem in the CIS region. The training was conducted among undergraduate and graduate students of Russian and CIS universities. The new course design is considered in comparison with similar courses previously conducted by the SPbPU SAP Academic Competence Centre (SAP ACC SPbPU), the experience of which was taken into account and revised. The course structure, its theoretical and practical components are described in details, the course landscape and the tools used in its implementation are described as well. Separately, the aspects that are traditionally difficult while conducting mass online courses are considered: technical implementation of practical assignments, individual organisational and technical support for students. The summary of the authors’ experience presented in the article can be useful in the preparation and implementation of hybrid online courses of various directions.

The analysis, modeling and forecasting of the dissemination of information in social networks, the study of statistical patterns of the totality of information processes is currently a very urgent task. Some aspects of information process management in social networks are considered. To optimize management activities, a classification of communities and groups of users of social networks is proposed. The empirical distributions of various statistical indicators that characterize the information processes in such communities and the interaction between their participants are synthesized. Some statistical features of the relationships between these indicators and the general patterns of so-called mutual transitions were analyzed. The article proposes and justifies regression models that can be used to predict the behavior of user participants in the corresponding groups, in particular, their activity, as well as other parameters that characterize the functioning and evolution of various social network communities. The authors proposed a classification of VK communities (groups), for which the corresponding goals of creation and features of functioning are identified. Conclusions are made, and the directions of further research implementation are determined.

Currently, cyber-physical systems are being actively introduced in many areas of modern society. In particular, training on computer simulators is gaining popularity. Computer techniques make it possible to improve the quality of perception of educational content by applying virtual effects that simulate processes occurring on a real technological object. This paper describes the development of cyber-physical training systems using the example of a system for training operators of an assembly machine for the manufacture of automobile tires. The paper considers the specifics of the use of virtual simulators in the training of operators on the example of an assembly machine for the manufacture of automobile tires. The article discloses important features of the educational process for specialists of existing industries within the framework of the ongoing development. In the framework of the article, the structure and principle of the machine is considered. The main requirements for the developed simulator are formulated. The main scenarios of its work are presented. The simulator was developed on the basis of the Unity virtual engine, the article contains the main objects and components implemented. The work also provides examples of class diagrams developed to implement the logic of tasks performed by trainees. The virtual simulator consists of training subsystems, service functions, reporting and data storage, which communicate with each other through a local network.

Complex socio-economic processes and rapid changes lead to the fact that modern cities, which are complex systems, cannot withstand numerous challenges without using a new management model that will allow flexible adaptation to changing external conditions. The “smart city” concept is one of the most demanded models for managing a modern megapolis. This concept is based on several key points. Sustainable migration management is essential. Migration modeling is a reliable tool for managing a megapolis as a complex socio-technical system. Various models of migration are known, but from a practical point of view, simulation models that take into account the impact of migration on the sustainable development of a megapolis are of the greatest interest. The aim of the study is to develop a system-dynamic model for managing the migration flows of a metropolis, which will significantly improve the quality and efficiency of its sustainable development. For its formation the package iThink 9.1.3 is used. The model was verified on statistical data about migration flows of St. Petersburg.

People living on territories try to improve their well-being−by developing technology and better social organization in line with preserving the environment. In general, contemporary utilization of CPS/IoT implies mainly natural and engineering domains making those smart. In the study clarification through network science of the role of advanced technologies and that of CPS has been proposed for further elaborating programs of sustainable development of specific remote territories. Case study included three border area (RF-Mongolia) territories of Eastern Siberia, Russian federation. A previous combined stem network model was improved to find the balance between socio-economic development of territories in whole, environment protection, and personal hopes of local for future. The novel model implies that three groups of entities: actors, infrastructures, and natural assets might be observed so that each group forms network subspaces in network space. Thus, each subspace is represented by a pertinent combined stem network. Additional interdependency effects not taken into account in previous stem network models we identified with special kind of links. Imbedded with tens or hundreds of thousands of sensors and actuators a territory decision support system might be provided with thorough network metric values just to choose optimal ways for territory development and elaborating effective measures to counteract coming threats. Terminology and qualitative side of this model particularly should deliver network forma mentis to experts from diverse domains and bridge them for further effective utilization of modern technologies to increase quality of living on the territory.

Emotional-enhanced learning is a meaningful driver of engagement leading to long-term memory retention in learners, however, traditional approaches such as problem-based learning, and project-based learning, among others, do not consider brain-based learning guidelines concerning learner’s emotional experience design. The Neuroscience-based Learning (NBL) technique is a novel neuro-educational approach that applies the implicit neuro-physiological mechanisms underlying vivid and highly-arousal emotion-al experiences leading to long-term memory retention. The NBL is devised from a cybernetic system point of view, by explaining the novel neuro-physiological learning scheme describing the relation among the environment and the learner’s internal mental processes ranging from perceptions, comparison with previous experiences and memories, immediate sensations and reactions, emotions, desires, intentions, higher order cognitive functions, and controlled actions towards the environment. While explaining biological processes, the scheme also relates the types of memory systems with their non-associative and associative learning mechanisms, and the variables that modulate learning. NBL proposes the triggers for a vivid and highly-arousal emotional learning, which are novelty, unpredictability, sense of low control, threat to ego, avoidance (aversion-mediated learning), and reward (reward-based learning).