Aim. Currently, the mainline railway network does not fully meet the market’s needs in terms of capacity, which is why every day JSC RZD is forced to reject 2 to 4% of applications for the transportation of goods across routes passing through limited-capacity railway infrastructure facilities. Increasing the practical capacity of railway lines by building additional tracks at stations and on open lines is unacceptably long and requires massive capital investments. Methods. It is proposed to solve the above problem by increasing the utilization of railway infrastructure by reducing the succession times and station operation times in the traffic schedule by deploying moving block and virtual coupling. This approach allows reducing capital investments up to 32 times and the duration of project implementation up to 8 times with a comparable capacity increase. The paper details the functionality of train separation technology and its technical parameters. Conclusion. The train separation technology has been deployed and is in operation on the Far Eastern, Trans-Baikal, East Siberian, and Krasnoyarsk Railways, branches of JSC RZD, that are part of the Eastern Operating Domain of JSC RZD network. By the end of 2023, the total length of track covered by the train separation solution was 5,734 km. Conclusion. In 2023, the lines of the Eastern Operating Domain handled almost 36 thousand pairs of trains with reduced spacing. The train separation technology also allows improving the safety and the level of digitalisation of the transportation process.

The paper proposes a method for defining a dedicated operational context as part of the development and deployment of autonomous corporate governance systems. The case study of autonomous board of directors systems is examined. A significant part of the operational context for the autonomous corporate governance systems consists of the regulatory and legal framework that regulates the company’s operations. A special operational context for autonomous artificial intelligence systems can be defined by simultaneously formulating local regulatory documents in two versions, i.e., to be used by people and by autonomous systems. In such a case, the artificial intelligence system receives a clearly defined operational context that allows such a system to perform its functions with a required operational quality. Local regulations that take into account the specificity of operations involving individuals and autonomous artificial intelligence systems can become the foundation of the relevant legislation that would regulate the development and deployment of autonomous systems.

Problem definition. Many artificial intelligence systems are essentially event classification systems. They are widely used in predictive analytics. Their role as predictors of hazardous events in transportation is constantly growing. The efficient application of artificial intelligence methods largely depends on the results of misclassification. Therefore, the problem associated with the calculation or statistical evaluation of the probability of misclassification and boundary value definition is of relevance. Aim. To estimate the boundaries for the combined probability of misclassification due to two different categories of errors, i.e., misclassifications proper and statistical errors resulting from misclassification. Results. The threshold value that is used for classification was statistically evaluated. The boundary conditions for the combined probability of misclassification were established. A generalization for N-dimensional spaces and general distributions and shapes of threshold surfaces was presented. The theoretical findings were illustrated with an example of practical application.

Aim. In the context of machine vision, the problem of detected object boundary definition is usually solved using semantic segmentation that requires a high computational resource. Its application increases the complexity and the cost of the implemented hardware and software systems. This paper proposes an alternative method for defining the boundary of a segmented object, i.e., a railway track, for a train traffic tracking system. Methods. Since a railway track in an image can be represented by an n-th order polynomial, it is suggested to solve the problem of track boundary detection by using approximations in the form of straight lines. It is suggested using the Hough transform to detect straight lines. The former’s parametric space will be arranged in accordance with the problem to be solved. Conclusions. The proposed approximation will allow abandoning semantic segmentation and reduce the computational complexity and load.

The paper aims to study the primary dependability characteristics of restorable k-out-of-n systems with arbitrary distributions of failure-free time and time to component repair, as well as the total number of repair units. A k-out-of-n system is a system consisting of n components that fails when k out of its (k≤n) components fail. l repair devices are available for restoring failed components. Such a system is denoted as <GI_k≤n|GI|l>. The research employed marked Markov processes and the theory of order statistics. Using the proposed approach, a mathematical system model was constructed, marks transformations were mapped and analytic expressions for calculating their distributions were given. In the following part of the paper, using the proposed method, a simulation algorithm will be defined for the purpose of assessing the key dependability characteristics. It will not only enable a numerical study of such systems, but will also help analyse the sensitivity of the dependability characteristics to the initial system parameters.  

Aim. The paper aims to improve the security of IoT devices by applying machine learning algorithms to detect attacks against IoT networks. The relevance of the goal is defined by the ever-growing number of such attacks around the world and the widespread use of IoT systems. The paper provides relevant statistical data. An analysis of the available papers showed that various methods were examined individually and were not compared to each other, so the aim of this paper that consists in identifying the most promising machine learning algorithm for detecting attacks against IoT networks is of relevance. Methods. The paper used the following machine learning methods to detect attacks against IoT networks: logistic regression, SVC, random forest, K-nearest neighbour method, k-means method, naive Bayes classifier, and variants of gradient boosting (XGBoost, AdaBoost, and CatBoost). The novelty consists in the comparison of the outputs of the supervised algorithms with the unsupervised K-means in the context of detection of attacks against IoT networks. The attack detection systems under development were trained using the UNSWNB15 dataset that contains data on nine types of attacks. The number of entries is more than 80 thousand. More than half of the entries deal with attacks. The methods were compared using a number of metrics. Results. An intrusion detection system was structurally defined and implemented. The stages of its operation include the analysis of input data and the output of final statistical data. The results show that the random forest algorithm is the best one out of those examined. The method also performs well in terms of learning speed. That means that the algorithm can be deployed and applied with the greatest success. Conclusions. This paper presents the results of comparing various machine learning algorithms in the context of IoT device intrusion detection. The accuracy and the ROC-AUC curve are used to evaluate the efficiency of the employed models. Having compared the models of the employed algorithms we found that the RandomForestClassifier model has the highest accuracy and a high AUC, which means that this algorithm is the most efficient in terms of IoT network intrusion detection. Further research will be dedicated to distinguishing between the types of attack.

Aim. To propose a methodological approach to ensuring the functional dependability of industrial facilities using SMART documents in the context of import-independent digital processes. Methods. The evolving applicability of the dependability theory and information systems design has defined the methodological provisions for algorithmising the application of SMART standards for minimising functional faults and failures of industrial processes. Digital models of production lines were examined as tools for machine-recognisable representation of standards. The proposed form of regulatory documents is intended to enable the transition from machine-readable data to machine-understandable content. Findings. The application of the proposed methods of content systematisation in the context of manufacturing process standardisation shows that the classification of process data based on standardised characteristics of digital information exchange processes is one of the key methodological provisions ensuring unambiguous interpretation of regulatory requirements. Specialised projects often use individual provisions rather than complete regulatory documents. In this case, in order to ensure machine-understandability of the developed content, it is proposed using special identifiers, e.g., «paragraph», «graphic object», «table cell». The deployment of such identification tools will allow creating a class of SMART standards that collect into a single document the provisions regarding the delivery of intended results in situations when the operating equipment is diverse. Publications created using intelligent processing of SMART documents are considered as containers of structured and unstructured data that take into account the conditions of particular projects. The existing dynamics of the demand for advanced industrial products often determines the admissibility of its industrial production by various companies with the parallel use of unique technologies. The conditions of such an organisation impose their limitations on the comparability of various process requirements, therefore, it is proposed to harmonise the presentation of the results of managerial, design, and process engineering solutions using the term «manufacturing system asset». The introduction of relative result estimates as part of the digital format is an efficient mechanism for ensuring management consistency. In this context, the importance of a uniform regulatory framework as the foundation for such assessments is growing. As one of the ways of minimising faults and failures of industrial equipment, it is proposed to create a situation centre with a decision support system based on the regulatory data with harmonised manufacturing system asset requirements. An experience of Industry 4.0 regulation is described in the review of the ISA digital manufacturing series of standards. It was shown that the deployment of SMART standards improves the stability of interaction between information management systems and the functional dependability of production lines. The introduction of provisions regulating the use of algorithms for the development, editing, and examination of draft regulatory documents into the content machine-understandability facilities is an important factor in ensuring the functional dependability of products. Conclusions. The presented approach is focused on the methodology for developing standards that take into account requirements from various subject areas. In the context of ongoing import substitution in manufacturing, the paper examines the feasibility of minimizing digital process faults and failures caused by the use of regulatory frameworks that are not equivalent to the provisions of international documents and take into account the specificity of Russian industrial systems.

Aim. The degradation of CMOS microcircuits exposed to ionizing radiation was analysed. Three MOS defect formation processes at the Si-SiO₂ boundary were examined. Methods. Using the example of test logic elements, the dependence of the time of conditional speed failure on the gamma dose rate was analysed. Findings. A critical defect at the Si-SiO₂ boundary was identified. Conclusions. The approach proposed in the paper allows identifying the causes of CMOS microcircuit failures. Calculated failure times for three dose rates are presented.