Pub Date : 2021-09-01DOI: 10.34229/1028-0979-2021-5-11
Andrey Litvynchuk, Lesya V. Baranovska
Face recognition is one of the main tasks of computer vision. It has many applications, which has led to a huge amount of research in this area. And although research in the field has been going on since the beginning of the computer vision, good results could be achieved only with the help of convolutional neural networks. In this work, a comparative analysis of facial recognition methods before convolutional neural networks was performed. A set of neural network architectures, methods of metric learning and optimization are considered. There were performed bunch of experiments and comparative analysis of the considered methods of improvement of convolutional neural networks. As a result a universal algorithm for training the face recognition model was obtained. To compare different approaches of face recognition, we chose a dataset called VGGFace2. It consists of 3,31 million images of 9131 people. It was created using images from the Google search engine. Initially, pre-trained neural networks were used to select photographs with humans. The images were then checked mannualy. For the validation sample, we set aside 50 images of 500 people, for a total of 25,000 images. Almost all experiments were performed iteratively. For example, we choose the best optimizer and then we use it to search for best arctitecture. As expected, neural networks with more parameters and more sophisticated architecture showed better results in this task. Among the considered models the best was Se-ResNet50. Metric learning is a method by which it is possible to achieve good accuracy in face recognition. Without this method it would be impossible to solve the problem. To optimize neural networks, we considered both adaptive and simple optimizers. It turned out that the stochastic gradient descent with moment is the best for this problem, and adaptive methods showed a rather poor result. In general, using different approaches, we were able to obtain an accuracy of 92 %, which is 25,5 % better than the baseline experiment. We see next ways for the further development of the research subject: improving neural network architecture, collecting more data and applying better regularization techniques.
{"title":"IMPROVING FACE RECOGNITION MODELS USING CONVOLUTIONAL NEURAL NETWORKS, METRIC LEARNING AND OPTIMIZATION METHOD","authors":"Andrey Litvynchuk, Lesya V. Baranovska","doi":"10.34229/1028-0979-2021-5-11","DOIUrl":"https://doi.org/10.34229/1028-0979-2021-5-11","url":null,"abstract":"Face recognition is one of the main tasks of computer vision. It has many applications, which has led to a huge amount of research in this area. And although research in the field has been going on since the beginning of the computer vision, good results could be achieved only with the help of convolutional neural networks. In this work, a comparative analysis of facial recognition methods before convolutional neural networks was performed. A set of neural network architectures, methods of metric learning and optimization are considered. There were performed bunch of experiments and comparative analysis of the considered methods of improvement of convolutional neural networks. As a result a universal algorithm for training the face recognition model was obtained. To compare different approaches of face recognition, we chose a dataset called VGGFace2. It consists of 3,31 million images of 9131 people. It was created using images from the Google search engine. Initially, pre-trained neural networks were used to select photographs with humans. The images were then checked mannualy. For the validation sample, we set aside 50 images of 500 people, for a total of 25,000 images. Almost all experiments were performed iteratively. For example, we choose the best optimizer and then we use it to search for best arctitecture. As expected, neural networks with more parameters and more sophisticated architecture showed better results in this task. Among the considered models the best was Se-ResNet50. Metric learning is a method by which it is possible to achieve good accuracy in face recognition. Without this method it would be impossible to solve the problem. To optimize neural networks, we considered both adaptive and simple optimizers. It turned out that the stochastic gradient descent with moment is the best for this problem, and adaptive methods showed a rather poor result. In general, using different approaches, we were able to obtain an accuracy of 92 %, which is 25,5 % better than the baseline experiment. We see next ways for the further development of the research subject: improving neural network architecture, collecting more data and applying better regularization techniques.","PeriodicalId":54874,"journal":{"name":"Journal of Automation and Information Sciences","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2021-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"48058042","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-09-01DOI: 10.34229/1028-0979-2021-5-4
Valery Severyn, E. Nikulina
The structure of information technology for modeling control systems, which includes a block of systems models, a module of integration methods and other program elements, is considered. To analyze the dynamics of control of a nuclear reactor, programs of mathematical models of a WWER-1000 nuclear reactor of the V-320 series and its control systems in the form of nonlinear systems of differential equations in the Cauchy form have been developed. For the integration of nonlinear systems of differential equations, an algorithm of the system method of the first degree is presented. A mathematical model of a WWER-1000 reactor as a control object with division into zones along the vertical axis in relative variables of state is considered, the values of the constant parameters of the model and the initial conditions corresponding to the nominal mode are given. Using information technology for ten zones of the reactor, the system integration method was used to simulate the dynamics of control of a nuclear reactor. Graphs of neutron and thermal processes in the reactor core, as well as changes in the axial offset when the reactor load is dumped under the influence of the movement of absorbing rods and an increase in the concentration of boric acid, are plotted. The analysis of dynamic processes of reactor control is carried out. The programs of integration methods and models of the WWER-1000 reactor of the V-320 series are included in the information technology to optimize the maneuvering modes of the reactor.
{"title":"APPLICATION OF INFORMATION TECHNOLOGY FOR MODELING THE CONTROL DYNAMICS OF A NUCLEAR REACTOR ZONING ON THE VERTICAL AXIS","authors":"Valery Severyn, E. Nikulina","doi":"10.34229/1028-0979-2021-5-4","DOIUrl":"https://doi.org/10.34229/1028-0979-2021-5-4","url":null,"abstract":"The structure of information technology for modeling control systems, which includes a block of systems models, a module of integration methods and other program elements, is considered. To analyze the dynamics of control of a nuclear reactor, programs of mathematical models of a WWER-1000 nuclear reactor of the V-320 series and its control systems in the form of nonlinear systems of differential equations in the Cauchy form have been developed. For the integration of nonlinear systems of differential equations, an algorithm of the system method of the first degree is presented. A mathematical model of a WWER-1000 reactor as a control object with division into zones along the vertical axis in relative variables of state is considered, the values of the constant parameters of the model and the initial conditions corresponding to the nominal mode are given. Using information technology for ten zones of the reactor, the system integration method was used to simulate the dynamics of control of a nuclear reactor. Graphs of neutron and thermal processes in the reactor core, as well as changes in the axial offset when the reactor load is dumped under the influence of the movement of absorbing rods and an increase in the concentration of boric acid, are plotted. The analysis of dynamic processes of reactor control is carried out. The programs of integration methods and models of the WWER-1000 reactor of the V-320 series are included in the information technology to optimize the maneuvering modes of the reactor.","PeriodicalId":54874,"journal":{"name":"Journal of Automation and Information Sciences","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2021-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"45774978","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-09-01DOI: 10.34229/1028-0979-2021-5-2
V. Soloviev, O. Rybalsky, V. Zhuravel, A. Shablya, Evgeny Timko
When testing the most advanced speaker identification systems on specialized databases, their minimum efficiency, estimated by the error probability at the point of intersection of the error curves, is only a few percent. However, many factors are known that affect the variability of the characteristics of the speaker's voice, each of which has its own, different from the others, influence on the results of the speaker's identification by the characteristics of the voice. The complexity of creating and testing speaker identification systems is the need to quantitatively formalize a number of specific factors that affect the characteristics of his voice. The article discusses the proposed method for accounting for a variety of factors affecting the parameters of the characteristics of the speaker's voice, which provides the fundamental possibility of indirectly accounting for their practically unlimited number. According to this method, «atomic» structures are distinguished from speech signals, which depend on the totality of the main factors that affect the speaker's identification process. With this method, all significant factors affecting the characteristics of the voice will be indirectly taken into account at the level of these structures. Subsequent decisions are made on the combinatorial set of a huge number of these «atomic» structures. «Atomic» speech structures are understood as the spectra of any fragments of any vowel sounds allocated in a time window of 20 ms. «Atomic» structures are selected automatically. The proposed method provides a rational consideration of the multifactorial influence of various parameters, since the spectra of these structures are influenced by all the main factors that characterize the individuality of the voice of a particular speaker. The decision on the identity of the voices of the announcers recorded on different phonograms is carried out on the basis of combinatorics of «atomic» spectra of vowel sounds in both phonograms. The method has shown high efficiency in the examination of phonograms of short duration.
{"title":"TAKING INTO ACCOUNT THE MULTIFACTORIAL CHARACTER OF VOICE CHARACTERISTICS IN THE PROBLEMS OF SPEAKER IDENTIFICATION","authors":"V. Soloviev, O. Rybalsky, V. Zhuravel, A. Shablya, Evgeny Timko","doi":"10.34229/1028-0979-2021-5-2","DOIUrl":"https://doi.org/10.34229/1028-0979-2021-5-2","url":null,"abstract":"When testing the most advanced speaker identification systems on specialized databases, their minimum efficiency, estimated by the error probability at the point of intersection of the error curves, is only a few percent. However, many factors are known that affect the variability of the characteristics of the speaker's voice, each of which has its own, different from the others, influence on the results of the speaker's identification by the characteristics of the voice. The complexity of creating and testing speaker identification systems is the need to quantitatively formalize a number of specific factors that affect the characteristics of his voice. The article discusses the proposed method for accounting for a variety of factors affecting the parameters of the characteristics of the speaker's voice, which provides the fundamental possibility of indirectly accounting for their practically unlimited number. According to this method, «atomic» structures are distinguished from speech signals, which depend on the totality of the main factors that affect the speaker's identification process. With this method, all significant factors affecting the characteristics of the voice will be indirectly taken into account at the level of these structures. Subsequent decisions are made on the combinatorial set of a huge number of these «atomic» structures. «Atomic» speech structures are understood as the spectra of any fragments of any vowel sounds allocated in a time window of 20 ms. «Atomic» structures are selected automatically. The proposed method provides a rational consideration of the multifactorial influence of various parameters, since the spectra of these structures are influenced by all the main factors that characterize the individuality of the voice of a particular speaker. The decision on the identity of the voices of the announcers recorded on different phonograms is carried out on the basis of combinatorics of «atomic» spectra of vowel sounds in both phonograms. The method has shown high efficiency in the examination of phonograms of short duration.","PeriodicalId":54874,"journal":{"name":"Journal of Automation and Information Sciences","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2021-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"44561546","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-09-01DOI: 10.34229/1028-0979-2021-5-5
Arkady Chikriі, I. Rappoport
The problem of a guaranteed result in game problems of group approach of controlled objects is considered. A method for solving such problems is proposed, which is associated with the construction of some scalar functions that qualitatively characterize the progress of the approach of a group of controlled objects and the efficiency of the decisions made. Such functions are called resolving functions. The attractiveness of the method of resolving functions lies in the fact that it makes it possible to use effectively the modern technique of multivalued mappings and their selection in substantiating game constructions and obtaining meaningful results on their basis. In any form of the method of resolving functions, the main principle is the accumulative principle, which is used in the current summation of the resolving functions to assess the quality of the game of the group approach until a certain threshold value is reached. In contrast to the main scheme of the mentioned method, the case is considered when the classical Pontryagin condition does not hold. In this situation, instead of Pontryagin’s selection, which do not exist, some shift functions are considered and, with their help, special multivalued mappings are introduced. They generate upper and lower resolving functions with the help of which sufficient conditions for the completion of the game of group approach in a certain guaranteed time are formulated. Comparison of guaranteed times for different schemes of group approach of controlled objects is given.
{"title":"GUARANTEED RESULT IN GROUP APPROACH GAME PROBLEMS OF CONTROLLED OBJECTS","authors":"Arkady Chikriі, I. Rappoport","doi":"10.34229/1028-0979-2021-5-5","DOIUrl":"https://doi.org/10.34229/1028-0979-2021-5-5","url":null,"abstract":"The problem of a guaranteed result in game problems of group approach of controlled objects is considered. A method for solving such problems is proposed, which is associated with the construction of some scalar functions that qualitatively characterize the progress of the approach of a group of controlled objects and the efficiency of the decisions made. Such functions are called resolving functions. The attractiveness of the method of resolving functions lies in the fact that it makes it possible to use effectively the modern technique of multivalued mappings and their selection in substantiating game constructions and obtaining meaningful results on their basis. In any form of the method of resolving functions, the main principle is the accumulative principle, which is used in the current summation of the resolving functions to assess the quality of the game of the group approach until a certain threshold value is reached. In contrast to the main scheme of the mentioned method, the case is considered when the classical Pontryagin condition does not hold. In this situation, instead of Pontryagin’s selection, which do not exist, some shift functions are considered and, with their help, special multivalued mappings are introduced. They generate upper and lower resolving functions with the help of which sufficient conditions for the completion of the game of group approach in a certain guaranteed time are formulated. Comparison of guaranteed times for different schemes of group approach of controlled objects is given.","PeriodicalId":54874,"journal":{"name":"Journal of Automation and Information Sciences","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2021-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"49162738","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-09-01DOI: 10.34229/1028-0979-2021-5-1
P. Stetsyuk, O. Lykhovyd, V. Zhydkov, A. Suprun
Mathematical models of two classes of problems of modernization of the capacity of arcs of fault-tolerant oriented networks are considered. A network is considered to be fault-tolerant for which it is possible to satisfy all the demands for the transmission of flows when there will be one, but any failure, from all possible single network failures. For the first class of problems (problem A), all possible paths in the network can be used for the transmission of flows. For the second class of problems (problem P), only paths from a predetermined set of paths are used to transfer flows. Mathematical models are represented by linear, Boolean and nonlinear programming problems with a block structure of the constraint matrix.The material of the article is presented in five sections. The first section describes the concepts of a single failure and the scenario of network failures, the content of optimization problems A and P for modernization of capacity of arcs of a fault-tolerant network, a test network (6 vertices and 19 arcs) to test algorithms for solving the problems of modernization of fault-tolerant networks. In the second section, basic models of linear programming problems for finding the capacities of arcs of the fault-tolerant physical structure of a network (problem A) and the fault-tolerant logical structure of a network (problem P) are described, and their properties are considered. The third section describes problems A and P in the form of mixed Boolean linear programming models. Optimal solutions of problem A for various failure scenarios are given for the example of the test network. The solutions were found using the Gurobi program from the NEOS server, where the mathematical model of problem A is described in the AMPL modeling language.The fourth section describes nonlinear convex programming models for problems A and P, developed to find the optimal capacities of fault-tolerant networks according to the selected criterion, and a decomposition algorithm for their solution. The fifth section describes software in the FORTRAN programming language for the decomposition algorithm based on efficient implementations of Shor’s r-algorithms. The decomposition algorithm is compared with the IPOPT program based on the results of solving test problems.
{"title":"OPTIMIZATION PROBLEMS OF MODERNIZATION OF THE CAPACITY OF ARCS OF FAULT-TOLERANT NETWORKS","authors":"P. Stetsyuk, O. Lykhovyd, V. Zhydkov, A. Suprun","doi":"10.34229/1028-0979-2021-5-1","DOIUrl":"https://doi.org/10.34229/1028-0979-2021-5-1","url":null,"abstract":"Mathematical models of two classes of problems of modernization of the capacity of arcs of fault-tolerant oriented networks are considered. A network is considered to be fault-tolerant for which it is possible to satisfy all the demands for the transmission of flows when there will be one, but any failure, from all possible single network failures. For the first class of problems (problem A), all possible paths in the network can be used for the transmission of flows. For the second class of problems (problem P), only paths from a predetermined set of paths are used to transfer flows. Mathematical models are represented by linear, Boolean and nonlinear programming problems with a block structure of the constraint matrix.The material of the article is presented in five sections. The first section describes the concepts of a single failure and the scenario of network failures, the content of optimization problems A and P for modernization of capacity of arcs of a fault-tolerant network, a test network (6 vertices and 19 arcs) to test algorithms for solving the problems of modernization of fault-tolerant networks. In the second section, basic models of linear programming problems for finding the capacities of arcs of the fault-tolerant physical structure of a network (problem A) and the fault-tolerant logical structure of a network (problem P) are described, and their properties are considered. The third section describes problems A and P in the form of mixed Boolean linear programming models. Optimal solutions of problem A for various failure scenarios are given for the example of the test network. The solutions were found using the Gurobi program from the NEOS server, where the mathematical model of problem A is described in the AMPL modeling language.The fourth section describes nonlinear convex programming models for problems A and P, developed to find the optimal capacities of fault-tolerant networks according to the selected criterion, and a decomposition algorithm for their solution. The fifth section describes software in the FORTRAN programming language for the decomposition algorithm based on efficient implementations of Shor’s r-algorithms. The decomposition algorithm is compared with the IPOPT program based on the results of solving test problems.","PeriodicalId":54874,"journal":{"name":"Journal of Automation and Information Sciences","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2021-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"46351411","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-09-01DOI: 10.34229/1028-0979-2021-5-10
V. Volosov, V. Shevchenko
A general structure of the kinematic equations for attitude evolution of a spacecraft (SC) (coordinate system associated with a spacecraft (SCS)) relative to the reference coordinate system (RCS) is proposed. It is assumed that the origins of the coordinate systems coincide and are located at an arbitrary point of the spacecraft. Each of the coordinate systems rotates at an arbitrary absolute angular velocity (relative to the inertial space) specified by the projections on their axes. Attitude parameters can be the Euler–Krylov angles, Rodrigues–Hamilton parameters, and modified Rodrigues parameters. It is shown that the well-known representations of the attitude evolution equations of the SCS relative to the RCS using the Rodrigues-Hamilton parameters (components of normalized quaternions) can be simply obtained from the solution of the Erugin problem of finding the entire set of differential equations with a given integral of motion. The advantages and disadvantages of use for each of the specified attitude parameters are considered. A method of attitude control synthesis is proposed which is common for all these equations and based on the decomposition of the original problem into kinematic and dynamic ones and the use of well-known generalizations of the direct Lyapunov method for their solution. The property of structural roughness according to Andronov–Pontryagin [27–29] of the obtained algorithm is illustrated with the help of computer simulation. Particularly, a specific example illustrates the possibility for even a structurally simplified algorithm of stabilizing a specified constant spacecraft attitude to track the program of its change with sufficient accuracy. The tracking task is typical for the control of spacecraft docking, spacecraft de-orbiting, and performing route surveys of the Earth's surface.
{"title":"MATHEMATICAL MODELS OF ANGULAR MOTION OF SPACE VEHICLES AND THEIR USE IN ORIENTATION CONTROL PROBLEMS","authors":"V. Volosov, V. Shevchenko","doi":"10.34229/1028-0979-2021-5-10","DOIUrl":"https://doi.org/10.34229/1028-0979-2021-5-10","url":null,"abstract":"A general structure of the kinematic equations for attitude evolution of a spacecraft (SC) (coordinate system associated with a spacecraft (SCS)) relative to the reference coordinate system (RCS) is proposed. It is assumed that the origins of the coordinate systems coincide and are located at an arbitrary point of the spacecraft. Each of the coordinate systems rotates at an arbitrary absolute angular velocity (relative to the inertial space) specified by the projections on their axes. Attitude parameters can be the Euler–Krylov angles, Rodrigues–Hamilton parameters, and modified Rodrigues parameters. It is shown that the well-known representations of the attitude evolution equations of the SCS relative to the RCS using the Rodrigues-Hamilton parameters (components of normalized quaternions) can be simply obtained from the solution of the Erugin problem of finding the entire set of differential equations with a given integral of motion. The advantages and disadvantages of use for each of the specified attitude parameters are considered. A method of attitude control synthesis is proposed which is common for all these equations and based on the decomposition of the original problem into kinematic and dynamic ones and the use of well-known generalizations of the direct Lyapunov method for their solution. The property of structural roughness according to Andronov–Pontryagin [27–29] of the obtained algorithm is illustrated with the help of computer simulation. Particularly, a specific example illustrates the possibility for even a structurally simplified algorithm of stabilizing a specified constant spacecraft attitude to track the program of its change with sufficient accuracy. The tracking task is typical for the control of spacecraft docking, spacecraft de-orbiting, and performing route surveys of the Earth's surface.","PeriodicalId":54874,"journal":{"name":"Journal of Automation and Information Sciences","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2021-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"42333552","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-09-01DOI: 10.34229/1028-0979-2021-5-3
Kirill Kadomskіy
In Industrial IoT (IIoT) systems, timed automata provide a highly useful abstraction for diagnosis and control tasks. Applying them requires automaton to be learned in passive online manner using positive samples only. Such kind of learning is supported by Hybrid timed Automata (HTA) and algorithm OTALA, but requireds a sequence of discrete events rather than continuous analog time series typically found in IIoT. Recent attempts to cover this gap, taken by A. von Birgelen, O. Niggemann, and others, involved pre-processing observations with a self-organized map (SOM) and watershed transform, yet resulting models have proven ineffective in some real-world systems. In this paper, incremental model-based clustering (IMCF) is employed to learn timed automaton from analog IIoT data. IMCF is a sequential algorithm that processes observed time-series online and splits them into a sequence of discrete states with either crisp or fuzzy transitions between them. Such transitions are then treated as events required for HTA identification with OTALA. Obtained models are evaluated in a case of IIoT system that has proved to be challenging for existing modelling techniques. Experimental results show 24,9–76,8% increase in model’s performance and suggest that discretizing obtained with IMCF has higher informativeness for HTA identification. Finally, wider perspectives of applying HTA in IIoT are discussed, and remaining principal limitations are identified as discrete nature of state transitions, and lack of long-term memory for transitions.
在工业物联网(IIoT)系统中，时间自动机为诊断和控制任务提供了非常有用的抽象。应用它们需要自动机以被动在线的方式学习，只使用正样本。这种学习由混合时间自动机(HTA)和OTALA算法支持，但需要一系列离散事件，而不是IIoT中常见的连续模拟时间序列。最近，a . von Birgelen、O. Niggemann等人尝试用自组织地图(SOM)和分水岭变换对观测结果进行预处理，但结果模型在一些现实世界的系统中被证明是无效的。本文采用基于增量模型的聚类(IMCF)从模拟IIoT数据中学习时间自动机。IMCF是一种顺序算法，它在线处理观察到的时间序列，并将它们分成一系列离散状态，它们之间有清晰或模糊的过渡。然后将此类转换视为与OTALA进行HTA识别所需的事件。在工业物联网系统的情况下评估获得的模型，该系统已被证明对现有建模技术具有挑战性。实验结果表明，模型性能提高了24.9 ~ 76.8%，表明用IMCF进行离散化对HTA识别具有较高的信息量。最后，讨论了在工业物联网中应用HTA的更广泛视角，并将剩余的主要限制确定为状态转换的离散性质，以及缺乏对转换的长期记忆。
{"title":"USING SEQUENTIAL CLUSTERING TO IDENTIFICATION OF HYBRID TIMED AUTOMATA FROM ANALOG IIOT DATA","authors":"Kirill Kadomskіy","doi":"10.34229/1028-0979-2021-5-3","DOIUrl":"https://doi.org/10.34229/1028-0979-2021-5-3","url":null,"abstract":"In Industrial IoT (IIoT) systems, timed automata provide a highly useful abstraction for diagnosis and control tasks. Applying them requires automaton to be learned in passive online manner using positive samples only. Such kind of learning is supported by Hybrid timed Automata (HTA) and algorithm OTALA, but requireds a sequence of discrete events rather than continuous analog time series typically found in IIoT. Recent attempts to cover this gap, taken by A. von Birgelen, O. Niggemann, and others, involved pre-processing observations with a self-organized map (SOM) and watershed transform, yet resulting models have proven ineffective in some real-world systems. In this paper, incremental model-based clustering (IMCF) is employed to learn timed automaton from analog IIoT data. IMCF is a sequential algorithm that processes observed time-series online and splits them into a sequence of discrete states with either crisp or fuzzy transitions between them. Such transitions are then treated as events required for HTA identification with OTALA. Obtained models are evaluated in a case of IIoT system that has proved to be challenging for existing modelling techniques. Experimental results show 24,9–76,8% increase in model’s performance and suggest that discretizing obtained with IMCF has higher informativeness for HTA identification. Finally, wider perspectives of applying HTA in IIoT are discussed, and remaining principal limitations are identified as discrete nature of state transitions, and lack of long-term memory for transitions.","PeriodicalId":54874,"journal":{"name":"Journal of Automation and Information Sciences","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2021-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"47956104","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-07-01DOI: 10.34229/1028-0979-2021-4-9
Valentin Bahatskyi, Aleksey Bahatskyi
Currently, the measurement of electrical and non-electrical quantities is performed using analog-to-digital conversion channels, which consist of analog signal conditioning circuits and analog-to-digital converters (ADC) of electrical quantities into a digital code. The paper considers the case when the defining errors of the measurement and control channel are systematic errors of the ADC. The reliability of measurements is assessed by their errors, and the reliability of control - by the likelihood of correct operation of the control device. In our opinion, evaluating the reliability of such similar processes as measurement and control using different criteria seems illogical. The aim of the work is to study the effect of systematic errors of an analog-to-digital converter on the errors of parameter control depending on the type of conformity functions and the width of the control window, as well as the choice of the resolution of the ADC for various control tasks. The paper analyzes the transfer functions of measurement and control. It is shown that they are formed using step functions. It is proposed to use not a step function as a control transfer function, but other functions of conformity to the norm, for example, a linear function or functions of higher orders. In this case, the control result is assessed not according to the criterion of the probability of correct operation, but using the control error. Analyzed from the point of view of reconfiguring the errors of the line, parabolic and state parabolic functions of the norms for the development of changes windows in control. A recommendation has been given for the selection of functions for the conformity of standards and for the distribution of analog-to-digital conversions for industrial control enterprises.
{"title":"ERRORS CONVERSION ON TRANSFER FUNCTIONS OF MEASUREMENT AND CONTROL","authors":"Valentin Bahatskyi, Aleksey Bahatskyi","doi":"10.34229/1028-0979-2021-4-9","DOIUrl":"https://doi.org/10.34229/1028-0979-2021-4-9","url":null,"abstract":"Currently, the measurement of electrical and non-electrical quantities is performed using analog-to-digital conversion channels, which consist of analog signal conditioning circuits and analog-to-digital converters (ADC) of electrical quantities into a digital code. The paper considers the case when the defining errors of the measurement and control channel are systematic errors of the ADC. The reliability of measurements is assessed by their errors, and the reliability of control - by the likelihood of correct operation of the control device. In our opinion, evaluating the reliability of such similar processes as measurement and control using different criteria seems illogical. The aim of the work is to study the effect of systematic errors of an analog-to-digital converter on the errors of parameter control depending on the type of conformity functions and the width of the control window, as well as the choice of the resolution of the ADC for various control tasks. The paper analyzes the transfer functions of measurement and control. It is shown that they are formed using step functions. It is proposed to use not a step function as a control transfer function, but other functions of conformity to the norm, for example, a linear function or functions of higher orders. In this case, the control result is assessed not according to the criterion of the probability of correct operation, but using the control error. Analyzed from the point of view of reconfiguring the errors of the line, parabolic and state parabolic functions of the norms for the development of changes windows in control. A recommendation has been given for the selection of functions for the conformity of standards and for the distribution of analog-to-digital conversions for industrial control enterprises.","PeriodicalId":54874,"journal":{"name":"Journal of Automation and Information Sciences","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2021-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"44818713","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-07-01DOI: 10.34229/1028-0979-2021-4-5
Elchin Aliyev, E. Aliyev, A. Ali
A comprehensive analysis of the applicantʼs solvency for obtaining a microcredit precedes the conclusion of a loan agreement with him. This allows to determine the risk factors associated with the possibility of non-repayment of a bank loan in due time, and, on the contrary, to assess the likelihood of timely repayment of the loan. Therefore, the assessment of the clientʼs creditworthiness is an integral part of the work of commercial banks and microfinance organizations to determine the possibility of issuing microloans to one or another applicant. The paper proposes a balanced approach to the multi-criteria assessment of the solvency of individuals, based, among other things, on a fuzzy analysis of their solvency indicators. The developed fuzzy inference system in combination with statistical methods for assessing solvency, can serve as an analytical core for a credit decision support system. Based on the example of ten hypothetical alternative borrowers, characterized by their current indicators, the corresponding assessments of their solvency were made, including scoring, Pareto method, Bord method and using a fuzzy inference system. Such a combined approach is distinguished by the ability to identify reliably a group of individuals with high credit discipline and the characteristics of those in relation to whom credit decisions are classified as high-risk.
{"title":"ANALYTICAL SUPPORT OF THE MICROCREDITING PROCESS USING A FUZZY INFERENCE SYSTEM","authors":"Elchin Aliyev, E. Aliyev, A. Ali","doi":"10.34229/1028-0979-2021-4-5","DOIUrl":"https://doi.org/10.34229/1028-0979-2021-4-5","url":null,"abstract":"A comprehensive analysis of the applicantʼs solvency for obtaining a microcredit precedes the conclusion of a loan agreement with him. This allows to determine the risk factors associated with the possibility of non-repayment of a bank loan in due time, and, on the contrary, to assess the likelihood of timely repayment of the loan. Therefore, the assessment of the clientʼs creditworthiness is an integral part of the work of commercial banks and microfinance organizations to determine the possibility of issuing microloans to one or another applicant. The paper proposes a balanced approach to the multi-criteria assessment of the solvency of individuals, based, among other things, on a fuzzy analysis of their solvency indicators. The developed fuzzy inference system in combination with statistical methods for assessing solvency, can serve as an analytical core for a credit decision support system. Based on the example of ten hypothetical alternative borrowers, characterized by their current indicators, the corresponding assessments of their solvency were made, including scoring, Pareto method, Bord method and using a fuzzy inference system. Such a combined approach is distinguished by the ability to identify reliably a group of individuals with high credit discipline and the characteristics of those in relation to whom credit decisions are classified as high-risk.","PeriodicalId":54874,"journal":{"name":"Journal of Automation and Information Sciences","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2021-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"42546204","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-07-01DOI: 10.34229/1028-0979-2021-4-10
Valeriy Severyn, E. Nikulina
Mathematical models of the WWER-1000 nuclear power reactor have been developed with division into zones along the vertical axis in the form of nonlinear systems of differential equations with dimensionless relative state variables. Models in a given number of zones along the vertical axis represent neutron kinetics, gradual heat release, thermal processes in fuel, cladding and coolant, changes in the concentration of iodine, xenon and boron. The parameters of mathematical models have been calculated based on the design and technological parameters of the V-320 series nuclear reactor. A general model of the reactor as a control object with division into zones along the vertical axis, as well as models with control of absorbing rods and boric acid, are obtained. Integration of the obtained systems of differential equations for given initial conditions allows one to obtain changes in all state variables in the reactor zones along the vertical axis. In particular, from the change in power in the zones along the vertical axis, the axial offset is calculated as the relative value of the difference between the powers of the upper and lower halves of the reactor core. The developed reactor models with dimensionless relative state variables use a minimum number of calculations, allow calculating the change in the axial offset, and are included in the information technology for controlling the power units of nuclear power plants to optimize the maneuvering modes of the WWER-1000 V-320 series reactor.
{"title":"MODELS OF WWER-1000 NUCLEAR REACTOR WITH DIVISION INTO ZONES ON VERTICAL AXIS FOR INFORMATION TECHNOLOGY OF CONTROL","authors":"Valeriy Severyn, E. Nikulina","doi":"10.34229/1028-0979-2021-4-10","DOIUrl":"https://doi.org/10.34229/1028-0979-2021-4-10","url":null,"abstract":"Mathematical models of the WWER-1000 nuclear power reactor have been developed with division into zones along the vertical axis in the form of nonlinear systems of differential equations with dimensionless relative state variables. Models in a given number of zones along the vertical axis represent neutron kinetics, gradual heat release, thermal processes in fuel, cladding and coolant, changes in the concentration of iodine, xenon and boron. The parameters of mathematical models have been calculated based on the design and technological parameters of the V-320 series nuclear reactor. A general model of the reactor as a control object with division into zones along the vertical axis, as well as models with control of absorbing rods and boric acid, are obtained. Integration of the obtained systems of differential equations for given initial conditions allows one to obtain changes in all state variables in the reactor zones along the vertical axis. In particular, from the change in power in the zones along the vertical axis, the axial offset is calculated as the relative value of the difference between the powers of the upper and lower halves of the reactor core. The developed reactor models with dimensionless relative state variables use a minimum number of calculations, allow calculating the change in the axial offset, and are included in the information technology for controlling the power units of nuclear power plants to optimize the maneuvering modes of the WWER-1000 V-320 series reactor.","PeriodicalId":54874,"journal":{"name":"Journal of Automation and Information Sciences","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2021-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"48129422","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：