+7 (495) 987 43 74 ext. 3304
Join us -              
Рус   |   Eng

Authors

Puchkov A.

Degree
Cand. Sci. (Eng.), Associate Professor, Information Technologies in Economics and Management Department, Branch of the National Research University “MPEI” in Smolensk
E-mail
putchkov63@mail.ru
Location
Smolensk, Russia
Articles

Algorithms for the formation of images of the states of objects for their analysis by deep neural networks

Algorithms of visualization of numerical data characterizing the state of objects and systems of various nature with the aim of finding hidden patterns in them using convolutional neural networks are presented. The algorithms used methods for obtaining images from numerical data on the basis of the discrete Fourier transform of time series fragments, as well as on the basis of the application of visualization using three-component system diagrams, if such a three-component representation of the system is possible. The software implementation of the proposed algorithms was performed in the Linux environment in the Python 3 language using the Keras open neural network library, which is a superstructure above the TensorFlow machine learning framework. For the learning process of the neural network, a Nvidia graphics processor was used that supports the technology of the CUDA parallel computing software and hardware architecture, which significantly reduced the learning time. The proposed approach is the recognition States of the objects according to their visualized data are based on the recognition of no boundaries or forms of the figures in the images and their textures. Also presented is a program that generates sets of images to implement the process of learning and testing convolutional neural networks in order to pre-tune them and assess the quality of the proposed algorithms.Keywords: Internet, Internet security, parental control applications, user security, information security, Internet threats.
Read more...

Preliminary assessment of the pragmatic value of information in the classifiсation problem based on deep neural networks

A method is proposed for preliminary assessment of the pragmatic value of information in the problem of classifying the state of an object based on deep recurrent networks of long short-term memory. The purpose of the study is to develop a method for predicting the state of a controlled object while minimizing the number of used prognostic parameters through a preliminary assessment of the pragmatic value of information. This is an especially urgent task under conditions of processing big data, characterized not only by significant volumes of incoming information, but also by information rate and multiformatness. The generation of big data is now happening in almost all areas of activity due to the widespread introduction of the Internet of Things in them. The method is implemented by a two-level scheme for processing input information. At the first level, a Random Forest machine learning algorithm is used, which has significantly fewer adjustable parameters than a recurrent neural network used at the second level for the final and more accurate classification of the state of the controlled object or process. The choice of Random Forest is due to its ability to assess the importance of variables in regression and classification problems. This is used in determining the pragmatic value of the input information at the first level of the data processing scheme. For this purpose, a parameter is selected that reflects the specified value in some sense, and based on the ranking of the input variables by the level of importance, they are selected to form training datasets for the recurrent network. The algorithm of the proposed data processing method with a preliminary assessment of the pragmatic value of information is implemented in a program in the MatLAB language, and it has shown its efficiency in an experiment on model data. Read more...

Algorithm for predicting the parameters of a system for processing waste apatite-nepheline ores

Within the framework of the concept of a circular economy, research in the field of creating technological systems for recycling waste from mining and processing plants occupies one of the key positions. This is connected, on the one hand, with significant volumes of such waste, reaching tens of millions of tons and posing a significant environmental hazard to air and water basins, human health, and, on the other hand, with their rich chemical and mineralogical composition, which makes it possible to call them accumulations of technogenic deposits. In this regard, the task of creating control systems for technological processes of processing such waste and their information support, including support for all stages of the passage of information processes, is urgent. The novelty of the presented research lies in the proposed structure of an intelligent control system for a complex chemical and energy technological system for processing apatite-nepheline ores, as well as in an algorithm for predicting technological parameters, which is part of the information support of the control system under consideration. The algorithm is based on the use of the apparatus of deep recurrent neural networks and Kalman filtering, which is used at the stage of data preprocessing to train the neural network. The paper describes the proposed algorithm for predicting multidimensional time series, adapted to the considered technological process, presents the software executed in the MatLab environment to demonstrate the efficiency of the specified combination of methods for processing technological parameters. In a model experiment, it has been shown that the use of filtering makes it possible to increase the accuracy of the forecast, which is especially noticeable at its large horizons. The practical significance of the research results is the proposed structure of an intelligent control system for the processing of apatite-nepheline ore waste and software for predicting its parameters, which can be used in various decision support systems. Read more...

Neural network analysis method of heat treatment processes of pelletized phosphate ore raw materials

Currently, there is an acute problem of waste disposal of mining and processing plants, which accumulate in significant volumes in the territories adjacent to them and pose a serious threat to the environment. In this regard, the creation of technological systems for processing ore waste and the improvement of their information support represent an urgent area of research. An example of such a system is a complex chemical and energy technology system for the production of yellow phosphorus from waste apatite-nepheline ores. The purpose of the study was to develop a model for collecting data on the parameters of the processes of heat treatment of pelletized phosphate ore raw materials in such a system, as well as a method for identifying dependencies between these parameters. The identification of dependencies in the information support of the yellow phosphorus production system will improve the quality of its functioning in terms of management criteria, energy and resource efficiency. To achieve this goal, the tasks of choosing a mathematical concept for the basis of the method being developed, constructing an algorithm and creating software implementing this method, conducting model experiments were solved. The method is based on the use of deep recurrent neural networks of long-term short-term memory, which have a high generalizing ability and are used in solving problems of regression and classification of multidimensional time sequences, in the form of which, as a rule, the parameters of a chemical and energy technology system are presented. The method is implemented as an application created in the MatLab 2021 environment. The application interface allows you to interactively conduct experiments with various sets of input and output parameters to identify the relationship between them, as well as change the hyperparameters of neural networks. As a result of the application, a repository of trained neural networks is created that simulate the relationships found between the specified parameters of the technological system and can be applied in decision support systems, management and engineering. Read more...

Multilevel algorithms for evaluating and making decisions on the optimal control of an integrated system for processing fine ore raw materials

The results of studies aimed at developing multi-level decision-making algorithms for management of energy and resource efficiency, technogenic and environmental safety of a complex multi-stage system for processing fine ore raw materials are presented (MSPFORM). A distinctive feature of such a system is its multidimensionality and multiscale, which manifests itself in the presence of two options for implementing technological processes for processing finely dispersed ore raw materials, the need to take into account the interaction of the aggregates included in the system, as well as the hierarchy of describing the processes occurring in them - mechanical, thermophysical, hydrodynamic, physical and chemical. Such a variety of processes characterizes the interdisciplinarity of research and the complexity of obtaining analytical, interconnected mathematical models. This situation inspired the analyze use of artificial intelligence methods, such as deep machine learning and fuzzy logic, to describe and analyze processes. The scientific component of the research results consists in the developed generalized structure of the MSPFORM, the conceptual basis of multilevel algorithms for evaluating and making decisions on the optimal control of this system, the proposed composition of the parameters and the form of the optimization criterion. The task of the study was to analyze possible options for the processing of ore raw materials, to develop a concept for the construction of the MSPFORM allowing the possibility of optimizing its functioning according to the criterion of energy and resource efficiency while meeting the requirements of environmental safety. The application of evolutionary algorithms for solving the problem of optimizing the MSPFORM according to the criterion of minimum energy consumption is announced and its stages are specified. The structure of the block of neuro-fuzzy analysis of information about the parameters of processes in MSPFORM is presented, which is based on the use of deep recurrent and convolutional neural networks, as well as a fuzzy inference system. The results of a simulation experiment on approbation of the software implementation of this block in the MatLab environment are presented. Read more...

An intelligent model for managing the risks of violation of the characteristics of electromechanical devices in a multi-stage system for processing ore raw materials

The results of studies on the development of the structure of an intelligent model for managing the risks of violation of the characteristics of electromechanical devices in a multi-stage system for processing ore raw materials are presented. Such devices are involved in all cycles of the technological process, so the assessment of this risk for them is an urgent task. A method for assessing such risks is proposed, which is based on the assessment of the useful life of equipment, performed on the basis of the prediction of characteristics by a deep recurrent neural network, with further generalization of the results of such an assessment in a fuzzy inference block. Recurrent neural networks with long short-term memory were used, which are one of the most powerful tools for solving time series regression problems, including predicting their values for long intervals. The use of deep neural networks to predict the characteristics of electromechanical devices made it possible to obtain a high prediction accuracy, which made it possible to apply a relatively less accurate recurrent least squares method for the iterative process of estimating the useful life of equipment. This approach made it possible to build a computational evaluation process with its constant refinement as new results of measurements of the characteristics of electromechanical devices become available. The results of a model experiment with a software implementation of the proposed method, performed in the MatLab 2021a environment, are presented, which showed the consistency of the program modules and obtaining a risk assessment result that is consistent with the expected dynamics of its change. Read more...

Algorithm for identifying threats to information security in distributed multiservice networks of government bodies

The results of studies are presented, the purpose of which was to develop an algorithm for identifying information security threats in distributed multiservice networks that provide information interaction of regional government bodies, as well as their communication with the population of the region. The relevance of the research topic is due to a significant increase in various types of cyber attacks on the computer networks of public authorities and the need to increase the level of security of these networks by intellectualizing methods for combating information security threats. The algorithm is based on the use of machine learning methods to analyze incoming traffic in order to identify events that affect the state of information security of public authorities. The algorithm provides for input traffic preprocessing, as a result of which a set of images (signatures) obtained from Wasm binary files is formed, and then the image classifier is launched. It contains a sequential inclusion of deep neural networks – a convolutional neural network for signature classification and a recurrent network that processes the sequences obtained at the output of the convolutional network. Features of the formation of signatures in the proposed algorithm, as well as sequences at the input to the recurrent network, make it possible to obtain the resulting assessment of information security, taking into account the history of its current state. The output of the recurrent network is aggregated with the result of comparing the actual signatures with those available in the database. The aggregation is performed by the fuzzy inference system of the second type, using the implication according to the Mamdani algorithm, which generates the final assessment of information security threats. Software was developed that implements the proposed algorithm, experiments were carried out on a synthetic data set, which showed the efficiency of the algorithm, confirmed the feasibility of its further improvement. Read more...

Fuzzy model of a multi-stage chemical-energy-technological processing system fine ore raw materials

The results of the study, the purpose of which was to build a software model of a multi-stage integrated system for processing finely dispersed ore raw materials, are presented. The role of such raw materials can be processed waste at mining and processing plants of apatite-nepheline and other types of ores, which accumulate in large volumes in tailing dumps. They create a significant environmental threat in the territories adjacent to the plants due to weathering, dust formation, penetration into the soil and aquifers of chemical compounds and substances hazardous to human health. Therefore, the improvement of existing production processes, the development of new technological systems for mining and processing plants, including the application of the principles of the circular economy, waste recycling, justifies the relevance of the chosen research area. The proposed program model is based on the use of trainable trees of systems (blocks) of fuzzy inference of the first and second types. This approach made it possible to avoid unnecessary complication of the bases of fuzzy inference rules when using only one fuzzy block when building a multi-parameter model of the entire multi-stage complex system. The use of several fuzzy inference blocks that describe the behavior of individual units of the system and their configuration in accordance with the physical structure of the system allows the use of relatively simple sets of rules for individual blocks. The joint selection of their parameters when training a tree of fuzzy blocks makes it possible to achieve high accuracy of the solutions obtained. The novelty of the research results is the proposed software fuzzy model of an integrated system for processing finely dispersed ore raw materials. The results of a simulation experiment conducted in the MatLab environment using a synthetic data set generated in Simulink are presented. The results showed that the trained fuzzy model provides good fidelity of the parameters and variables from the test part of the synthetic set. Read more...

Neuroregulator of the complex technological system for processing ore waste

The study is devoted to improving the management system of a complex technological system for processing ore waste. Such waste accumulates in large volumes in the territories adjacent to the mining and processing plants, posing a great environmental threat to both the population and the environment due to dust formation and the penetration of harmful compounds into the soil and groundwater. Therefore, the task of improving the management systems for the processing of ore waste, as one of the priorities, is on the current agenda of the management of mining and processing plants. The complexity of the technological system is manifested in the presence of two processing lines that differ in the set of units, and the choice of line depends on the granulometric composition of ore waste. The scientific novelty of the research results is the proposed structure of the neural network controller based on the reference model for the technological system, which is used as deep recurrent neural networks. The general structure of the neuroregulator includes several local neurocontrollers for each of the units of the technological system. Recurrent neural networks make it possible to create high-precision digital copies of individual units of two processing lines and use them to simulate the response of control objects when setting up controllers. Approbation of the proposed structure of the neuroregulator was carried out in the MatLab-Simulik environment, neural networks were designed using the Deep Network Designer tool. The results of testing showed that the speed of the control system is increased compared to other architectures of neuroregulators available in the Simulik environment, which can positively affect the operation of the entire technological system in transient conditions, in particular, reduce technological losses. Read more...

A method for solving the inverse kinematics problem based on reinforcement learning for controlling robotic manipulators

A method for solving the inverse kinematics problem for a three-link robotic manipulator is proposed based on one of the types of machine learning - reinforcement learning. In the general case, this task consists of finding the laws of change in the generalized coordinates of the manipulator’s gripping device that provide the specified kinematic parameters. When solving the problem analytically, the basis for calculating inverse kinematics is the Denavit – Hartenberg parameters with further implementation of numerical matrix calculations. However, taking into account the kinematic redundancy of multi-link manipulators, this approach is labor-intensive and does not allow automated consideration of changes in the external environment in real time, as well as the features of the robot’s field of application. Therefore, an urgent research task is to develop a solution whose structure contains a self-learning block that provides a solution to the inverse kinematics problem under conditions of a changing external environment, the behavior of which is unknown in advance. The proposed method is based on simulating the process of achieving the goal of robot control (positioning the gripping device of the manipulator) at a given point in space using the trial and error method. For approaching the goal at each learning step, a reward function is calculated, which is used when controlling the robot. In the proposed method, the agent is a recurrent artificial neural network, and the environment, the state of which is observed and assessed, is a robotic manipulator. The use of a recurrent neural network made it possible to take into account the history of the movement of the manipulator and overcome the difficulties associated with the fact that different combinations of angles between links can lead to the same point in the workspace. Testing of the proposed method was carried out on a virtual model of the robot, made using the MatLAB Robotics System Toolbox and the Simscape environment, which showed high efficiency in terms of the “time – accuracy” criterion of the proposed method for solving the inverse kinematics problem. Read more...

A method for predicting bank customer churn based on an ensemble machine learning model

The results of research are presented, the purpose of which was to develop a method for predicting the outflow of clients of a commercial bank based on the use of machine learning models (including deep artificial neural networks) for processing client data, as well as the creation of software tools that implement this method. The object of the study is a commercial bank, and the subject of the study is its activities in the B2C segment, which includes commercial interaction between businesses and individuals. The relevance of the chosen area of research is determined by the increased activity of banks in the field of introducing digital services to reduce non-operating costs associated, in particular, with retaining clients, since the costs of attracting new ones are much higher than maintaining existing clients. The scientific novelty of the research results is the developed method for predicting the outflow of commercial bank clients, as well as the algorithm underlying the software that implements the proposed method. The proposed ensemble forecasting model is based on three classification algorithms: k-means, random forest and multilayer perceptron. To aggregate the outputs of individual models, it is proposed to use a learning tree of fuzzy inference systems of the Mamdani type. Training of the ensemble model is carried out in two stages: first, the listed three classifiers are trained, and then, based on the data obtained from their outputs, a tree of fuzzy inference systems is trained. The ensemble model in the proposed method implements a static version of the forecast, the results of which are used in a dynamic forecast performed in two versions – based on the recurrent least squares method and based on a convolutional neural network. Model experiments carried out on a synthetic dataset taken from the Kaggle website showed that the ensemble model has a higher quality of binary classification than each model individually. Read more...