Óêðà¿íñüêà   Ðóññêèé
DonNTU   Masters' portal

Abstract

Contents

Introduction

In the current conditions of modern scientific and technical progress, any enterprise in any industry cannot do without the help of scientific management methods, automation and computer technology for designing automatic and automated systems that allow to effectively manage various objects and processes. The set of existing control objects in real situations exhibit various properties and are poorly described or not at all described entirely in terms of classical differential systems. Such systems, in which it is completely possible to neglect their stochasticity, exhibit combined continuous-discrete behavior, and neglecting the discrete component does not provide adequate models. For this class of systems in the mid-sixties, a new approach began to develop, called hybrid systems, which is the subject of this work.

Classical methods of analysis and synthesis of control systems are based on the assumption that the mathematical model of an object is well-known and describes its behavior exactly. However, for modern approaches to the formulation and solution of control problems, a more critical look at the accuracy of mathematical models available to the developer is characteristic. At the same time they talk about the uncertainty of the mathematical model of the object.

The metallurgical plant is a rather complex and many-sided enterprise. There is a strong misconception that metallurgical plants are engaged only in smelting metals, which is not true. Also, various steel mills can have a rather different structure of the enterprise, the production cycle, the method of steelmaking and much more. However, it is not always the processes occurring in such a large enterprise that can be described by a certain functional dependence, often we have to deal with stochastic models and fuzzy control, especially in the management of some organizational or technological processes. In addition to the above, in the face of uncertainty, the importance of studying possible risks, which also influence decision making, increases.

1. Relevance of the topic

The relevance of this work lies in the fact that the metallurgical plant is a rather important part in the life of any city, because it provides a significant number of jobs and large revenues to the budget. The positive dynamics of the financial performance of enterprises of this type can serve as a serious help for economic growth and increasing the welfare of the population. The effective work of large enterprises, such as the one on which the study is based, is achieved through a set of different approaches.

The development of scalable systems equipped with a functional capable of solving problems in the face of uncertainty is one of the promising areas for working in crisis conditions that have recently been increasing more and more often.

2. The purpose and objectives of the study, the planned results

The aim of the work is to develop such a scalable process control system for an enterprise in the considered industry, which was applicable regardless of the current state of the organization and could react to those situations that may arise in the course of its economic, production and other activities.

The main tasks of the developed system:

  1. Creating a scalable system with a basic graphical user interface that implements the basic functionality when connecting and programming individual modules.
  2. For the technological part of production: the ability to create a layer between the upper and middle levels of the process control system, SCADA systems and PLC, which could control the technological processes better using the implemented methods.
  3. For the managerial part of production: creating the possibility of introducing automation of business processes using plug-ins.
  4. The implementation of modules for solving problems related to fuzzy logic, game theory, automatic control theory and others.
  5. Creation of a functioning system based on the received modules on the basis of an industry enterprise.

Development Object : a process control system under uncertainty

As part of the master’s work, it is planned to obtain relevant scientific results in the following areas:

  1. Application of game theory and fuzzy logic to control technological processes.
  2. Evaluation of the practical and economic efficiency of the proposed approach in comparison with the one that currently exists.
  3. Evaluation of the practicality of using neural networks trained in managing automated process nodes in comparison with humans.

3. Review of research and development

The issue of improving management efficiency due to fuzzy logic is, at the moment, one of the developing areas of the theory of automatic control. At the same time, the possibility of expanding the functionality implemented for these purposes to other areas of the enterprise’s economic activity has been studied rather poorly.

3.1 Overview of international sources

The use of fuzzy control in industry is often mentioned in various papers [1]. This topic is most often raised in the context of setting PIDs [2 - 3] because it is one of the effective methods settings. At the same time, the problem of assessing the economic condition of an enterprise, as one of the main sources of information for managing its conditions, has been widely studied and presented in the form of various standards. In the bachelor’s study, when developing a management system model for the overall enterprise assessment, a standard was chosen for the FERMA risk assessment [4], in turn, for evaluating the company and comparing it with other metallurgical enterprises The methodology of the rating agency Moody`s was chosen [5].

With regards to directly hybrid intelligent systems, some of the most significant works are [6 - 8]. Also, an important study regarding the use of hybrid intelligent systems is [9]

3.2 Overview of national sources

The theoretical basis for the study of management in the face of uncertainty in this paper is the work of Blyagoz and Blumin [10-11], which give sufficiently detailed information on the application of various methods for control in conditions of uncertainty (used in this work of game theory and fuzzy control). Kolesnikov's work [12] played a key role in the development of the architectural design of a hybrid intellectual system and the formation of an admissible realizable functional.

4. Methods of building an adaptive intelligent hybrid control system

The central issue in the development of various adaptive intelligent systems is to design a knowledge base about the subject area. Knowledge based systems use information from databases and knowledge bases. Such systems are widely used to solve various control problems in which there is no a priori necessary description of the state and structure of the system. In this case, the method of constructing an adaptive intelligent hybrid control system, considered in [12], involves the following steps:

  1. analysis of the problem situation;
  2. the formation of the subject area;
  3. development of methods for the interaction of classical and fuzzy logic;
  4. structuring the subject area and building a model based on classical and fuzzy logic;
  5. forming a knowledge base with a rule base as a control component;
  6. construction and description of the model in the form of separate concepts;
  7. performing computational experiments;
  8. modeling of individual subsystems of the hybrid system;
  9. testing (analysis of the adequacy of the model) of the hybrid system;
  10. quality assessment of computational experiments;
  11. correction or refinement of the resulting model.

Development of a hybrid research method for an adaptive intelligent control system. It is known that the basis of hybridization consists of three laws: the law of mutual adaptation, the law of discrete series of structures and the law of transformations.

The law of mutual adaptation. The dynamics and synthesis of the development of any system (hybrid) is a process of mutual adaptation of the components of the system among themselves and the system with the external environment, i.e. other autonomous methods and methods — hybrids.

The law of discrete series of structures. Any hybrid can be implemented through one of its possible structures from a discrete series. The law states that there is some method of obtaining one structure that is included in a discrete series from another structure of this series. Also, in this series there should be target structures that will allow to determine the problem of quality of a complex system, i.e make hybridization focused.

The law of transformation. Transformation of one hybrid structure into another can occur only through knowledge common to both structures, i.e. through the state of the system. The transformation law describes the formation of new system states, which are displayed by the intersection of their characteristic curves, their interference with each other. In accordance with the law of transformation, a new structure cannot be generated as such and arises only on the basis of the previous structure. At the same time, the mutual adaptation of a part of the components achieved under the old structure, sufficient for building a new structure, is preserved.

The first level is represented by traditional formal-logical thinking. This level is based on classical methods and management processes, which are described using a programming language. At the second level, fuzzy modeling is used, which, together with the obtained results of the first level, explores various aspects of uncertainty. Two-level positioning allows us to consider the overall structure of the data processing system from different points of view, while the interaction of elements (components) of the structure is not only mechanical or electrical, but also informational, which is an important attribute of modern organizational and technical systems. In accordance with the presented data processing system, the adaptive hybrid control system is implemented on the basis of a combination of algorithms and methods of traditional formal logical thinking and fuzzy logic.

The sequence of actions for solving the control problem is shown in the figure below.

The sequence of actions for solving the control problem

Figure 1 — The sequence of actions for solving the control problem

The proposed algorithm for solving the control problem makes it possible to solve one problem by several autonomous methods, and the mathematical or software implementation of this algorithm can be performed by one of the methods at a specific iteration. In this case, hybridization is based on two methods and the sequence looks like: method 1 -> method 2.

The structure of the algorithm consists of several stages. Identify variables:

  1. the fulfillment of method 1 (classical logic) corresponds to the condition I = 1;
  2. not fulfilling method 1 meets condition I = 0;
  3. a positive score corresponds to the condition L = 1;
  4. A negative score corresponds to the condition L = 0.

The content of the main stages of hybrid modeling has the following structure.

  1. The problem statement is implemented, i.e. exact wording of conditions with a description of the input and output data used to solve it.
  2. Collection of information. The received signals from the external environment are input influences for the preprocessing block of initial data.
  3. Preprocessing the source data, i.e. creation of a classified system for all methods used in the hybrid system to prepare information for the decision block.
  4. Building a knowledge base.
  5. Making a decision according to which the choice of the method for solving the control problem is made. In relation to the task in question, the hybridizer needs to have sufficient knowledge about whether enough knowledge is obtained after the first method is completed to continue the process of solving the control problem, therefore if condition I = 1 is met, then go to step 7, otherwise — Go to step 6.
  6. The implementation of the classic hybrid system method.
  7. Implementation of the fuzzy method of the hybrid system.
  8. Evaluation of the effectiveness of the decision. If the condition L = 1 is met, then go to step 10, otherwise — Go to step 9.
  9. Adjust the results to continue.
  10. The completion of the algorithm, the conclusion of the solution.

When implementing one of the methods of hybridization in management, a hybrid is created in the process of solving a management problem. The process of creating a hybrid is the formation of a certain sequence of interrelated relations Ri. A prerequisite for the functioning of a hybrid is the prioritization between the autonomous methods used by the hybrid.

Forming technology stack

Under the technology stack, it is customary to call the list of technologies, programming languages, libraries, extensions, etc., which are used to solve a particular problem.

One of the most rapidly developing programming languages ​​is Python. This language has weak IDEs for developing interfaces, and those that have sufficient functionality for implementing a project of this level & mdash; work with Python through binding components. So, for working with the graphical interface in Visual Studio, the IronPython extension is used, which currently only works with Python version 2, which imposes a restriction on the development process. For the development of the interface, it is convenient to use the WPF technology, since it contains a Canvas element that can be effectively used to build mnemonic schemes of technological processes, unlike Windows Forms. An alternative to this IDE can only be Qt Creator, which works through the PyQt extension, but it requires further study of the Qt environment, so preference was given to Visual Studio.

Since the development environment was chosen by Visual Studio, you should consider the possibility of creating a multilingual project. In this case, it makes sense, since only C# has a sufficiently detailed guide for creating knowledge bases, allows you to work directly with the NuGet system, which you may need for an effective development process, etc., you can use a bunch of these languages. Given the fact that Visual Studio is more suited to working with the C # language, the highest priority should be attributed to C#; at the same time, Python will act as an auxiliary for machine learning.

The choice of a priority development language C# allows you to uniquely define the fact that the operating system in which the developed GIS will work will be Windows, moreover, in the x86 edition, which will make it possible to achieve compatibility with older computers that are most often found in enterprises.

TensorFlow was chosen as a library for working with neural networks and machine learning, since it is the undisputed leader in popularity among cross-platform libraries, which also has sufficiently detailed documentation and has community support. When directly designing the system, you should also consider the possibility of completely eliminating the PS Python and using the less popular and less documented unofficial branch of TensorFlow — TensorFlowSharp.

To connect to the Allen-Bradley controller, preference was given to the LibPlcTag library as the most common; Both the S7.NET library and the small open source libraries found on GitHub can be used to connect to a Siemens S7 PLC — 300;

So the technology stack will look like this:

  1. GIS development and operation operating system: Windows x86
  2. programming languages: C# and Python (optional, version 2);
  3. development environment: Visual Studio (preferably the current version);
  4. DBMS: Microsoft SQL Server (preferably the current version);
  5. plug-in libraries and extensions: TensorFlow (TensorFlowSharp if Python is abandoned), LibTagPlc, S7.NET, IronPython.
  6. Other technologies used: WPF.

Conclusions

At the moment, the following results were obtained:

  1. a management model has been developed in the face of uncertainty (the result of undergraduate work);
  2. a list of implemented system functionality has been generated;
  3. describes the generally accepted method of constructing hybrid intelligent systems in the context of the system being developed;
  4. a stack of technologies has been formed that will be used to create the system, as well as basic system requirements for technical support.

When writing this essay, the master's work is not yet completed. Final Completion: June 2019. The full text of the work and materials on the topic can be obtained from the author or his manager after the specified date.

References

  1. Timothy T. Ross, Fuzzy Logic. — United Kingdom, 2010. — 530 p.
  2. Natsheh, E. and Buragga, K. A. Comparison between conventional and fuzzy logic pID controllers for controlling DC motors, IJCSI — International Journal of Computer Science Saudi Arabia, 2010. — 128–134 pp.
  3. Jelena Godjevac, Comparison between PID and fuzzy control, LAMI–EpFL — Ecole polytechnique Federale de Lausanne, Switzerland [ ]. — : https://www.polytech.univ-smb.fr/fileadmin....
  4. FERMA a risk management standard. – Federation of European risk management association, 2002. – 16 .
  5. Global Mining Industry. Rating methodology. – Moodys Investors Service, 2014. – 24 .
  6. Medsker L.R. Hybrid Intelligent Systems. – Boston: Kluwer Academic publishers, 1995. – 298 .
  7. Wermter S., Sun R. Hybrid Neural Systems. – Heidelberg, Germany: Springer-Verlag, 2000. — 401 p.
  8. Jain L.C., Martin N.M. Fusion of Neural Networks, Fuzzy Systems and Genetic Algorithms: Industrial Applications. – CRC Press, CRC Press LLC, 1998. — 368 p.
  9. Pacheco, R.A hybrid intelligent system applied to financial statement analysis / R. Pacheco // Proc. of 5th FUZZ IEEE. New Orleans. – 1996. –  1007 — 1012 pp.
  10.  .. / .. , .. . – : , 2001. – 138 .
  11.  .. / .. , .. // , 2006. – 4. – . 28—34.
  12. , .. : / ..  – . : - , 2001. – 600 .