CC BY
35
AN ANALYSIS OF THE OBJECT AND THE DEVELOPMENT STRUCTURES OF CONTROL SYSTEM NEURONET CONTROL SYSTEM OF THE VACUUM DEAERATOR VD-400
Master of 2 years of study Alekseev P. P. Candidate of Engineering Sciences Shcherbatov I. A.
Russia, Astrakhan, Department Automation and control, Institute of information technology and communication, Astrakhan state technical University
ABSTRACT_
In the article is considered algorithms which are used for program realization neural network system of managing oxygen concentration of deaerated water. The analyze of vacuum deaerator VD-400 is made as an object of control. Identified regulate and perturbing effects and also regulated options VD-400. Especial attention was for software program which creates multilayer neural networks models direct spread and teaches them online. There are shown interfaces of teaching and modeling. Also the general circuit control vacuum deaerator VD-400.
Introduction. The systems based on knowledge are one of the key directions of researches in the field of artificial intelligence. The complex problems arising in practice are solved with use of experts' knowledge [1]. This area of artificial intelligence includes the algorithms, models and methods directed to automatic accumulation and formation of knowledge, using procedures of the analysis and synthesis of data. Artificial neural networks imitate a natural prototype which plays the predominating role in the organization of higher nervous activity of the person and his mental abilities.
The vacuum VD-400 deaerator (LLC LUKOIL-Astrakhanenergo of division the Astrakhan HPS-2) intended for removal of corrosion and aggressive gases from make-up water of power coppers acts as object of research. According to GOST 16860-77 VD-400 has to provide average heating of water at a size from 15°C to 25 °C at change of productivity in the deaerator from 30% to 120% from nominal, the content of oxygen in deaerated water shouldn't exceed 30 mkg/kg, free carbonic acid has to be absent.
Nowadays in a control system of process of chemical water treatment of a boiler room of HPS-2 in Astrakhan local automatic equipment on the basis of analog technical means is used. Technological process is considered as a set of the isolated parameters, control of which is exercised of local systems of automatic control. Each measured parameter from primary measuring converter arrives on the analog control device installed in a case in operator. Indication and registration of parameters is carried out by the analog indicators and recording registrars established on a board in operator room.
Regulation of technological parameters can be also carried out in the manual mode from a board in operator room by means of buttons.
On the basis of results of the analysis of the existing analogs ACS (automated control system), a current state of automation of technological processes of regulation of deaeration of feed water, it is possible to draw the following conclusions:
- The instrument reading defining concentration of oxygen isn't used as adjustable parameter, and the applied control systems provide stabilization only of indirect parameters: entrance Temperature (feed water temperature in front of the deaerator) and exit Temperature (feed water temperature after the deaerator).
- Any of the analyzed systems doesn't use neural network control though there is a possibility of use of advanced technology, for the purpose of increase of operational characteristics of the operating technological installations.
ARTICLE INFO_
Received 27 October 2016 Accepted 03 November 2016 Published 15 November 2016
KEYWORDS
Vacuum deaerator, Multilayer neural network, Training and simulation of neural networks, Linear function
© 2016 The Authors.
http://ijite. org/
3(5) November 2016 31
- Creation of a highly effective neural network control system of process of deaeration of feed water, in particular regulation of temperature of the chemical desalinated water and pressure of feed water, is an actual task and has practical value.
Task definition. Subject of this research are algorithms of work of a neural network control system of concentration of oxygen of deaerated water of the vacuum VD-400 deaerator, and also their program realization.
Objective of this research is to increase effectiveness of the control of the vacuum deaerator, due to development of the software which main functions will be in regulation of temperature of the chemically desalinated water and pressure of feed water. The choice of optimum parameters will be made by means of creation of models of multilayered neural networks of direct distribution and their training in real time. As possible functions of activation of a network the linearfu) =u function and a hyperbolic tangent off(u)=(ell-eu)/(eu+eu).
Analysis of the control object. The block diagram (fig. 1) gives a general idea about the principle of action of a control system and about her main functional parts. The offered scheme of control of the vacuum deaerator consists of two levels: the 1st ACS level - a local network of controllers, the 2nd ACS level - a workstation.lst ACS level - the local network of controllers represents chemical water treatment process, including VD-400 which technological parameters, measuring devices via input modules, are transferred to the programmable logical S7-1200 controller. In its turn, the controller through output modules, by executive mechanisms operates the regulating and locking fittings. On the communication line data are transferred to the 2nd ACS level - a workstation (computer). According to the standard of physical level for the asynchronous interface, data are regulated by electric parameters of the differential communication line as «general tire» RS-485 in an enterprise control system.
The developed control system has two-level structure. The lower level includes the measuring converters and executive mechanisms connected to the programmable logical SIMATIC S7-1200 controller.
Fig. 1. Block diagram of the developed control system
In monitoring systems and regulations data transmission about the measured parameters from sensors on input modules, and also the operating influences from modules of a conclusion to executive mechanisms is carried out by means of the unified current signals.
Considering the vacuum deaerator as a control object, it is possible to allocate the following influences and parameters (fig. 2).
Fcfipw
T.
cihpjw
chpw
I amb
Ql
CO2 L
Fig. 2. Structure of the existing control system on AHPS-2 the vacuum deaerator
32 3(5) November 2016 http:/Ajite.org/
The regulating influences: Fs - a steam consumption on the vacuum deaerator; Rs - a steam pressure. Adjustable parameters: CO2 - concentration of deaerated water from the vacuum deaerator; L - level of deaerated water in a tank of collecting deaerated water; P - pressure in the vacuum deaerator.
The revolting influences: Controllable: Tchpw - temperature of chemically purified water; Ts -temperature of steam; Pv - vacuum pressure; Fevap - an expense of evaporate; Pchpw - pressure of chemically purified water. Uncontrollable: Tamb - ambient temperature; Qi- thermal losses.
Further in entrance and output sizes it is possible to choose structure of a neural network. It is known that when modeling any process of rather neural network with one hidden layer (fig. 3), with necessary amount of neurons equal 2*n+1. [5]
Fig. 3. Scheme of training ofneural network model
Before setting up the neural network model, it is important to perform a measurement filter. When performing objectives of a filtration and the forecast when processing results of measurements in real time it is possible to use a recurrent method of the smallest squares, and also a method of the singular stochastic analysis [4].
Conclusion. The vacuum VD-400 deaerator intended for removal of corrosion and aggressive gases from make-up water of power coppers has acted as an object of research. The analysis of the deaerator as an object of control is made. The regulating influences (an expense and a vapor pressure), adjustable parameters (concentration, pressure and level of deaerated water), and also the revolting influences are revealed (controllable and uncontrollable).
Further in entrance and output sizes the structure of a neural network is chosen. The technique of creation of neural network model in which the control algorithm consists of the repeating computing operations between the moments of receipt of the measured information on variables of a condition of object is given.
REFERENCES
1. Bobyr M.V., Titov V.S. Intellectual control system of the temperature deformations at the cutting // Avtomatizacija. Sovremennye tehnologii. 2011. N° 5.pp. 3-7.
2. Bobyr M.V. Diagnostics equipment CNC by methods of fuzzy logic // Promyshlennye ASU i kontrollery. 2010. № 1. pp. 18-20.
3. Nemchinov D.V., Protalinskiy O.M. Reducing the risk of emergency at work sites // Vestnik Astrakhanskogo gosudarstvennogo tekhnicheskogo universiteta. Seriya: Upravleniye, vychislitel'naya tekhnika i informatika. 2009. № 1. pp. 111-116.
4. Glazkov V.P., Bol'shakov A.A., Kulik A.A. Application neural network compensator for stabilized motion of the semiautomatic prosthetic systems // Mehatronika, avtomatizacija, upravlenie. 2014. № 1. pp. 13-17.
5. Protalinskiy O.M. Verification of information in primary pcs with fuzzy sets // Izvestiya vysshikh uchebnykh zavedeniy. Severo-Kavkazskiy region. Seriya: Tekhnicheskiye nauki. 2003. № 3. p. 60.
http:/ijite. org/
3(5) November 2016 33
