Resume
Збережено в:
| Дата: | 2010 |
|---|---|
| Формат: | Стаття |
| Мова: | English |
| Опубліковано: |
Інститут проблем штучного інтелекту МОН України та НАН України
2010
|
| Назва видання: | Штучний інтелект |
| Онлайн доступ: | http://dspace.nbuv.gov.ua/handle/123456789/56165 |
| Теги: |
Додати тег
Немає тегів, Будьте першим, хто поставить тег для цього запису!
|
| Назва журналу: | Digital Library of Periodicals of National Academy of Sciences of Ukraine |
| Цитувати: | Resume // Штучний інтелект. — 2010. — № 2. — С. 170-179. — англ. |
Репозитарії
Digital Library of Periodicals of National Academy of Sciences of Ukraine| id |
irk-123456789-56165 |
|---|---|
| record_format |
dspace |
| spelling |
irk-123456789-561652014-02-13T03:13:12Z Resume 2010 Article Resume // Штучний інтелект. — 2010. — № 2. — С. 170-179. — англ. 1561-5359 http://dspace.nbuv.gov.ua/handle/123456789/56165 en Штучний інтелект Інститут проблем штучного інтелекту МОН України та НАН України |
| institution |
Digital Library of Periodicals of National Academy of Sciences of Ukraine |
| collection |
DSpace DC |
| language |
English |
| format |
Article |
| title |
Resume |
| spellingShingle |
Resume Штучний інтелект |
| title_short |
Resume |
| title_full |
Resume |
| title_fullStr |
Resume |
| title_full_unstemmed |
Resume |
| title_sort |
resume |
| publisher |
Інститут проблем штучного інтелекту МОН України та НАН України |
| publishDate |
2010 |
| url |
http://dspace.nbuv.gov.ua/handle/123456789/56165 |
| citation_txt |
Resume // Штучний інтелект. — 2010. — № 2. — С. 170-179. — англ. |
| series |
Штучний інтелект |
| first_indexed |
2025-07-05T07:24:05Z |
| last_indexed |
2025-07-05T07:24:05Z |
| _version_ |
1836790836867629056 |
| fulltext |
«Искусственный интеллект» 2’2010 170
3-R
N.A. Bazhenov
Combining Probabilistic Tagging with Rule-based Multilevel Parsing
for Requirements Elicitation
The concerned paper deals with the problem of natural language processing and ana-
lyzes with regard to requirements elicitation and analyze, which are realized by requirements
and software engineers in cooperation with stakeholders and end-users. To bridge gap bet-
ween raw requirement texts generated by system’s end-users and the conceptual models
which are needed for system designers and architects the hybrid NLP approach is proposed.
The developed approach is based on POS probabilistic tagging combined with rule-based
chunking (shallow parsing) enriched with some additional heuristic procedures of different
linguistic characteristics identification. There are the following processing steps established
the developed methodology: (1) Tokenization and correction of wrong split tokens; (2) POS
tagging provided by a probabilistic trigram-based tagger with possibility of ad-hoc additio-
nal lexicons plug-in; (3) The mapping engine for splitting up and reinterpreting the standard
tagset into ontology-oriented one; (4) The pre-chunking procedure including compound noun
identification, extraction and generation of inflectional word forms, verb transitivity disam-
biguation, passive voice identification, nouns classification and other morphological infor-
mation extraction; (5) multilevel chunking procedure including the standard chunking rules
engine, multi-words and idiomatic expressions identification and complex phrase sequence
identification (such as prepositional post-modifier, parenthesis, homogeneous parts, relative
clause, infinitive group, subordinate sentence). This approach focuses on the enriching stan-
dard tags optimizing the automatic elicitation of concepts, relations and attributes for requi-
rements engineering purposes. Free text processing in this way can only be successful, if cer-
tain NLP standard techniques like probabilistic tagging are combined with special procedures
like pre-chunking and chunking.
Based on theoretical foundations of proposed approach the software application for re-
quirements documents processing was developed. The implemented tool was successfully
adopted and integrated into the project NIBA (Natürlichsprachliche Information Bedarf Ana-
lyse, founded at Alpen-Adria Universität Klagenfurt, Austria), whose aim is natural language
processing to support requirements engineering and conceptual modeling.
Al-Ammouri Ali, O.P. Shkurko
Techno-economic Self-control of Informative-control
Systems Functioning Efficiency
Nowadays the economy is closely connected with information technology and automa-
tion of production processes. Any enterprise achieves great economic success in case the
production is not expensive and qualitative.
To attain two such important attributes of the production as low cost price and high
quality it is possible only with method of manufacturing processes quality automation. Elect-
ronics is the basis of automation, and on its modern development level is microprocessor
technology.
Interface is the most pressing issue of the day while applying microprocessor techno-
logy in the technical processes of production including, foremost, efficient hardware input
data base to microprocessor for the very limited time.
Resume
«Штучний інтелект» 2’2010 171
3-R
This article deals with matrix structures of computing means connections, intended
for sequential information processing in real time, arriving randomly at a Poisson law on
several fronts (lines) and processed consistently by phases (matrix columns).
The selection method of optimal matrix structure of computing facilities is offered when
data is transferred through random flow by means of every channel (matrix line). Such struc-
tures can be attributed to multichannel and multiphase queuing systems.
A.S. Voronoj, A.A. Yegoshina
Tools for Integration Domain Ontologies for Knowledge Bases
of Intelligent Tutoring Systems
Intelligent educational systems based on systematized knowledge are intensively de-
veloping currently. Ontology is an effective way of knowledge presenting and organizing.
But the standard ontologies can't be used in specific and dynamic fields of knowledge that
makes their unification an actual task. The article deals with the basic problem of a phased
creation of ontologies in the knowledge basis of intelligent educational systems. Also it con-
tains the analysis of existing ontology integration tools, such as mapping tools, ontology
alignment and unification.
It is shown that the source of complicated ontology is a multiagent search system for
ontological information in the subject areas. Searches in this system are formed on the the-
saurus basis of subject area and on the expert local knowledge basis objectives. Relevant
components of the ontology from external repositories of educational resources and other se-
mantic Web search system, that are loaded in the working database of integrated association
of ontologies are found.
The authors created the algorithm of ontology associations, including: ontology order
alignment, resolution of names conflict, classes, and their properties and relations associa-
tion. Based on the proposed algorithm software for integrating ontologies is created, that is
an important part of information resources integration tool, which can be used for creation
of the ontological knowledge bases.
Developed tools can be used to speed the creation of intelligent education environ-
ment for students’ training in the direction of Computer Science. Further task is the realizati-
on of other integration system components, formalizing integration process and development
of a universal algorithm on ontologies.
M.I. Gorbijchuk, S.T. Samulyak, I.V. Shchupak
Classification Functions Synthesis Based on Genetic Algorithms
The task of building the classifiers based on the partition functions synthesis can be
found in many technical applications, for example in image recognition systems, in a prob-
lem of division of the diagnosis signs space into classes during technical condition evalua-
tion process etc. In most cases it is assumed that the structure of the partition function is gi-
ven and there is only a need to find its parameters. However the structure of the function is
unknown beforehand and the problem of its optimal structure definition is the topic of this
study.
The authors used the genetic algorithm in order to solve this problem. The essence of
the algorithm is that the structure of the partition function is chosen to be linear. The chro-
mosome associated with that structure consists of binary genes. “Null” value is assigned
Resume
«Искусственный интеллект» 2’2010 172
3-R
when the parameter value in function structure is equal to zero, otherwise the value is co-
unted as “one”. The problem of the optimal partition surface synthesis is solved the follo-
wing way – one chromosome is found from an initial pool of relatives, so that the fitness
function acquires its minimal value.
The novelty in this study is the classification function synthesis method development
based on the genetic algorithms. The practical value of the work is that the authors created
software for solving the partition functions synthesis problem and the particular example is
presented showing its efficiency.
G.V. Dorokhina, V.Y. Trunov, E.V. Shilova
Russian Words Morphological Analysis without Dictionary Module
The article is devoted creation of software for Russian words morphological analysis
without a dictionary (RwdMA_IAI). Such system is needed when the word is to be proces-
sed by a module of morphological analysis is out of dictionary. Also there was a plan to use
this module in word’s paradigms adding automation task.
The database of RwdMA_IAI is formed by using the Russian words declarative mor-
phological analysis module (RDMA_IAI). It contains table of pseudo-inflection paradigms.
Each pseudo-inflection paradigm is an ordered set of pairs: information about grammatical
form, pseudo-inflection. Two pseudo-inflection paradigms are equal if their sets of pairs
are same.
The filling of RwdMA_IAI database includes following stages: successive paradigms
sampling, pseudo-inflection paradigm forming for each paradigm, pseudo-inflection para-
digm finding in RwdMA_IAI database, rang increasing for found one and adding of that
pseudo-inflection paradigm which is absent.
As a result RwdMA_IAI is filled with paradigms of verbs, nouns, adjective from
RDMA_IAI database. It is able to process the morphological analysis of unknown words
and to synthesize their potential paradigms. The database of RwdMA_IAI contains about
19 thousand pseudo-inflection paradigms that includes (table of 680 thousand rows, 32.5 thou-
sand different pseudo-inflections).
Module using while Forum “Methods of automatic analysis of text estimation: mor-
phological parser for Russian” (http://ru-eval.ru/) shows that 4.5 % of processed words is
unknown and is to be processed by RwdMA_IAI module. An average time of word morpho-
logical analysis by RwdMA_IAI is over 0.01 second.
Module using for synthesizing the potential paradigms of unknown words shows that
there are too many hypotheses of paradigms for one word to automate words’ paradigms
adding process. The correct paradigm could be selected by analyzing the set of potential pa-
radigms of large text’s unknown words.
Resume
«Штучний інтелект» 2’2010 173
3-R
Vladimir S. Mayorov, Sergey V. Mayorov, Sergei M. Komarov
Program System for Laser Cutting
Laser cutting is widely used in modern industry for manufacturing of different parts
from metal, plastic, wood etc. Depending on the type of material and laser machine itself,
different parameters of processing are required. A source drawing for cutting can also in-
clude different graphical objects designed by means of different software applications, such
as AutoCAD, Corel Draw, etc. Usually outlines are composed of arcs, lines, polylines, Be-
zier curves, ellipses or NURBS.
The goal of this project is to create an intellectual program system for laser cutting,
capable of selecting the best parameters for laser cutting of complex outlines. The system
presented for your consideration consists of two main modules: geometry processor and
converter (GPC) and Decision Support System (DSS). The GPC module imports vector dra-
wings such as .DXF, .AI or .EPS files and converts all outlines into contiguous polylines
consisting of lines and arcs. A potential User can save the converted drawing and manually
edit it, or immediately turn it into a CNC program. All technological parameters for the CNC
are supplied by the DSS module. The DSS module intends to help technicians to choose
the best possible conditions for laser processing of materials. This module includes several
important blocks and components: physical models and equations, data bases, practical
knowledge of experts, experimental results bank, some important criteria, algorithms for
analysis and decision. A set of parameters for the description of laser technological proces-
ses has been developed. Each process is characterized by the values of such parameters chan-
ging within some reasonable diapasons. The DSS module processes technological data and
offers the solution using the modern Case-Based Reasoning (CBR) technology (solution of
problems by the analysis of precedents). Method “k” of the nearest neighbors (k-NN) brings
the best results. The practical tests prove the high efficiency of our program system for la-
ser cutting.
A.S. Minenko
Mathematical Modeling of the Processes Crystallization of Metal
There is researched Stephan’s space task regarding to an ingredient and convective
movement in a liquid phase, described by Navier-Stokes equations. The method of study-
ing of this problem consisting of the decision expansion for series of a small parameter is
offered. In the no stationary case the decision of corresponding boundary-value tasks for
definition of the expansion members is formed as fixed points of operators.
Convection influence on the front of crystallization is explored. The method of the con-
jugation tasks decision arising while Stephan’s tasks research in space is also developed.
This paper extend to the rime-dependent case some result obtained by the authors for
the steady-state Stefan problem with convection. Referring to the ice-water system, water
is assumed to be incompressible and to obey the Stokes equation (1)( )x
v v p f u
t
ν∂
− ∆ +∇ =
∂
( v velocity, ν kinematic viscosity, p pressure, f buoyancy force, (1)u water temperature),
while temperature (1)u satisfies the heat conduction-convection equation with temperature-
dependent thermophysical properties. The temperature field (2)u in the solid phase is go-
verned by diffusion only. At the ice-water interface, (1)u = (2)u =0 and the Stefan condition
holds. The scheme is completed by initial and boundary conditions.
Resume
«Искусственный интеллект» 2’2010 174
3-R
The authors present an existence theorem for the case of two space dimensions. The ma-
in difficulty consists in the fact that, to interpret the Stokes equation in a weak sense, some
information is needed on the region where the temperature is positive, which is in turn in-
fluenced by the velocity field itself. The precise formulation of the problem requires a te-
chnical choice on function spaces. Existence of a solution is proved by introducing a tem-
perature dependent penalty term in the fluid flow equation in order to define both the appro-
ximating temperature uε and the approximation velocity ε in the whole domain. Compactness
arguments are used to get a convergent subsequence, whose limit is shown to solve the ori-
ginal problem. The question of uniqueness is left open.
A.S. Minenko
Approximation Analysis of Many Dimensional Mathematical Model
of Crystallization of Metal with Convection
Under of the studying of the heat physics processes accompanied by substance phase
transitions, described by the mathematical model in which the each phase temperature sa-
tisfies the equation of heat transition with its heat physics coefficients, on border of phases
division, both temperatures are constant and equal to the temperature of a phase transition
and on the set parts of border certain schedule is supported. The phases division surface (“a
free border”) is unknown and Stephan’s condition is set additionally for its definition. This
condition turns mathematical model into a nonlinear problem of the large difficulty. The
Navier-Stokes equations are used to describe speed fields in a liquid phase. The method of
a small parameter is offered to solve the task.
Up to now only one paper has been know which contained the proof of the existence
of the classical solution to the two-phase multi-dimensional Stefan problem. The paper un-
der review is the second one containing result of no less significance. The author considers
the nondegenerate two-phase Stefan problem in a rectangle, assuming that there exists a
solution of the corresponding stationary Stefan problem and that the boundary and initial
data of the problem are only slightly different from the corresponding characteristics of the
stationary solution. The existence of the smooth solution of the problem is proved under
not too restrictive assumptions concerning the input data of problem. Moreover, the author
indicates a simple condition guaranteeing the existence of the corresponding stationary so-
lution and its stability, the latter being understood in its usual sense. This paper may be strong-
ly recommended to all interested in parabolic free boundary problems in general and in their
classical treatment in particular.
L.P. Mironenko, I.V. Petrenko, I.A. Novikova
The Unique Approach of Obtaining of the Canonical Equations
of the Second Order Curves
There is a number of approaches to the description and classification of the second
order curves. This question is not new in analytical geometry. From our point of view there
is a possibility for more laconic and clear statement for the problem of getting of the 2-nd
order canonical equations and their classification.
In the offered theory this problem has been solved on the basis of unique definition for
all the second order lines. In the base of this definition the eccentricity notation and the ec-
centricity distinction for various types of curves were applied. Let’s define the second or-
Resume
«Штучний інтелект» 2’2010 175
3-R
der line as the set of points of a plane, for which the relation of distance r from the fixed
point F named focus of the given curve, to any point M of a curve to distance d to some
straight line D named the directress, is a constant e named eccentricity.
Let’ apply this definition to the 2-nd order general equation with two variables x and y
,0222 332313
2
2212
2
11 =−++++ ayaxayaxyaxa
where 02
22
2
12
2
11 ≠++ aaa and 331211 ,...,, aaa are the equation factors, which are real num-
bers. Let’s arrange axis y of Cartesian system perpendicularly to directress D . The focus
of the asked line must be on the axis x and differ from
Cartesian system origin. At such conditions the curve must be symmetric with respect
to the axis x and factors 12a and 23a must be zeroes. In fact we’ll get the equation
.02 3313
2
22
2
11 =−++ axayaxa
In this expression it is easy to write down the full square with respect to the variable х.
It means there are two different cases, when 033
2
22
2
11 =−+ ayaxa and 02 13
2
22 =+ xaya .
Let’s write down the both cases by using Cronecker’s symbol δ , which is defined as
≠
=
=
nm
nm
mn if ,0
if ,1
δ .
Finally we’ll get ,0)1(2)1( 33131,
2
22
2 =−−++− axayax λδλ λ
where λ is a new parameter and 0111 >−= λa . Since a is the big semiaxis and b is the
small semiaxis of an ellipse, where aab <−= λ1 at 1<λ , therefore we’ll obtain:
,1
)1( 2
2
2
2
=
−
+
a
y
a
x
λ
where ,111)1( 2
2
2
2
2
22 <−=→=−=→−=
a
bee
a
bab ellipseλλ
where ellipsee is the eccentricity of an ellipse.
If 1>λ , then it follows 1>e and we’ll get the expression for the eccentricity of a hy-
perbole:
.11)1( 2
2
222 >+=→−=
a
beaeb hyperbole
If 1=λ , then it follows 1=e . If assume 0/ 2213 >−= aap we’ll get the canonical equa-
tion of a parabola: pxy 22 = , where 2/p is the distance from the focus to the origin of Car-
tesian co-ordinates.
Now it is possible to write down all the second order lines equation by using the uni-
que expression, which depends from the eccentricity e :
.2))(1( 1,
2222 pxyaxe eδ=+−−
At 0=e we’ll get the circle equation: .222 ayx =+ ..
In the traditional definitions of the ellipse and the hyperbole take place only one fo-
cus and one directress, but the symmetry of the last equation with respect to the axis y
means that these curves have two focuses and two directresses. Further in the paper has been
considered the main geometrical properties of the second order curves.
Resume
«Искусственный интеллект» 2’2010 176
3-R
K.V. Murygin
Detection of Car License Plate on the Basis
of the Mixed Cascade of Classifiers
The automation of auto transport traffic control is now actual and important problem
in which frameworks systems of automatic recognition of license plates actively are deve-
loped. There are many the similar systems focused on national standards of license plates
which can essentially differ under many characteristics: the size, the form, color of symbols
and a background, quantity and structure (letters and figures) symbols, the alphabet of sym-
bols. Therefore the problem of recognition of license plates of vehicles accepted in Ukraine
is actual and important in practice.
The first development stage is the decision of a problem of a car license number detec-
tion which is greatly defines quality of algorithm complex of license number recognition
system.
For detection of car license number on digital images the wide variety of heuristic ap-
proaches and their combinations based on prospective differences in properties of images
of license plates, and the images which are not concerning them is used. Some of such app-
roaches is based on: maps of vertical borders; spectral characteristics; fuzzy maps of an
membership of points to license plates; calculation of number of the borders following on
certain distance from each other; morphological transformation for search of lines of the sym-
bols having certain width, etc. Unlike listed approaches, the used method is based on the
general principles of search of objects on images. As the initial data for getting of effective
algorithm of search the set of grayscale pictures of car numbers, and background images is
used. Thus any heuristic knowledge of properties of images of numbers it is not used.
Detection was based on a cascade of classifiers and multiscale image scanning techni-
que. The classifiers of two types are used:
1. MCV-classifiers – at the cascade initial stages;
2. The linear classifiers received by means of algorithm AdaBoost – at the cascade en-
ding stages.
At the expense of use at the initial stages of MCV-classifiers the number of properties
used for classification is reduced at classification quality preservation that in a combination
to possibility of treelike representation of the MCV-classifiers allows reducing search time
essentially.
During training the cascade of classifiers having following characteristics has been re-
ceived: positive detection rate – 0.941, an error of false detection – <10-8. Rate of processing
of images in the size of 640*480 pixels on computer Pentium Core2Duo with frequency of
the processor 2.33GHz was 12-15 frames per second. The considered algorithm allows using
successfully it as the first stage of processing of images in systems of recognition of car li-
cense numbers. The further researches in a direction of development of automatic car plate
recognition system can be bound up with normalizations of the car plate image, segmenta-
tion of symbols and recognition of the symbols presented on the images of license plate.
Resume
«Штучний інтелект» 2’2010 177
3-R
A.N. Serebrovskyi
About Technology of Knowledge Extraction from the Informative
Resources of Expert System Subject Domain
The approach to creation of knowledge extraction technology from text electronic re-
sources which have not preliminary semantic description is offered. Technology is inten-
ded for forming and update of knowledge base of expert system and includes the following
processes: constructions of ontology and thesaurus of subject domain; the automated seman-
tic markup of text documents; extractions of knowledge from marked texts.
Description of thesaurus construction technology and ontology of subject domain
Thesaurus and ontology of subject domain are built one time at creation of expert sys-
tem and then repeatedly used for its exploitation.
Technology includes the following stages: forming of text set characterizing a subject
domain; compiling of terms dictionary of subject domain; forming of list of notions of sub-
ject domain; classification of notions according to criteria: object, process, event, property, va-
lue; establishment of base semantic connections between notions. As a result the description
of number of base relations between concrete notions of subject domain is formed, and the
accordance between ontology notions and their lexical presentations in the test documents
is set.
Semantic markup of electronic texts
Semantic markup includes stages: breaking up of electronic documents on fragments
(sections, pages, paragraphs); assignment of set of tags to every fragment corresponding to me-
tadata described in a fragment.
Extraction of knowledge from marked texts
Extraction of knowledge includes stages: forming of query for targeted knowledge ex-
traction; search on the formed query in the library of marked texts; ordering of the found text
fragments; filtration of found fragments package; presentation of knowledge from every frag-
ment as the set of suggestions in the limited natural language; internal knowledge represen-
tation of formalized suggestions and their loading in a knowledge base.
S.A. Subbotin
The Methods of Synthesis and Models of Neuro-fuzzy Networks
for Diagnosis and Classification Problem Solving on Features
Currently, there are several types of neuro-fuzzy network synthesis methods and mo-
dels, which have different performance. Therefore the problem of determining the approp-
riateness of methods using for the in practical problem solving is actual. This problem is
decided in the paper as a problem of evaluation of temporal and spatial complexity of the
neuro-fuzzy network synthesis methods.
The complexity of the learning process for the Mamdani-Zadeh and Takagi-Sugeno-
Kang neuro-fuzzy networks depends on the complexity of the network structure (number
of connections and number of neurons), of the dimension of the training sample, of the com-
plexity of the problem, of the weight initial values in the network, as well as the required
accuracy and iteration complexity of the error backpropagation method.
The methods of synthesis of neuro-fuzzy networks with a non-iterative weight eva-
luation based on a preliminary analysis of the training sample and on the automatic selection
of fuzzy terms. These methods do not require a user actions for the number and parameter
Resume
«Искусственный интеллект» 2’2010 178
3-R
of fuzzy terms setting, automatically form a partition of feature space, allow to take into ac-
count the information about the feature and fuzzy term informativeness in the structure and
parameters of the network, allow to ensure a high level of interpretability of constructed
models.
For these methods the analytical estimations of temporal and spatial complexity are
presented in the table. To simplify the resulting estimates put the assumption of the model
dependencies and values of the parameters affecting the complexity. This would facilitate
the assessment of complexity and makes them comparable to the different methods and
models.
These complexity estimations show that the most expedient method is the method of
synthesis of three-layer neuro-fuzzy network with binary output, the most economical use
of memory has the three-layer network with discrete output synthesis method, the most ra-
pid and least demanding of memory is a model of three-layer network with a binary output.
The analytical evaluation of temporal and spatial complexity can be used to automate
the selection of the synthesis method and model of neuro-fuzzy networks for solving speci-
fic practical problems.
He.S. Tsibulnik, V.N. Pigus
Intellectual Control Unit of Access
However, if as a reading device to put the microcontroller having onboard АЦП and
with big enough speed (25 MIPS and more) it is possible to improve security of the moni-
toring system of access considerably.
It is reached as follows. The internal power supply system of key DS1990A has no
voltage stabilizer and the target field transistor of a key has final internal resistance. It means
that level “0” at the key answer depends on pressure of an external food, capacity of the con-
denser and internal resistance of the target transistor of a key. If to measure values of pres-
sure in the beginning of a phase of the answer of a key - presence pulse, in middle this phase
and in the end of a phase, to write down in memory of the monitoring system of access
these values together with a key code this data can serve as additional parameters of identi-
fication of a key.
For enough high-speed microcontroller it is possible to write the program which not
simply sets parameters of time intervals (time slots) at key data reading, but also changes
them at key data recording in access system. The data which correspond to the minimum
duration time slots thus register.
Then key data reading can be spent to three stages: at the first reading standard time
parameters are used, the additional data for this concrete key is taken, repeated data reading
with the minimum values of time is made. The key is completely identified, if repeated
reading is spent successfully and its data coincides with the first reading.
Resume
«Штучний інтелект» 2’2010 179
3-R
V.Е. Tsyganash
The Analysis and Resolution of the Preferred Initiative Task
during Electrotheat Installation Control
In article the basic problems of management are formulated and analysed by the three-
phase arc furnace which control system represents set of three automatic regulators and one
adjustable object - furnaces into which energy is entered by three electric arches.
The situation of carrying out of fusion is considered at the increased share of a brea-
kage in batch when regular systems of automatic control become ineffective, and the per-
son-operator because of the possibilities limitation is compelled to make а decision in the
conditions of the raised levels of uncertainty.
In such conditions traditional construction technologies using of control systems by
powerful electrothermal installations, for example, on the basis of parametrical optimization,
does not give desirable results. Therefore for decrease in level of uncertainty at decision-
making the new method – a method of motions divisions on slow and fast is chosen, allo-
wing substantially to reduce the level of fast movements, at approach of a working point of
the unit to an optimum trajectory. It is carried out by introduction of absolute, portable and
relative systems of readout for the variables characterizing the process. Thus results of the
power supply interaction with the energy consumer are represented in relative system of
readout. For better physical possibilities of coordination the person that making decision
(operator) with variables of process, as a defining variable power gets out. Such choice is
caused by range expansion of the analyzed signals representation, as towards low frequen-
cies, and aside higher frequencies and decrease in regularity of signals space because of
transition to a uniform defining variable, notably, to the actual capacity arriving in loading
from the power supply.
Simplicity and efficiency of the offered method are confirmed by results of introduc-
tion on 12-ton of the arc steel-smelting furnace of Joint-Stock Company NKMZ. Its introduc-
tion has allowed to reduce considerably quantity of breakages of electrodes and to reduce
the expense of the electric power on 20 – 30 thousand kilowatt-hours monthly.
|