Often, when conducting
scientific research related to the peculiarities of the operation of a nuclear
reactor, one encounters the tasks of analyzing archived data. For example,
power plant workers can analyze existing archives in order to adjust the
further operation of the reactor. At the same time, during the data analysis,
it is possible to identify anomalies in the operation of the reactor and, by
conducting a study on the causes of the occurrence of such behavior, prevent
their further manifestations, thereby increasing the safety of operation of the
reactor itself. Also, when using additional support, operators of nuclear power
plants can increase the safety of facilities by quickly responding to changes
in parameters or exceeding any of the parameters of the maximum permissible
values.
The safe operation of
powerful nuclear power reactors is ensured by the availability of information
and computational systems that allow measuring, calculating, and monitoring the
most important parameters of a nuclear power unit.
At the stations, there
is a station-wide data exchange network between the process control systems.
Information from various systems enters the data warehouse at regular intervals.
Further, the workers of the power unit analyze the data obtained and correct
the operation of the reactor.
The files located in
the data warehouse are not convenient for analysis, therefore, the task arises
of forming such a data warehouse into which an archive containing the values of
the main parameters of the power unit functioning according to the selected
sections would be loaded. The archival information accumulated in this
repository can be further used for various studies.
Depending on the task,
the requirements for the nature, type and volume of stored information, as well
as the degree of its granularity, may differ significantly. For example, to
solve the problem of assessing the quality of the work of operating personnel,
it is necessary to store information on the type and number of monitored
parameters, the values of which have gone beyond the limits established by the
regulations,
the number
and type of operator impacts on the control object (movements of the control
and protection system organs (CPS), adjustment of the coolant flow rate, etc.),
the degree of spatial stability of the three-dimensional energy release field,
etc. for several hours. When solving the problem of identifying the causes of
failure, for example, fuel elements, fuel assemblies and cassettes, information
for a period from several days to several years may be required. In this case,
the prehistory of behavior of limiting parameters of the reactor and the
power
unit may be of interest. For
example, for a RBMK-type reactor (high-power channel-type reactor), such
parameters as the power of each channel, its power generation, coolant flow
through the channel, indications of the systems for monitoring the tightness of
the cladding of fuel elements (CFE) and monitoring the integrity of
technological channels (TC) [5], linear load on the fuel element, stock up to
the crisis of heat exchange, the number of permutations of the fuel cartridge,
etc. For a VVER-type reactor (water-water power reactor), the limiting
parameters partially coincide with the above, but also contain significant
additions and differences associated with the difference in the design of the
reactors. For example, the limiting parameter is the concentration of boric
acid, the absence of boiling of the coolant, the pressure in the reactor, etc.
However, despite the
fact that the tasks listed above are of different nature, the archive of
operational parameters of RBMK and VVER reactors should allow solving each of
them and, moreover, serve as an information base for solving newly arising
problems. With the current level of development of computing systems at NPPs
(nuclear power plants), it is possible to organize the storage of all
experimental and calculated information with high detail in time over a long
period of operation, however, problems arise with the express analysis of a
large amount of data.
The fundamental concept
in the development of software for the express analysis of an archive is the
method of scientific
visualization. Scientific visualization is a modern effective approach to data
analysis that allows visualizing arrays of data of different nature - abstract
or real. Visual information is better perceived and allows to convey the result
to the user quickly and efficiently. Physiologically
, the perception of visual information is fundamental for
humans. The success of visualization directly depends on the correctness of its
application, namely on the precise structuring of the approach and the data
itself. The essence of the scientific visualization method lies in the fact
that the initial analyzed data is associated with some of their static or
dynamic graphic interpretation, which is visually analyzed, and the results of
the analysis of this graphic interpretation are then interpreted in relation to
the original data.
The paper [9] describes
a modern storage of parameters of a nuclear power unit with an RBMK reactor and
a specially created module for visualizing archived data in a user-friendly
form for viewing the archive database from remote workplaces. This module also
serves to easily export data for subsequent analysis and calculations.
The module interface is
shown in Fig. 1. At startup, the main window of the program opens, the current
state of the database is analyzed and a list of available "time
slices" is formed. The title of the main window displays information about
the time of the last update. There is a button on the toolbar to manually
update the list of available slices. In the center of the program window, a
cartogram of the parameter that is selected in the list on the left is
displayed. Periodically (by default, once every 30 seconds) the program
automatically updates the list of available slices.
Figure 1: The window of the visualization module after
selecting the required information
The data of the SKALA-MICRO system are used to extract such
parameters as the reactor loading, the coolant flow rate in each channel,
information on the position of the control rods, the readings of the energy
release control sensors, as well as the calculated parameters - the reactor
power, the power generation of the cassettes and the margin before the crisis.
In many respects, a similar approach was used by us in the
development of software for creating an archive of operational parameters of a
power unit with a VVER-type reactor. In [10], a software package is
described that allows to load
archives of VVER reactor parameters into the database and view the already
loaded data in a convenient user interface.
The display of data on
the 3D core model has been optimized by simplifying the cell models. To display
the parameter values, two separate visual modules have been developed: a table
and a 3D model of the VVER reactor core.
On the left in all
modules there is a tree-like display of the state of the database (a complete
set of loaded data). The data in the tree view is structured as follows:
campaign-slice-parameter-level (if there are
several levels). Below the tree-like
display there is a window for displaying extended information on the cell to
which the mouse is pointing. Fig. 2 shows the main window of the application.
Figure 2: 2D renderer window after selecting the
required information
There
was developed a 3D data flow animation module (fig. 3) that allows viewing
changes
in parameter values
throughout the active zone in the form of animation with customizable
parameters, a difference module for displaying the difference in parameter
values for two different time slices, a module for displaying the history of
cell values and a module for calculating the correlation of a parameter value
[11].
Figure 3: 3D visualization of the reactor core
From the point of view of express analysis of the archive, software
[10] has been previously developed and presented which allows
to export, preprocess,
and visualize multidimensional data of the RBMK
archive. This software was developed using effective modern data visualization
tools. Since the archive is mostly multidimensional data, the problem of
reducing the dimension of the space of variables was solved to present 2D or 3D
visualization for further processing and making judgments about the safety of a
nuclear reactor.
The first method used to solve the problem is the method of
principal components. The multidimensional vector of parameters and maximum
allowable values were projected onto the first three principal components,
which, as a rule, reflected most of the variance. Further, the trajectory of
the motion of the parameters in the main components was deduced. At the same
time, the boundary of the maximum permissible values was visualized in the
space of variables. Going beyond the settings of at least one of the parameters
leads to an unplanned decrease in power, which can lead to an emergency.
In Fig. 4, you can see
when parameters are approaching or crossing limits and analyze time slices to
identify the cause. The color of the point for the time slice
indicates the deviation of the
values from the mathematical expectation in accordance with the scale. In the
role of the missing lower (upper) limits, the minimum values over the entire
(specified) time interval were chosen.
Figure 4: An example of using an application for
visualizing the trajectory of movement
The second, no
less popular method for
visualizing multidimensional data, was the Chernoff Faces method. This approach
allows to visualize multidimensional data using a human face for these purposes
[15]. Individual parts of the face, such as eyes, eyebrows, nose, lips, depict
the meanings of various signs, changing their size, shape or position. Usually,
the Chernoff faces method is
used
when it is necessary to group (cluster) objects according to several
characteristics, or when it is necessary to analyze presumably complex
relationships between variables.
The result of
visualization by the Chernoff faces method is observed in Fig. 5. Analyzing the
results obtained, it is possible to draw conclusions when some of the
parameters deviate greatly from their average values, when maximums or minimums
are reached, or to find implicit relationships between the parameters (when
visualizing more parameters). Such an approach, with the simultaneous use of a
larger number of parameters, would make it possible to visually highlight
clusters of similar facial expressions or find hidden dependencies of parameters
among themselves.
Figure 5: An example of using the application for
visualization by the Chernoff faces method
Performing a visual
analysis of the visualization result, you can manually select clusters of
similar values, or outliers. For example, in Fig. 5, you can note that on days
17-25 the eyes on the face are greatly expanded, however, starting from day 26,
the eyes are narrow, which suggests that it is necessary to pay attention to
this transition and analyze the archive in more detail on days 25-26. You can
also pay attention to day 16 - you can notice that the nose on the face is
practically absent, which indicates that there was some deviation on this day
and a deeper analysis is needed to identify the causes. Similar outliers can be
observed on days 15 and 19 - the eyebrows on the face are heavily frowned,
which also indicates any deviations from the normal operation of the reactor.
Thus, this method
allows you to visually identify deviations on the face. This is due to the
peculiarity of human perception of the face and the ability to instantly detect
even the slightest changes in it.
Unlike the archives of
the SKALA-MICRO system, the archives of the VVER reactor parameters are
transmitted as a set of files with the .txt extension, distributed over
directories.
Figure 6: The structure of the VVER reactor archive
The name of the
campaign is written in the name of the archive file, the root of the archive
contains intermediate directories within one campaign, each of which contains
directories whose names indicate a time slice. Each time slice contains a set
of files with a .txt extension for each parameter. The archive structure is
shown in Fig. 6.
Each file contains
parameter values for that slice for each cassette. Bindings to
cassettes are divided into three types: binding to the cassette number, binding
to the KNI number, and binding to the control system number. Along with the
archive of parameters, a text configuration file is supplied that defines the
relationship between cassette numbers, KNI numbers and CPS numbers. Several
lines from this file are presented in Table 1.
Table 1.
Fragment
from the configuration file of accordance between KNI and SUZ to cassettes
|
Relation
type
|
Value
type
|
Cassette
number
|
Related
element
number
|
|
Kni_Pos
|
Int
|
1
|
158
|
|
Kni_Pos
|
Int
|
2
|
163
|
|
Kni_Pos
|
Int
|
3
|
133
|
|
Kni_Pos
|
Int
|
4
|
93
|
|
Kni_Pos
|
Int
|
5
|
97
|
|
…
|
|
|
|
|
Suz_Pos
|
Int
|
1
|
17
|
|
Suz_Pos
|
Int
|
2
|
74
|
|
Suz_Pos
|
Int
|
3
|
151
|
|
Suz_Pos
|
Int
|
4
|
13
|
|
Suz_Pos
|
Int
|
4
|
90
|
Each file in the
archive begins with a header containing the campaign name, cut time,
measurement units and parameter name. Further along the lines, the values
of the parameter for this slice are recorded with reference to
the cassette number, KNI number or CPS number and possible reference to the
level for parameters whose values can be multilevel.
Determination of the type of binding is carried out analytically and depends on
the number of records in the file. The configuration file compares the number
of records in the file and the number of cassettes, KNI and CPS. The matched
element number is a parameter bindable number.
Also, a file with the
.pdf extension is attached to the archive, containing descriptions of each
parameter that may be present in the archive.
To start working with
the database, you need to execute a script to initialize the tables that store
information about the unloaded archive. You also need to create a configuration
file first.
The configuration files
provided with the archive are recorded in the “config”, “cassettes” and “parameters”
tables. After successful configuration, the configured field in the “config”
table is set to true.
For each parameter
presented in the archive, a separate table is created, called “p1”, “p2”, etc.,
where the digit is the parameter id. Some parameters may not be present in the
archive, therefore tables are created only for the present parameters. After
creating a table of values for a certain parameter, the presence
field in the “parameters” table for this parameter is set to true.
When reading the
archive, it is sequentially written to the “campaigns” tables, “timepoints” and
the table for a specific parameter.
The campaign parameters
table is filled in as the archive is read. This table identifies the presence
of parameters in a specific campaign. The table is primarily used to speed up
the retrieval of data for display in the user interface.
Fig. 7 shows the
structural model of the database of the archived values of the
VVER reactor parameters.
Figure 7: The structure of the database
Scripts were developed
for reading parameters and further processing in order to build visual graphs.
Also, a universal script was implemented for building the same type of charts.
The connection to the
database from the program occurs through SQL queries, namely the pyodbc
library. The third party pyodbc module makes it easy to access databases
through the Open Database Connectivity (ODBC) programming interface.
Further, to retrieve
the parameter from the database, you need to execute the SQL query using the
above class using the pandas library. Pandas is a high-level Python library for
data analysis. In the Python ecosystem, pandas is the most advanced and fastest
growing library for data processing and analysis.
Thus, it is possible to
extract parameters into convenient structures provided by the same pandas
library - DataFrame. The structure is an indexed multidimensional array of
values (Fig. 8).
Figure 8: A two-dimensional indexed DataFrame
To build graphs, a
method was developed using the matplotlib and seaborn libraries. Matplotlib is
a 2D (and 3D) graphics library for the python programming language that can be
used to create high-quality drawings in various formats. Matplotlib is a python
package module. Seaborn is a high-level API based on the matplotlib library.
Seaborn contains more adequate default charting settings.
Based on the
existing developments for carrying out express analysis in the
field of visualization of the RBMK archive, software for the VVER archive was
developed. The developed
software
makes it possible to analyze
the archive by constructing high-altitude or time graphs and
Chernoff faces, allowing to select the parameters and settings of the desired
visualization. Fig. 9 shows the main window of the application. This complex
allows connecting to a database storing pre-downloaded data and customizing the
configuration of the desired
visualization.
Figure 9: Main window of the implemented software
To build any type of graphs offered by the
complex, it is first needed to
select
the parameter to be
visualized. At the moment, the complex supports the following parameters:
•
Power generation by the DPZ;
•
Burnout;
•
Leaked charge;
•
Load factor of the central
TVEL.
At the same time, the program was developed considering the
possibility of adding parameters for visualization.
Also, to reduce the number of displayed parameters and
further visualization, the core was divided into 60-degree sectors of symmetry
(Fig. 10).
For the same reasons,
the active zone was divided into orbits separately in a similar way (Fig. 11).
Further, for each sector (orbit), the parameter value is averaged for all TCs
contained in the sector (orbit), and the average value is visualized in the form
of graphs to study the behavior of generalized parameters over time.
Figure 10: Division of the core into 60-degree sectors of symmetry
Figure 11: Division of the core into orbits
The
developed software package can be conceptually divided into 4 parts.
Each part is a separate type of visualization that can be built by
preconfiguring the required configurations. Also, the developed software allows
to work with
graphics: change the
scale, adjust colors and line types, save the graph as a graphic image.
At the same time, each of the
visualizations allows to preprocess data
for a more general visualization. The following options are available for
preparing data for further visualization
using one of the proposed methods:
•
Cassette (output of the
parameter value for the cassette);
•
Sector average
(sector-averaged parameter value);
•
Orbital average
(orbital-averaged parameter value);
•
Height distribution (only
for Chernoff faces);
•
Offset of the average by
sector (only for Chernoff faces);
•
Orbital average offset (only
for Chernoff faces).
Applying these methods, it is possible to draw conclusions
when some of the parameters deviate greatly from their average values, when maximums
or minimums are reached, or to find implicit relationships between the
parameters (when visualizing more parameters).
1.
Diagram of the parameter
distribution over the height of the reactor core (Fig. 12). To build a graph,
you need to configure the following parameters:
•
Cassette numbers;
•
Date in the format
"YYYY-MM-DD".
Figure 12: Diagram of the parameter distribution over the
height of the reactor core
2.
The timeline graph of the
parameter (Fig. 13). To build a graph, you need to configure the following
parameters:
•
Cassette numbers;
•
Level;
•
Start date of the timeline
in the "YYYY-MM-DD" format;
•
Date of the end of the
timeline in the format "YYYY-MM-DD".
Figure 13: Parameter timeline
3.
The timeline graph of
parameter offset (fig. 14). To build a graph, you need to configure the
following parameters:
•
Cassette numbers;
•
Start date of the timeline
in the "YYYY-MM-DD" format;
•
Date of the end of the
timeline in the format "YYYY-MM-DD".
Figure 14: Parameter offset timeline
4.
Visualization by Chernoff
Faces method (Fig. 15). To build a graph, you need to configure the following
parameters:
•
Start date of the timeline
in the "YYYY-MM-DD" format;
•
Date of the end of the
timeline in the format "YYYY-MM-DD".
Figure 15: Static rendering with Chernoff faces
It should be noted that the type of visualization with the
use of Chernoff faces also supports dynamic changes in the characteristics of
the face, which allows visually, "in real time", to trace the
dynamics of changes in the face in comparison with the "average face"
– face, that characteristics take average values for the selected time slice.
An example of such visualization is shown in Fig. 16.
Figure 16: Dynamic rendering with Chernoff faces
This article presents a software package for visualizing
the archive of a VVER-type nuclear reactor. This complex has an advantage over
the previously developed ones in terms of performing express analysis and
building dynamic visualization. With the help of this complex and the developed
methods, it is possible to construct visualizations with the desired settings
and, thus, to reveal anomalies or patterns in the
behavior of parameters over time. Also,
adapted visualization based on the Chernoff faces method allows to cluster
generalized parameters, or to identify strong deviations, thereby speeding up
the process of the analysis itself. After conducting a generalized study, it is
possible to move on
to a more
detailed, in-depth study of the movement of individual parameters to identify
the reasons for this behavior.
The novelty of this software is using of new methods
applied to the archives of operational parameters of nuclear reactors. In this
area, the use of new methods can reveal new patterns, as well as improve speed
and quality of the analysis of the reactor behavior in different time slices
(for example, during the day or throughout the year). Also, a structured
approach based on the scientific visualization method allows the implementation
of software that can be used both for post-analysis of the archive and for use
by operational personnel. In this case, the software is developed in such a way
that, when using various archives and configuration settings, a result can be
obtained that can be interpreted for the initial object of consideration - the
reactor.
Also, the implementation of new software systems for
visualizing the archives of a nuclear reactor makes it possible to approach the
solution of the analyzed problem from the point of view of an express analysis
of the archive and thus conduct a study of multidimensional archive data using
appropriate methods.
The proposed software can be used both by the operating
personnel of the NPP as an auxiliary one in order to increase the efficiency of
monitoring the operation of the power unit, and in order to analyze the
existing archive database.
It is planned to continue working on improving the quality
and stability of the developed software package, as well as expanding the
functionality and adding new features.
[1]
G. Tikhomirov, I. Saldikov,
E. Malikova, L. Kuchenkova, V. Pilyugin. NRNU MEPhI experience in development
and application of visualization software in nuclear power plants education.
Scientific Visualization, 2012, volume 4, number 2, pp. 57-63.
[2]
A.O. Bukalin, A.M. Zagrebaev,
V.N. Samanchuk. Visualization of the process of neutron-physical calculation of
a nuclear reactor. Scientific Visualization, 2020, volume 12, number 1, pp.
112-119, doi: 10.26583/sv.12.1.10
[3]
I. Ivanov, S. Bychkkov, V.
Druzhinin, I. Moiseyev, U. Shmonin, N. Shchukin. Application of visualization
tools for analysis of statistical uncertainties in monte carlo calculations of
channel-wise functionals in rbmk-1000. Scientific Visualization, 2012, volume
4, number 1, pp. 22-30.
[4]
D. Zinakov. ReQt – software
visualization and analysis of calculation results of RBMK-1000 reactor cores.
Scientific Visualization, 2012, volume 4, number 1, pp. 31-41.
[5]
Dollezhal' N.A., Emel'yanov
I.Ya. Kanal'nyi yadernyi ehnergeticheskii reaktor. [Channel Nuclear Power
Reactor] Atomizdat Publ., 1980.
[6]
M.A. Abramov, V.I. Avdeev,
E.O. Adamov et al.Edited by Yu.M. Cherkashova. Kanal'nyi yadernyi
ehnergeticheskii reaktor RBMK.[Channel Nuclear Power Reactor RBMK] GUP NIKIEhT
Publ., 2006. 632 p.
[7]
Ovchinnikov F.Ya., Golubev
L.I., Dobrynin V.D., Klochkov V.I., Semenov V.V., Tsybenko V.M.
Ehkspluatatsionnye rezhimy vodo-vodyanykh ehnergeticheskikh yadernykh
reaktorov. [Operational modes of water-cooled nuclear power reactors] Atomizdat
Publ., 1975.
[8]
Nuclear Power Engineering.
Problems. Solution. Edited by M.N. Strikhanov. Part 1. M.: Social Forecasts and
Marketing Center, 2011. 424 p.
[9]
A.M. Zagrebayev, R.N.
Ramazanov. Nuclear Reactor RBMK Archive Data Visualization. Scientific
Visualization, 2015, volume 7, number 2, pp. 1-11
[10]
Compression and
Visualization of the Operational Parameters Archive. Andrey Zagrebayev, Victor
Pilyugin, Stanislav Ten. GraphiCon 2020. Computer Graphics and Machine Vision. pp.
1-11. doi: 10.51130/graphicon-2020-2-3-11
[11]
A.M. Zagrebayev, I.Yu.
Leveev, V.V. Pilyugin, S. Ten. Creation and visualization of archives of
operational RBMKs parameters and VVER reactors. Scientific Visualization, 2020.
volume 12, number 4, pp. 33-45. doi: 10.26583/sv.12.4.04
[12]
Yurova L.N., Naumov V.I.,
Savander V.I., Zagrebayev A.M. Kompaktnoe predstavlenie vynutrireaktornoi
informatsii o potoke neitronov / Fizika yadernykh reaktorov. [Compact
representation of in-reactor neutron flux information. / Physics of nuclear
reactors.] Atomizdat Publ., 1975, number 4, pp. 19-23
[13]
Alimov A.L., Shchadilov A.E.
Optimal'noe adaptivnoe szhatie tsifrovykh soobshchenii po algoritmu
kusochno-lineinoi approksimatsii. [Optimal adaptive compression of digital
messages using piecewise linear approximation algorithm] Avtometriya Publ.,
1983, number 3, pp. 14-18.
[14]
A.M. Zagrebayev, V.A.
Nasonova, N.V. Ovsyannikova. Matematicheskoe modelirovanie yadernogo reaktora
pri sluchainykh vozmushcheniyakh tekhnologicheskikh parametrov. [Mathematical
modeling of a nuclear reactor with random perturbations of technological
parameters] NRNU MEPhI Publ., 2009. 400 p.
[15]
Chernoff, H. The Use of
Faces to Represent Points in K-Dimensional Space Graphically. Journal of the
American Statistical Association 68 (342). 1973. pp. 361–368. doi: 10.1080/01621459.1973.10482434
