=Paper=
{{Paper
|id=Vol-2258/paper28
|storemode=property
|title=Fuzzy Reliability Model of Systems for Decision Support in Technical Diagnostics
|pdfUrl=https://ceur-ws.org/Vol-2258/paper28.pdf
|volume=Vol-2258
|authors=Evgenii Gavriliuk,Sergei Mantserov
}}
==Fuzzy Reliability Model of Systems for Decision Support in Technical Diagnostics==
https://ceur-ws.org/Vol-2258/paper28.pdf
Fuzzy reliability model of systems for decision support in
technical diagnostics
E A Gavriliuk1 and S A Mantserov2
1
Engineering center, Gazprom transgaz Nizhny Novgorod, 11 Larina Street, Nizhny Novgorod,
Russia
2
Department of Engineering automation, Nizhny Novgorod State Technical University n.a.
R.E. Alekseev, 24 Minina Street, Nizhny Novgorod, Russia
Abstract. The paper is devoted to the decision support method in technical diagnostics
problems. The analysis of two approaches to the determination of the systems technical state:
probabilistic and deterministic are carried out. The classical reliability model is considered on
the basis of statistical-probabilistic approach, a method for its modification is proposed. A
structural-parametric reliability model of complex technical systems (in particular, industrial
equipment) is proposed for the purpose of systematizing heterogeneous data using the
mathematical apparatus of fuzzy sets. The fuzzy reliability model allows one to process
diagnostic information about an object in a unified manner and bring it to a unified form. The
problem of reducing the parameters dimension under multifactorial analysis and the problem of
objects ordering (ranking) is solved. The technical state concept in terms of fuzzy sets theory is
formalized in the paper. The physical meaning of technical state index is described. The
presented fuzzy reliability model makes it possible to apply methods of computer information
processing for use in problem-oriented control systems and decision-making, allowing to
improve equipment reliability and management efficiency. The complex application of the
classical reliability model and fuzzy reliability model can improve the decisions foundation
made in technical diagnostics problems.
1. Introduction and task statement
The increasing importance of complex and expensive technical systems, especially in hazardous
industrial enterprises, safety reliability and durability requirements make it very important to assess
the state of the system, its reliability. Technical diagnostics is a field of knowledge which
encompasses theory, methods and means of determining the technical state (TS) of objects and
systems [1]. Technical diagnostics studies methods of obtaining, processing, evaluating diagnostic
information, diagnostic models, and decision-making algorithms. The purpose of technical diagnostics
is to increase reliability and lifetime of technical systems.
As is known, the most important reliability indicator is the absence of failures of the technical
system during its operation. Due to early detection of defects and malfunctions, technical diagnostics
can prevent failures with severe consequences during maintenance. This increases the reliability and
efficiency of operation and creates the possibility of operating technical systems in accordance with
their TS. This approach also involves the application of the maintenance and repair concept –
Reliability-Centered Maintenance, or its particular case – Condition Based Maintenance [2, 3]. In
operation, the object is operated up to the limit state in accordance with the recommendations of the
222
technical diagnostics system. Such an approach can bring an economic effect to the enterprise,
equivalent to the cost of 30% of the total equipment park [4].
R. Barlow and F. Proshan [5, 6] are experts in the reliability theory, who were among the first to
begin research on systems consisting of "ageing" elements. This problem arose because of practical
need to make quantitative estimates of the reliability indicators of systems for which only limited
statistical information was available. Such systems include, for example, aircraft: there is little failure
information due to sufficiently high reliability, but it is known that over time various reliability
characteristics are monotonically deteriorating. These studies are based on estimates for the mean
values of the time between failures and the fact that the failure intensity function is monotonic.
Technical diagnostics solves a wide range of problems, many of which are adjacent to the tasks of
other scientific disciplines. The main task of technical diagnostics is to determine the state of the
technical system.
The state of the system is described by a set of parameters (characteristics) which define it. This set
can be different. The system state recognition is defined as referring the system state to one of the
possible classes (diagnoses). It is often necessary to choose one of two diagnoses (differential
diagnosis or dichotomy); for example, "perfect (flawless) state" and "imperfect (flaw) state" [2].
However, in other cases it may be necessary to characterize the imperfect state in more detail. In most
cases of technical diagnostics, diagnoses (classes) are established in advance, and in this case the
recognition problem is often called the classification problem [4].
The analysis of the state is carried out under operating conditions, in which obtaining information
is extremely difficult. Thus, the information is limited, inaccurate or fuzzy. The theoretical foundation
for solving the main task of technical diagnostics is the pattern recognition theory. Algorithms for
recognition in technical diagnostics are partly based on diagnostic models that establish a connection
between the states of a technical system and their mappings in the space of diagnostic signals. An
important part of the problem of recognition is the rules of decision-making (decision rules).
The solution of the diagnostic task (referring the object to the perfect or imperfect one) is always
associated with the risk of false alarm or defect missing. The false alarm is the case when a decision is
made about the presence of a defect, but in reality, the system is in good order. The defect missing is
making a decision about a perfect state, while the system contains a defect. To make an informed
decision, it is advisable to involve new methods and use an integrated approach. This will reduce the
error risk of decision-maker.
Another important area of technical diagnostics is the controllability theory. Controllability refers
to the ability of the object to provide a reliable estimate of its technical condition and early detection
of malfunctions and failures. A major task of the controllability theory is the study of tools and
methods of obtaining diagnostic information.
Thus, technical diagnostics is characterized by two interpenetrating and interrelated directions: the
recognition theory and the controllability theory. The recognition theory contains sections related to
the construction of recognition algorithms, decision rules and diagnostic models. The controllability
theory includes the development of tools and methods for obtaining diagnostic information, automated
monitoring and troubleshooting. Technical diagnostics is traditionally considered as a section of the
general reliability theory.
Technical diagnostics is sometimes called “without disassembling” diagnostics, i.e. diagnostics,
carried out without disassembly of the object and termination of its operation. The task of technical
diagnostics is to determine the degree of wear of the object by measurement of indirect parameters. As
was mentioned earlier, one of the important features of technical diagnostics is recognition in the
situation of limited information, when it is required to be guided by certain methods and rules to make
an informed decision. A set of sequential actions in the recognition process is called the recognition
algorithm. An essential part of this process is the process of choosing parameters which describe the
state of the system.
Since technical diagnostics involves processing of large amounts of information, decision-making
(recognition) is often carried out with the help of a computer.
The solution of technical diagnostics problems is always connected with reliability forecasting for
the nearest period of operation (until the next technical inspection). The object TS forecast consists of
223
evaluating its possible state at a certain point in time in the future, based on known information about
changes in the past and the results of determining the actual state at the present time.
To solve the prediction problem, the time of object operation is divided into two intervals: T1 - the
interval of monitoring the state of the object (past and present) and T2 - the interval at which
forecasting is performed (in the future). The larger the interval T1 is, the greater the amount of
information about the nature of the process of changing the object state of is and the more reliable the
forecast is. However, an increase in the observation interval leads to an increase in costs associated
with experimental studies and diagnostic results processing. The forecast reliability also depends on
the given time in T2.
The TS change is determined by the nature of the change in the object properties in connection
with the constantly occurring processes of internal degradation changes and physicochemical
transformations under the environmental influence and operation mode. In this case, a continuous or
discrete change in the characteristics characterizing these properties is observed, which leads to the
displacement of the state vector of the object in the region of serviceable (perfect) states S1 to its
boundary along a certain trajectory. The task of forecasting in the general case is to predict the type of
such trajectory. The nature of change in the properties, and, accordingly, the object state in time, can
be described by the dependence:
F (Z , t ) A(Z , t ) X ( Z , t ), (1)
where A(Z, t), X(Z, t) – deterministic and probabilistic (random) components of the process,
respectively.
If the influence degree of the second component in the expression is negligible X(Z,t) → 0, the
process of object state changing is described as deterministic, otherwise A(Z,t) → 0 it must be
considered as random.
Thus, the recognition of the object TS in technical diagnostics (in particular, technical forecasting)
is based on two principles - deterministic (analytical) and probabilistic.
2. Probabilistic method of systems technical state recognition
The problem of probabilistic prediction reduces to an estimate of the object reliability indices at given
instants of time in the region T2. In this case, the procedure for prediction and the reliability of the
results obtained are largely determined by the amount of initial information about the reliability of the
object and its elements.
With probabilistic prediction it is assumed that malfunctions and failures of elements and systems
are random events, then probability theory and mathematical statistics are the main apparatus for
investigating reliability. The reliability characteristics are chosen from the number of indicators
accepted in probability theory.
The main advantage of statistical recognition methods is the ability to simultaneously take into
account the signs of a different physical nature, since they are characterized by dimensionless
quantities – the probabilities of their appearance at different states of the system. If the state of the
object is changed randomly and, consequently, its diagnostic parameters are, the probabilistic forecast
can be calculated only from the results of observations of a group of identical objects operating under
the same or similar conditions. The calculation is based on the fact that the characteristics of the
position of the random process are non-random functions of time.
Currently, many manufacturers offer operating time to failure as the main technical specification
for the prediction of failures. However, this value has an idealized basis, since general statistics are
used that do not take into account the specific conditions of operation of the facility. This accuracy in
practice is simply not enough for a reliable prediction of equipment failures. As examples of such
conditions it is possible to allocate installation, climatic conditions, vibration, noise, electromagnetic
field guidance and other. Even from the place that this element will occupy in the common system, the
amount of time between failures will depend.
Thus, if for a single case (operation in production 1) an element is influenced by a set of conditions
A, then for another case (operation in production 2) the same element will be affected by the set of
conditions B.
224
To increase the reliability of probabilistic prediction, it is necessary for a particular operating
company to maintain a particular failure statistics throughout the life cycle of an element to calculate
the risk degree – probability of failure. The risk degree should be understood as the expected
frequency or likelihood of equipment failure. The application of risk degree concept, thus, allows to
translate the danger into a discharge of measured values. The risk degree is actually a measure of
danger [7].
Thus, we will receive the magnitude of the risk directly for a specific case, thereby maximally
bringing it closer to the true value. The risk degree is a constant value, but for different production
conditions it is different. It should be noted that the proposed approach is suitable for enterprises with
a geographical extent and a large number of units of unified equipment.
Any transition from one state to another can be gradual with a relatively long duration in time and a
sudden one with a relatively short duration in time (figure. 1).
Figure 1. Graphic explanation.
Figure 1 shows the dynamic change in the object TS I in time t. In the interval from 0 to t1, the TS
changed gradually but in the interval from t1 to t2 the TS changed suddenly. The prediction of the
moment t1 has the greatest interest and complexity at the same time. A statistical-probabilistic method
is used to determine this sudden moment. It should be noted that a sudden failure means that the
reasons for this refusal are not established, that the process of changing the TS of this object has not
been studied to the end. Either the set of diagnostic parameters is not determined in due measure, that
is the model describing the object TS is incorrectly defined, otherwise there would be no surprise
factor at all. In other words, the use of probability theory as a predicting method of objects TS (in
particular, failures) should be the last resort.
On the basis of the foregoing, the objective of technical diagnostics can be depicted as the
following function (objective function):
F (Z , t ) A(Z , t ), (2)
or:
A(Z , t ) 0, (3)
In other words, all occasional failures (and consequently their causes) must be transferred into the
category of deterministic ones. Then the failures nature will not be sudden.
Due to the idealized (theorized) nature of the classical reliability theory, the risk of an incorrect
decision to classify an object as "perfect" or "imperfect" only on the basis of statistical-probability
methods is high. Accordingly, it is necessary to consider the possibility of additional support for
decision-making. This problem can be formulated as follows: the development of a method for
processing and systematization of various types of diagnostic information (a deterministic approach to
the recognition of TS).
225
3. Deterministic method of systems technical state recognition
When solving the problem of deterministic predicting in direct formulation, the sought-for
characteristics are the values of the diagnostic parameters. Since the state recognition of system with a
large number of diagnostic parameters is associated with significant difficulties associated with
monitoring and systematization of various types of parameters, they are set to minimum for prediction
from the condition of ensuring the required reliability of the prediction. In practice, usually one
diagnostic parameter is used – critical or integrated. If there is a critical parameter, this greatly
simplifies the diagnostic process. A situation is possible where the control of a critical parameter is
rather difficult due to various reasons and it is more economical for an enterprise to admit a failure,
but to be ready for it than to spend huge resources on diagnosing. The integrated diagnostic parameter
requires the application of special mathematical methods of processing and systematization of
heterogeneous data - various diagnostic parameters. We can definitely say that to increase the
reliability of solutions it is necessary to apply a set of methods.
TS, being a complex property of the object without a definite formalization is difficult for the
perception of the decision-maker. Thus, there is an urgent task of developing a mathematical model
that systematizes various types of diagnostic data for making decisions in the technical diagnostics
tasks.
The object TS can be characterized (estimated) by means of some finite set of parameter values:
С { p1 ; p2 ; p3 ;... pn }, (4)
where С – TS of object;
pn – technical parameter value;
n – number of parameters.
In figure 2 on one axis five TS are marked, defined by the standards [2, 3]. If we assume that the
TS can be quantitatively expressed by the number I from 0 to 1, where 1 means a perfect TS, then
using the numerical representation of the TS instead of discrete values: perfect, imperfect, up state,
down state and limiting state, one can observe "Serviceability degree" of the object. This
dimensionless quantitative evaluation of the TS is called the Technical State Index (TSI) [8, 9]. The
value of TSI characterizes the object state from the point of view of its parameters' compliance with
normative values. The parameter is the passport characteristic of the object.
Figure 2. Graphic image of TS.
Thus, any object TS can be expressed using a finite set of TSI:
С {I1 ; I 2 ; I 3 ;...; I n }, (5)
where С – TS of object,
In – TSI according to parameter n,
n – number of parameters.
Then TSI Ii reflects the degree of correspondence of the i-th parameter to the required (ideal, nominal)
value.
226
It should be noted that in the classical reliability theory [4-6, 10] the system state at time t is described
by a random vector:
X (t ) X 1 (t ),..., X n (t ) , (6)
where Xi(t) – vector components, which can be the values of various system parameters, capable of
taking values on the entire real axis. X(t) can be a one-dimensional variable that always takes two
values: 1 if the system is operable, and 0 if the system is in a failed state. A random vector X(t) is
characterized by probability distribution:
F x1 ,..., xn , t , (7)
that is, the probability that
X 1 (t ) x1 ,..., X n (t ) xn . (8)
The state of the system can be described in various ways. The fuzzy sets theory [11] will be used as
a theoretical basis in this paper.
Assume that the TS is characterized by a single parameter x, then in the view of set theory we can
write:
x E , (9)
where E – set of values (range) of this parameter.
If E is the set of all possible values of the parameter, then it necessarily contains some subset of the
required (admissible) values of the parameter; denoting this subset as A, we get:
A E, (10)
If the element x of set E is the element of subset A (in other words, the value of the parameter x is
in the required (admissible) range of values), then:
x A. (11)
To express this membership in set theory, use the notion – the characteristic function whose value
indicates whether (yes or no) x is an element of A:
1, if x A,
A ( x) (12)
0, if x A.
The peculiarity of this function is in the binary nature of its values (figure 3).
Figure 3. The characteristic function for the set A.
In the production application, the characteristic function indicates whether the given technical
parameter of an element satisfies the required values. That is, TSI is a dimensionless estimate. The
227
characteristic function shows TSI dependence on specific values of the parameter. Figure 4 provides a
graphic explanation of the above.
Figure 4. Graphic explanation of TSI.
Figure 4 uses the following notation: Ii – ITS of the i-th parameter, pi – value of the i-th parameter,
pimin des., pirat. des., pimax des. – minimum, rated and maximum desired (admissible) value accordingly.
However, in practice, the boundary between admissible and inadmissible values of the parameter
may not have a clear character, which is why it is rational to consider the solution of the problem of
evaluating the TS in the view of the fuzzy sets theory. Then the characteristic function can take any
value in the interval [0; 1]. In accordance with this, the element xi of the set E can belong to A to a
certain degree A ( x ) .
Thus, according to [11], with the help of the notion of fuzzy subset it is possible to study
nonrigorous defined concepts (such as TS) using mathematical structures. Then the estimation of the
object TS will take its values in the set M = [0; 1] by means of a certain characteristic membership
function. And TSI is the current value of the characteristic membership function.
Since each parameter is unique (the set of all possible values, a subset of the required values, the
dimension, etc.), then a different membership function is constructed for each parameter. The basis for
constructing the membership function is the object manufacturer documentation. Let's consider
examples of membership functions of belonging of physical parameters of a technical object – power
unit (tables 1, 2).
Table 1. Diagnostic parameters of the power unit.
№ Diagnostic parameter Current value Minimum Maximum
acceptable value acceptable value
1 Operation life, years 6 0 10
2 Output voltage, V 23.5 23.0 24.0
Output voltage ripple,
3 400 0 1000
mV
Average failure time,
4 9.5 10 -
thousand hours
228
Table 2. Membership functions and TSIs of the power unit.
№ Membership function Current TSI
1 I1 = 0,40
2 I2 = 1,00
3 I3 = 0,60
4 I4 ≈ 0,46
I1 – TSI of power unit by parameter p1 – «Operation life ». Minimum acceptable value of parameter
p1 0 years, maximum acceptable value of parameter p1max accept . 10 years – this value is
min accept .
229
emergency, that is, the excess of the operation life declared by the manufacturer in the documentation
for the object increases the risk of failure.
I2 – TSI of power unit by parameter p2 – «Output voltage». Minimum acceptable value of
parameter p2min accept . 23,0 V, maximum acceptable value of parameter p2max accept . 24,0 V. In this
case p2min accept . and p2max accept . are emergency. The rated (ideal) value in this case is equal to
p2rat . 0,5 ( p2min accept . p2max accept . ).
I3 – TSI of power unit by parameter p3 – «Output voltage ripple». Minimum acceptable value of
parameter p3min accept . 0 mV, maximum acceptable value of parameter p3max accept . 1000 mV.
I4 – TSI of power unit by parameter p4 – «Average failure time». Minimum acceptable value of
parameter p4min accept . 10 thousand hours, minimum emergency value of parameter p4min emerg . 0
thousand hours. The normal Gauss distribution is chosen as the dependence I 4 ( p4 ) , where the mean-
square distance of the distribution is assumed to be σ = 0.4.
In this example, one of the parameters for evaluating TS is "Average failure time". This parameter
refers to statistical parameters of the classical reliability theory. It should be noted that authors of the
paper intentionally included it in a number of parameters for TS estimating to show the possibility of
using different types of data in the TS estimating. However, this issue requires further research. There
may be a situation where the object TSI will be close to 1, but the object life cycle at the stage of its
operation shows that it often fails. Against this background, it is quite difficult to work out a decision
regarding the attribution of the object to "serviceable" or "faulty". But, nevertheless, it can be stated
that, either a lot of parameters, by which the object is diagnosed, are not sufficient to determine the
exact diagnosis, or the object does not have the property of controllability. In this situation, many
parameters (deterministic component) should be reviewed, or the deterministic component in (1)
should be neglected in order to optimize the diagnostic process. In any case, the fact of a large
frequency of object failures when finding all diagnosed parameters in the acceptable zone should be
taken into account when evaluating its TS.
Thus, the TS of object A in terms of the fuzzy sets theory can be described (formalized) as follows
[12]:
x1 x2 ... xn
, (13)
A% A%1 ( x1 ) A%2 ( x2 ) ... A%n ( xn )
where xi – parameters for evaluating of TS of object A;
n – number of parameters;
A%i ( xi ) – membership function (each parameter has the required values subset A%i , and accordingly,
has its own membership function).
This structure forms the fuzzy reliability model of technical object.
If all current values of the parameters are aligned in the proposed model, the set A%will characterize
the current TS of the object:
x1 x2 ... xn
, (14)
A%curr . A%1 ( x1 curr. ) A%2 ( x2 curr. ) ... A%n ( xn curr. )
where xi curr. – current values of object parameters.
If the object TS is ideal (rated) for all parameters, then its state can be described as follows:
x1 x2 ... xn
(15)
A% 1 1 ... 1 Arat . .
230
That is, the ideal state of the object can be described in the form of a crisp set, the elements of
which are n ones.
To determine the integrated TS, we use two estimates – the relative linear distance (Hamming
distance) and the relative quadratic distance (Euclidean distance) between set A%curr. and set Arat. . As is
known, these two concepts give two estimates of the distance between fuzzy sets.
The relative linear distance between the above sets is determined by the formula:
( A%curr . , Arat . ) n 1 A ( xi ) A% ( xi ) n 1 1 A% ( xi ) .
n n
rat . curr . curr . (16)
i 1 i 1
Formula (16) can be transformed:
n
( A%curr . , Arat . ) 1 n 1 A% ( xi ) 1 Aarithm
curr . %
. av.
( xi ),
curr . (17)
i 1
where Aarithm
%curr .
.av.
( xi ) – arithmetic average of all values A%curr . ( xi ) .
The relative quadratic distance between the above sets is determined by the formula:
1/2 1/2
2
2
( A%curr . , Arat . ) n1 Arat . ( xi ) A%curr . ( xi ) n 1 1 A%curr . ( xi )
n n
(18)
i 1
i 1
Both relative distances satisfy the conditions:
0 ( A%curr . , Arat . ) 1, (19)
0 ( A%curr . , Arat . ) 1. (20)
Thus, using relative distances ( A%curr . , Arat . ) and ( A%curr . , Arat . ) , we can not only assess the
integrated state of the object, but also compare objects with each other (ranking).
Here is an example of calculating the integrated TSI of power unit, discussed above. The current
TS of the power unit is described using the reliability fuzzy model:
C 0, 40;1,00;0,60;0, 46. (21)
The reliability model for a given power unit with ideal TS:
C 1, 00;1, 00;1, 00;1, 00 . (22)
Then integrated TSI of the power unit is I = 0,62.
The presented model of object TS description (estimation) allows to solve a problem of
"incomparability" of objects among themselves. Even under the condition of different number of
parameters, according to which the TS of objects is evaluated, the possibility of comparing (ranking)
objects is opened up with the help of reliability fuzzy model. The ranking of objects occurs on a
unifying basis – the TS.
To construct the reliability model of a complex system (figure 5), a structurally-parametric method
for analyzing hierarchies – hierarchical decomposition is used [13].
231
Figure 5. Fuzzy reliability model of system.
The following notations are used in figure 5: I i –TSI of the i-th subsystem, I ij – TSI of the j-th
element of the i-th subsystem, I ilk – TSI by the k-th parameter of the j-th element of the i-th subsystem,
xijk – the k-th parameter of the j-th element of the i-th subsystem.
Thus, in order to carry out an evaluation of the TS of a complex system consisting of set of
elements, it is necessary to estimate the TS of each element of the system separately by the parameters
established for each element, then calculate the integrated TSI of hierarchical structures of the system.
It is modified in accordance with the evaluation of the manifestation of additivity and integrity
properties in the proposed model, (figure 6).
Figure 6. Modified fuzzy reliability model of system.
In figure 6 the system parameters and places that they occupy in the general hierarchical structure
are highlighted in color. The presented model makes it possible to take into account both the
parameters of the elements and the system parameters, which allows to balance the additivity and
integrity in the integrated TS system evaluation (including industrial equipment).
The fuzzy reliability model is a model of system analysis for technical diagnostics tasks to the full
extent. Various information management systems (MIS) and decision support systems (DSS) can be
developed on its basis.
MIS and DSS, built on the basis of a reliability fuzzy model, can provide a decision-maker with
organized information on the equipment condition. As a result, it will allow to perform coordinated
work, to make grounded administrative decisions and to increase management process efficiency on
the whole.
4. Summary and conclusion
TSI is an integrated indicator of system reliability, which systematizes heterogeneous diagnostic
parameters. TSI, like probabilistic characteristics, is a dimensionless value with a range of values from
0 to 1, which introduces an additional advantage when used together.
232
It is possible to increase accuracy in engineering diagnostics tasks only by means of the combined
use of any and all methods. The TS should be evaluated on the basis of both a deterministic approach
and a probabilistic approach (table 3) in order to reduce the risk of equipment failures.
Table 3. Comparative features of deterministic and probabilistic principles of systems technical
state recognition.
Deterministic component Probabilistic component
Failure causes determined does not matter
Failure frequency relatively high relatively low
immediate values of diagnostic
Predicted (evaluated) result probabilistic assessments
parameters, TSI
Obtaining information methods diagnosis statistical-probabilistic
Mathematical apparatus fuzzy sets theory probability theory
Theoretical basis reliability theory
Thus, the fuzzy reliability model and the classical reliability model (statistical-probabilistic method
of TS recognition) are complementary. Their integrated application increases the accuracy of the
decisions made, as well as the reliability of the systems.
5. References
[1] GOST 20911-89. Technicheskaya diagnostika. Terminy i opredeleniia [GOST 20911-89.
Technical diagnostics. Terms and definitions]. Moscow, Standartinform Publ., 2009. p 11.
[2] GOST 27.002-2015. Nadezhnost v tekhnike. Terminy i opredeleniia [GOST 27.002-2015.
Dependability in technics. Terms and definitions]. Moscow, Standartinform Publ., 2016. p 23.
[3] GOST 18322-2016. Sistema technicheskogo obsluzhivaniia i remonta techniki. Terminy i
opredeleniia [GOST 18322-2016. Maintenance and repair system of engineering. Terms and
definitions]. Moscow, Standartinform Publ., 2017. p 13.
[4] Birger I.A. Technicheskaya diagnostika Moskow, Mashinostroenie Publ., 1978. p 240.
[5] Barlow R., Proshan F., Matematicheskaya teoriya nadezhnosti. Per. s angl. [Mathematical
theory of reliability. Translated from English] Moscow, Sovetskoe Radio Publ., 1969. p 488.
[6] Barlow R., Proshan F., Statisticheskaya teoriya nadezhnosti. Per. s angl. [Statistical theory of
reliability. Translated from English] Moscow, Nayka Publ., 1984. p 328.
[7] Gavrilyuk E.A., Mantserov S.A., Panov A.Y., Prognozirovanie otkazov sistem
avtomaticheskogo upravleniya gazoperekachivayushhimi agregatami na osnove indeksa
texnicheskogo sostoyaniya i stepeni riska [The failure prediction of automatic gas-compressor
unit control systems on basis of technical state index and measure of risk]. Fundamentalnye
issledovaniya, 2015, no. 7-2, pp 309-313.
[8] Gavrilyuk E.A., Mantserov S.A., Sinichkin S.G., Kompleksnaya ocenka texnicheskogo
sostoyaniya sistem avtomaticheskogo upravleniya gazoperekachivayushhimi agregatami [The
complex technical state evaluation of automatic gas-compressor unit control systems].
Fundamentalnye issledovaniya, 2014, no. 11, pp 2141-2145.
[9] Gavriliuk E.A., Mantserov S.A. Razrabotka strategii obsluzhivania i remonta oborudovaniia
gazotransportnogo predpriiatiia na osnove indeksa technicheskogo sostoianiia [Development
of gas-transport equipment maintenance strategy on basis of technical state index]. NGTU,
2017, no. 3(118), pp 121-126.
[10] Shklyar V.N. Nadezhnost` sistem ypravleniya: ychebnoe posobie Tomsk, Tomsk politechnical
university Publ., 2009. p 126.
[11] Kofman A., Vvedenie v teoriiu nechetkih mnogestv [Fuzzy sets theory]. Moscow, 1982. p 432.
233
[12] Gavriliuk E.A., Mantserov S.A. Upravlenie technicheskim sostoyaniem slozhnich sistem na
osnove nechetkoi modeli [Technical state control of complex systems via fuzzy model].
Avtomatizacia processov upravlenia, 2018, no. 1(51), pp 91-98.
[13] Gavriliuk E.A., Mantserov S.A. Nechetkaya model nadezhnosti technicheskich sistem
[Technical systems dependability fuzzy model]. Politech. University Publ., 2018, vol. 2., pp
197-206
234