Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

JCSS Workshop on Reliability Based Code Calibration

Reliability based calibration of Eurocodes considering a


steel member
Haig Gulvanessian Milan Holický
BRE Watford, UK KI CTU in Prague, Czech Republic

Abstract
The reliability based study of partial and combination factors for various actions and
material properties concerns design methods recommended in EN 1990. Submitted
investigation of a steel structural member is primarily focused on application of
expression (6.10) and twin expressions (6.10a) and (6.10b) of prEN 1990. It appears
that partial factors in both cases may be adjusted to achieve required reliability level.
Modified expression (6.10a), considering permanent action only, seems to give an
unacceptable solution for countries of both authors. More investigations are required
to make a choice for the National Annex.

Keywords: Partial factor, Combination factor, Steel member, Reliability

1 Introduction
The reliability study used for the investigation described in this paper, is based on
probabilistic methods using the probability of failure Pf and reliability index β as
defined in EN 1990 [1] and ISO 2394 [2]. For Annex A of EN 1990 [1] Nationally
Determined Parameters need to be specified in the National Annex, where EN 1990
allows for National choice. Decisions concerning the choice of expressions (6.10) or,
alternatively, twin equations (6.10a) and (6.10b) with possible simplification of (6.10a)
using permanent actions only, together with the choice of the partial factors γG and γQ
for permanent and variable actions and ψ factors used in combination of actions is
one of the most important issues for the National Annex. Compared to ENV 1991-1
[3], which provided two alternative rules for combination of actions EN 1990 provides
three. In addition, EN 1990 will be a fully operative Code of Practice and therefore a
comprehensive study to establish the nationally determined parameters is essential.

Several recent studies [4,5,6,7] have investigated this important question.

The extensive study by SAKO [4] presents an analysis of alternative procedures


concerning the combination rules included in ENV 1991-1 and verification of the
partial safety factors for actions and materials. The results obtained indicate that
structural members made of different materials and designed according to the
alternative ENV 1991-1 procedures for combination of actions have a significant
variation in the reliability level.

The SAKO report shows that the calibration of partial safety factors within the
methodological principles provided in ENV 1991-1 is a complex task. It also shows
that it is impossible to achieve approximate constant levels of the reliability index β if
all the recommended partial safety factors are specified in accordance with [1] for

1
JCSS Workshop on Reliability Based Code Calibration

different proportions of permanent and variable actions. However, the SAKO


proposes gives a general procedure, on obtaining the minimum variation in reliability
index β.

A further study, based on the SAKO investigation, on the probabilistic calibration of


the partial safety factors was prepared in Finland [5]. An alarming range in reliability
for various structural members made from different materials indicates an urgent
need for a comprehensive international analysis with an agreed basis for reliability
models.

Another calibration study [6] shows that a more uniform reliability level has been
obtained in the newly revised Danish structural code using a reliability-based
calibration and optimisation technique.

The study [7] shows that a possible modification of safety factors and other reliability
elements from those recommended in EN 1990 [1], should be supported by
comprehensive studies and preferably limited to those cases, which are linked to
national (regional) conditions, e.g. to climatic actions and their combinations. This
recommendation is supported by a probabilistic optimisation, which also considers
the total expected cost for the life cycle of the structure as an objective function; and
the partial factors γG and γQ of permanent and imposed loads as decisive parameters.
From [7] it appears that the optimum values of the partial factors γG and γQ are very
close to the values recommended in EN 1990 [1].

This paper attempts to extend the previous studies on load factors by a reliability
study of the Eurocode rules for combination of actions, considering a steel structural
member. In addition, the EN 1990 combination rules are compared with those
specified in BS 5950 [10] and BS 8110 [11], and the National Application Document
to ENV 1991-1 included in the Czech prestandard [12]. This study is based on the
probabilistic concepts codified in EN 1990 [1], ISO 2394 [2] and on the reliability
methods recommended by JCSS [8] [9]. The study presented here utilises the
authors’ own computer programmes developed for the purpose of calibration of
safety factors using MATHCAD software. Selected numerical results were however
checked using software COMREL [14].

This study represents a very transparent investigation to identify preliminary


conclusions and items for future research, which will enable member states to initiate
their program of work to produce National Annexes for EN 1990..

2 Design in accordance with EN 1990


The equation for investigating the hypothetical structural member is written in a
general form covering three combination rules provided in EN 1990 by expressions
(6.10), (6.10a), (6.10b) and modified expression (6.10a) as

Rk /γΜ = (ξ) γG Gk + γQ (ψQ) Qk + γW (ψW) Wk (1)

It is assumed that quantities Gk, Qk and Wk denote appropriate load effects and then,
in case of linear behaviour, equation (1) is generally applicable for various structural

2
JCSS Workshop on Reliability Based Code Calibration

members (of different material – not only steel). Symbol γΜ denotes the partial factor
of material property, γG, γQ and γW denote partial factors of the permanent action G,
variable action Q and variable action W respectively. In order to keep the study more
general, different partial factors γQ and γW are introduced in equation (1), even though
it is recommended in EN 1990 to use the same safety factor γQ (=1,5) for all variable
actions.

In equation (1) the factors, ξ, ψQ and ψW are given in brackets. These parameters
given in the brackets should be applied accordingly to the rules described in EN 1990
for expressions (6.10), (6.10a) and (6.10b). This method of writing symbols in
brackets (where values of symbols given in brackets can be changed depending on
the combination of action rule that is used) will be applied in this paper in order to
simplify the equations.

The reduction factor ξ should be applied in accordance with equation (6.10b) to


reduce the permanent action G. Similarly, the combination factors ψQ and ψW should
be applied in accordance with expressions (6.10), (6.10a) and (6.10b) to reduce
action effects due to the variable actions. For example, in equation (2) below, which
corresponds to expression (6.10) of EN 1990 the factors ξ, ψQ and ψW given in
brackets in equation (1) will be 1,0, 1,0 and 0,6 respectively.

When combination of actions given in expression (6.10) of EN 1990 (denoted in this


study as combination A) is used and Qk is the leading action, then the design formula
(1) becomes

Rk /γΜ = γG Gk + γQ Qk + γW ψW Wk (2)

Equation (2) also corresponds to UK combination rule provided in BS 5950 [10] and
BS 8110 [11] and Czech combination rule provided in CSN P ENV 1991/1 [13].
However, the partial safety factors γG, γQ and γW are different from those
recommended in EN 1990 [1]. The partial safety factors in the UK codes make an
allowance for simultaneity of actions.

Similarly, for the combination rule (6.10a) equation (1) has the form

Rk /γΜ = γG Gk + γQ ψQ Qk + γW ψW Wk (3)

In equation (3) both variable actions are reduced by appropriate ψ factors. For
combination rule (6.10b) equation (1) becomes

Rk /γΜ = ξ γG Gk + γQ Qk + γW ψW Wk (4)

Here, the permanent action is reduced by ξ factor while the leading variable action Qk
is not reduced. When the pair of expressions (6.10a) and (6.10b) (which should
always be applied together) are used (denoted in this study as combination B) then
the structural members should be investigated using equations (3) and (4) together.

3
JCSS Workshop on Reliability Based Code Calibration

Finally, if the modified expression (6.10a), which has to be used together with
(6.10b), is being used (denoted in this study as combination C) then only permanent
actions are considered, equation (1) reduces to a simple form

Rk /γΜ = γG Gk (5)

The structural members, in this case, should be investigated using equations (4) and
(5) together.

The steel structural member considered in this paper might be a beam, or any other
structural member, that may be designed using the linear equation (1), in the
appropriate form of equations (2) to (5). In these equations Rk denotes the
characteristic value of the member resistance R, and Gk, Qk, Wk denote the
characteristic values of the permanent action G, and two variable actions Q and W
respectively. It should be noted that Gk, Qk, Wk generally denotes the relevant action
effects (not just directly the actions themselves), which are therefore, dependent on
load, structural arrangement and geometry of the structural member.

Taking into account linearity of the design relationship (1), the whole investigation
may be normalised choosing, for example, Rd = Rk /γΜ = 1 (in appropriate units). This
simplification is accepted in the developed programme using the MATHCAD
software. The characteristic values of the actions Gk, Qk, Wk must be obviously
expressed in the same units as the resistance Rk.

To investigate the effect of variable actions on the reliability of the structural member,
the characteristic values Gk, Qk, Wk are mutually related using the load ratio χ of the
variable action Qk+Wk to the total load Gk+Qk+Wk and the variable load factor k as a
ratio of the second variable action Wk to the first variable action Qk as follows

χ = (Qk+Wk)/(Gk+Qk+Wk),
(6)
k = Wk /Qk

The factor k takes an account of such condition that when two variable actions are
acting together, the ratio of their characteristic values (e.g. for imposed load and wind
action) may be different for different structural members, depending on the location of
the structural member within a structure.

Using equations (1) and (6), the characteristic values of actions Gk, Qk, Wk may be
expressed in terms of the load ratio χ and factor k for a given characteristic value Rk
of the resistance R, as

R k /γ M
Gk =
((ψ Q )γ Q + k (ψ W )γ W )χ
(ξ )γ G +
(1 + k )(1 − χ )
(7)
χ Gk
Qk =
(1 + k )(1 - χ )

4
JCSS Workshop on Reliability Based Code Calibration

It should be noted that expression (6.10) in EN 1990 is applicable for the whole
possible range of the load ratio 0 ≤ χ ≤ 1. Expression (6.10a) is applicable for the
interval 0≤ χ ≤ χlim and expression (6.10b) for the interval χlim ≤ χ ≤ 1, where the
relevant action effects given by one of the expressions is greater than the other. The
limit values χlim,B and χlim,C separating both domains in combinations B and C can be
found from equations (3), (4), (7) and expressions (6.10a) and (6.10b) in EN 1990 as

γ G (1 − ξ )(1 + k )
χlim,B= (8a)
γ G (1 − ξ )(1 + k ) + γ Q (a − ψ Q ) + γ W k (b − ψ W )

γ G (1 − ξ )(1 + k )
χlim,C= (8b)
γ G (1 − ξ )(1 + k ) + γ Qa + γ W kb

where auxiliary quantity a = 1 if k ≤ (1-ψQ)/(1-ψW), and a = ψQ if k > (1-ψQ)/(1-ψW),


similarly quantity b = ψW if k ≤ (1-ψQ)/(1-ψW), and b = 1 if k > (1-ψQ)/(1-ψW).

Within the Section 2 all variables are considered as deterministic quantities. In the
following Section 3 on reliability analysis of the hypothetical steel structural member,
the resistance R and actions G, Q, W and additionally introduced model factors
describing model uncertainty of the action effect and resistance are taken as random
variables.

3 Limit state function


Reliability analysis should first consider the limit state function g(X) corresponding to
the design relationship given in equation (1)

g(X) = θR R – θE (G + Q + W) (9)

Here X denotes the vector of basic variables (random variables entering the right
hand side of this equation), θR denotes the factor expressing the uncertainty of the
resistance model and θE the factor expressing the uncertainty of the loading model.

It should be noted that the limit state function given in equation (9) corresponds in
general to the design relationship given in equation (1), and in addition to the derived
equations (2) to (5). The reliability level of the structural member, which is being
designed using any of these equations (2) to (5) may be checked using the limit state
function given in equation (9). Obviously, only those variable actions applicable for
the design of the structural member should be considered in the limit state function
described by equation (9).

A critical ingredient in any reliability analysis is the specification of probabilistic


models for the basic variables. The limit state function, equation (9), has six basic
variables that describe the actions G, Q, and W, the resistance R and the model
uncertainties θR and θE. Simplified (e.g. averaged over a common loaded area)
probabilistic models are indicated in Table 1. These models may be refined when
particular loaded areas and structural conditions are considered.

5
JCSS Workshop on Reliability Based Code Calibration

The simplified probabilistic models of basic variables indicated in Table 1 are


primarily intended to be used as "conventional models" in time invariant reliability
analysis of structural members using Turkstra's combination rule [12], for example,
for the probabilistic calibration of the rules for combination of actions. Conventional
models should enable the objective comparison of results of various reliability studies
expected in the near future in connection with implementation of the present suite of
Eurocodes into the national systems of design codes. However, when the reliability of
different types of structural members under particular conditions is assessed, the
proposed models in Table 1 may have to be modified.

Table 1. Simplified probabilistic models of basic variables for time invariant reliability
analysis using Turkstra's rule (combination of 50 years maximum of Q and an annual
W).
No. Category of Name of basic Sym. Dim- Distrib Mean St. d. Prob.
variables variables X ension -ution µX σX ΦX(Xk)
1 Actions Permanent G kN/m2 N Gk 0,1µX 0,5
2 Imposed - 5 years Q kN/m2 GU 0,2Qk 1,1µX 0,995
2 Imposed - 50 y. Q kN/m2 GU 0,6Qk 0,35µX 0,954
3 Wind - 1 year W kN/m2 GU 0,5Wk 0,4µX 0,98
4 Wind - 50 year W kN/m2 GU 0,7Wk 0,25µX 0,94
4 Resistance Steel R kN/m2 LN Rk +2σ 0,08µX 0,016
5 Model Action effect fact. θE - N 1,00 0,10 0,5
6 uncertainties Resistance factor θR - N 1,10 0,07 0,16

In Table 1 it should be noted that the mean µX and standard deviation σX of each
variable are intentionally related to the characteristic value Xk used in the design
calculations. The last column of Table 1 indicates the actual probability of the
occurrence of values smaller than the characteristic value Xk, thus the probability

P{X<Xk}= ΦX(Xk) (10)

where ΦX denotes the relevant distribution function of the basic variable X. It should
be mentioned that due to several reasons (e.g. quality control, historical development
of codified values) these probabilities in general differ from those recommended for
the specification of the characteristic values Xk in the Eurocodes. For example, the
characteristic value of the resistance Rk has the probability of being infringed only
0,016 and not 0,05, as recommended in EN 1990. Only for the climatic action W
(wind or snow), the theoretical model is determined in such a way so that the
probability of the characteristic value Wk being exceeded by the annual maximum
corresponding to the recommended value 0,02.

It should be noted that, knowing the coefficient of variation w, the mean µ of a


material property, may be assessed from the characteristic value Xk using the
expression

µX = Xk/(1-2 wX) (11)

6
JCSS Workshop on Reliability Based Code Calibration

This relationship corresponds to the rules indicated in Table 1 and is applied in the
following reliability analysis. Equation (11) is well supported by available
experimental data obtained for steel. For example, in the case of steel S235, the
experimentally established mean 280 MPa (confirmed by several investigations)
satisfies equation (11) if the coefficient of variation 0,08, indicated in Table 1, is
considered.

The relationship given in equation (11) does not necessarily apply for reinforced
concrete members when two materials (reinforcing bars and concrete) are used. In
this case equation (11) may be used to obtain a first estimate of the mean resistance
only. For the reliability analysis in this investigation a more detailed analysis of the
composite cross section has been applied

The coefficients of model uncertainty θE and θR are described by random variables as


indicated in [8] (having the coefficients of variation of 0,10 and 0,07 respectively). For
other specific cases (depending on the material) different values of the mean and
standard deviation may be more appropriate. For example, the mean of the model
uncertainty θR may be greater than 1,1, e.g. 1,2 in case of a steel structural member
exposed to compression.

4 Reliability analysis
Using the limit state function given in equation (9), (as applicable to expressions
(6.10), (6.10a) and (6.10b), and (6.10a) modified and (6.10b)), the following five
cases of combination of actions (denoted by letters A, B, C, D and E) were
investigated for a steel member.

A) Combination based on equation (6.10) for γM=1,10, γG=1,35, γQ=1,5, γW=1,5.


B) Combination based on equation (6.10a) and (6.10b), γM=1,10, γG=1,35, γQ=1,5,
γW=1,5.
C) Combination based on equation (6.10a) and (6.10b), for γM=1,10, γG=1,35, γQ=1,5,
γW=1,5.
D) Czech combination based on equation (6.10), for γM=1,15, γG=1,20, γQ=1,4,
γW=1,4.
E) UK combination based on equation (6.10), for γM=1,05 derived from data provided
in BS 5950 and for the partial factors for actions provided in BS 8110 as given below.
F) Combination based on equation (6.10) for γM=1,00, γG=1,35, γQ=1,5, γW=1,5.

The partial safety factors for actions were applied in accordance with provisions
indicated in BS 8110 as follows:

γG = 1,40 and γQ = 1.60 and γW = 1,40 when considering one variable action,
γG = 1,20 and γQ = γW = 1,20 when considering two variable actions.

The following combination coefficients were used in the combinations A, B, C and D:


ψQ = 0,70, ψW = 0,60 and ξ = 0,85.

Figures 1 and 2 show the results obtained assuming the probabilistic models of the
basic variables defined in Table 1. It was considered that the imposed load Q is

7
JCSS Workshop on Reliability Based Code Calibration

dominant and wind action W accompanying. Two values of the factor k are
considered here: k = 0 and 0,5. Figures 1 and 2 show the reliability index β as a
function of the load ratio χ. The limiting value χlim,B and χlim,C given by equation (8),
separating the domains of equations (3) and (4) in combinations B, and (5) and (4) in
combinations C, is shown in Figures 1 and 2 by the vertical dashed lines. The
reliability level β = 3,8, recommended by EN 1990 [1] (in Table B 2), is indicated in
Figures 1 to 12 by the horizontal dashed lines.

β A

4.5 D

B B E
4

C C
3.5

χ lim ,C χ lim ,B .
3
0 0.2 0.4 0.6
χ 0.8 1

Figure 1. Reliability index β versus load ratio χ for k=0 and load combinations A, B,
C, D and E.

5
A
β
D
4.5
B B

E
4
C

C
3.5

χlim,C χlim,B
3
0 0.2 0.4 χ 0.6 0.8 1

Figure 2. Reliability index β versus load ratio χ for k=0,5 and load combinations A, B,
C, D and E..

8
JCSS Workshop on Reliability Based Code Calibration

In recent drafts of EN 1993-1 partial factor γM = 1 (instead of previously used value γM


= 1,10) is indicated as a recommended value. Obviously this decrease in γM would
lead to a decrease of reliability of steel members. This combination, denoted F) is
indicated in Figure 3 together with combination A) and D), all based on fundamental
combination rule described in EN 1990 by equation (6.10). Figure 3 shows again
reliability index β versus load ratio χ for k=0.

β A

4.5

F
4

3.5

.
3
0 0.2 0.4 0.6 0.8 1
χ

Figure 3. Reliability index β versus load ratio χ for k=0 and load combinations A, D
and F.

It follows from Figure 3 that the combination F provide lower reliability than
combination A and D, however still acceptable for load ratio χ within the important
range from 0,15 to 0,6. Comparing Figure 1 and 3 it follows that the partial factor γM =
1 could be applied only in the load combination based on equation (6.10) as a
modification of the combination A) (not as a modification of combination B) and C)).

5 Reliability differentiation
Methods for reliability differentiation are generally described in Annex B of EN 1990.
It is recommended to multiply load effect by factor KFI, having values 0.9, 1.0 and 1,1
for decreased normal and increased reliability respectively. Thus, equation (1)
becomes

Rk /γΜ = KFI [(ξ) γG Gk + γQ (ψQ) Qk + γW (ψW) Wk] (12)

9
JCSS Workshop on Reliability Based Code Calibration

Considering a steel member designed using equation (4) and load combination B,
figure 3 shows reliability index β versus KFI factor for the load factor χ = 0,3 and 0,6
assuming factor k=0.

4.5

4
β = 3,8

3.5

3
0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2
KFI

Figure 3. Reliability index β versus reliability differentiation factor KFI for two load
ratios χ = 0,3 and 0,6, for k = 0 and load combination B.

Figure 3 confirms that the factor KFI = 0,9 would lead to a decrease of the reliability
index β by about 0,5, while KFI = 1,1 would increase β by about 0,5 as indicated in
EN 1990. Similar results as for the factor k = 0 (considered in Figure 3) are obtained
also for k ≠ 0.

However, Figure 3 also indicates that the resulting effect of the simple operational
approach recommended for reliability differentiation in EN 1990 depends also on the
load ratio χ (and also on factor k). Thus the operational method of reliability
differentiation may further increase inconsistency in reliability level of structural
members having the same significance.

6 Discussion of obtained results and conclusions


Figures 1 to 2, obtained from this initial study, show that

a) Expression (6.10), when used with the partial factors recommended by EN 1990,
give for most practical cases of load ratio χ reliability levels above those desired (β =
3,8). The use of expressions (6.10a) and (6.10b) give reliability levels less than for
expression (6.10), but still in most cases above those desired (β = 3,8).

10
JCSS Workshop on Reliability Based Code Calibration

b) Modified expression (6.10a) used together with expression (6.10b) from EN 1990,
leads to a reliability level less than desired (β = 3,8), particularly when the load ratio χ
is close to zero (less than 0,2).

c) The combination rules in BS 5950 and BS 8110 lead to near similar results as
expression (6.10) from EN 1990, when considering one variable action only.
However, the use of the UK combination rules when two variable actions are being
considered together, leads to a substantially lower reliability than EN 1990
expressions (6.10) or (6.10a) and (6.10b), in particular for low load ratios χ.

d) The use of the combination D, in accordance with the Czech standard [13] based
on expressions (6.10), leads to a similar results as combination B, however to lower β
than combination A.

e) The use of expressions (6.10a) used together with (6.10b), leads to a more
uniform distribution of β, as the function of the load ratio χ, than expression (6.10).

f) The partial factor γM = 1 could be applied only in the load combination based on
equation (6.10) as a modification of the combination A)

f) Load factor KFI can be used for reliability differentiation; for a decrease or an
increase of the reliability index β by about 0,5 (as indicated in EN 1990), the factor KFI
should be 0,9 and 1,1 respectively.

The combinations of the permanent action and two variable actions considered in this
study indicate that in most cases, the obtained reliability level is not essentially
different from the one obtained when the permanent action is combined with one
variable action only. However, the combination effect of individual variables depends
on their probabilistic models. More significant differences may be expected when
refined theoretical models of relevant variables and when time variant reliability
analyses are considered. To cover such cases, further research is needed.

Considering the limited study of this paper, the National Annexes should not specify
the use of (6.10a) modified. The decision on the use of expression (6.10) and twin
expression (6.10) and (6.10b) cannot be made from this limited study. Moreover, this
investigation is directly applicable for bending and axial forces without buckling only.
Further work is required to investigate:

• probabilistic models including the parameters for the basic variables


used for the reliability analysis,

• additional materials (e.g composite steel and concrete, timber and


masonry) that will have different probabilistic models,

• additional structural members and other resistance modes (e.g


columns, joints, slabs, shear and punching shear) will have different
probabilistic models,

• a more comprehensive reliability investigation of reinforced concrete


and composite steel and concrete structural members.

11
JCSS Workshop on Reliability Based Code Calibration

7 References
[1] prEN 1990: Basis of Structural Design. (Transformation ENV 1991-1,
1994) Brussels, 2001.
[2] ISO 2394: General Principles on Reliability for Structures. Zurich, 1997.
[3] ENV 1991-1: Basis of Design. Brussels, 1994.
[4] SAKO; Joint Committee of NKB and INSTA-B. Basis of Design of Structures.
Proposal for Modification of Partial Safety Factors in Eurocodes. 1999.
[5] Ministry of the Environment, Housing and Building Department, Finland.
Probabilistic Calibration of Partial Safety Factors (Eurocode and Finish proposal). 2000.
[6] Sorensen J.D., Hansen S.O. and Nielsen T.A.: Partial Safety Factors and Target
Reliability Level in Danish Codes. Proc.: Safety, Risk, and Reliablity. IABSE, Malta,
2001, pp. 179-184.
[7] Holický M., Marková J.: Verification of load factors for concrete components by
reliability and optimization analysis: Background documents for implementing
Eurocodes. Progress in Structural Engineering and Materials, Vol. 2 No. 4, 2000, pp.
502-507.
[8] JCSS: Probabilistic model codes. Working materials, 2001.
[9] Vrouwenfelder T.: JCSS Probabilistic Model Code. Proc.: Safety, Risk, and
Reliablity. IABSE, Malts, 2001, pp. 65-70.
[10] BS 5950: British Standards Institution. Structural use of steel work in building:
Part 1: Code of practice for design in simple and continuous construction: hot rolled
sections. 1990.
[11] BS 8110: British Standards Institution. Structural use of concrete: Part 1: Code
of practice for design and construction. 1985.
[12] CSN P ENV 1991-1-1, Basis of Design. Brussels, CSNI Praha 1996.
[13] Tursktra, C.J. 1970 Application of Bayesian Decision Theory. Study No. 3
Structural Reliability and Codified Design. Solid Mechanics Division, University of
Waterloo, Ontario, Canada.
[14] RCP COMREL version 7.10. MUENCHEN, 1999.

Acknowledgement: This study is a part of the research project GACR 103/00/0758.

12

You might also like