* Membres du GT 22.06 (en septembre 1988) G. Orawski (RoyaumeUni), chef de file; G. Henrioul (Belgique), secrétaire; M. Barbarito (Italie); R de Weck
(Suisse); E. Ghannoum (Canada>; J.l. Gidiund (Suède); R Khadri (Algérie); R Kiessling (Rép. Féd. d’Allemagne); J.R Nolasco (Brésil); A.H. Peyrot (EtatsUnis);
D. de la Houssaye (France); K. Schjetne (Norvége); RD. Wells (Australie); A.B. Wood (Royaumeuni); R.J. Zoomer (PaysBas).
Load~ng and strength of overhead transmission Nnes
1. INTRODUCTION However, even before they were affected by an event, for the degree of
published, work had begun on their utilisation of various components, for a
1.1 Historical background selected sequence of failure etc.
revision to take into account various
Engineers involved n the internatio comments regarding presentation and Although the terms of reference of
nal activities of CIGRE realised, some techniques. The major effort was direc IEC TCII have not been changed, this
twenty or more years ago, that there ted towards a probabilistic assessment Technical Committee is now referred to
was a Iack of rationalisation in over of the performance of the unes by alto as “Loading and Strength of Overhead
head me regulations. Comprehension wing for the number of components Lines”.
and understanding were not easy even
among specialists. Deeper discussions
showed that concepts such as working
Ioads, ultimate loads, factors of safety, CHARGES ET RESISTANCES
working stress, permissible stress, LOADS AND STRENGTHS FIABILITE : RELIABILI7Y
allowable stress etc. did not necessari FACTORS ~ SECURITE SECURITY
ly convey the same images to aIl. It L SURETE
was decided by the CIGRE Study
Committees (SC6 and SC7) on over LOAO STRENGTH
head mes to form a new Working
Group on Factors of Safety (first mee
ting March 1961). LOAD S
STRENGTH
The present Working Group 22.06
under the name “Line Reliability and
Security” continues the work which Figure 1 A. Approche déterministe.
was started in 1961. Figure lA. Deterministic approach.
In the meantime, the International
Electrotechnical Commission (IEC) for
med a Technical Committee No. 11
“Recommendations for Overhead
Lines within which several Working
CUMULATIVE
Groups were nominated (1973/1974). STRENGTR
Thus for several years now both LOAO DISTRIBUTIONS
CIGRE Study committee 22 (which
inherited the activities of SC6 and DENSITY
SC7) and IEC TCll have followed
parallel paths. Luckily, a large propor
tion of the membership of the CIGRE R LOAOS
and IEC Working Groups was com STRENGTH
mon. Hence the findings within one Figure lB. Données statistiques.
Organisation were automatically avai Figure lB. Statistical data.
labletotheother.
CHARGE LOAD RESISTANCE STRENGTH
Within IEC TCII a stage has now
been reached when IEC Central Office
Reports have been published, under
the heading of:
“Overhead Line Support Loadings”
8261 (1985) Part 1 General
 
perature Ioadings R
8263 (1985) Part 3 Special loa
 
* SC22WGOS Membership (September 1988) G. Orawski (United Kingdom), Convenor; G. Henrioul (Belgium), Secretary; M. Barbarito (Italy); P. de Weck
(Switzerland); E. Ghannoum (Canada); J.!. Gidlund (Sweden); F. Khadri (Algeria); F. Kiessiing (Fed. Rep. cf Germany); J.F. Nolasco (Brazil); A.H. Peyrot (Uni
ted States); F. Pourbaix (France); K. Schjetne (Norway); F.D. WeIls (Australia); A.B. Wood (United Kingdom); R.J. Zoomer (Netherlands).
N°129 ELECTRA 67
Essentially, the IEC Documents wiIl The next step in the reasoning pro Security (structural) is the ability of a
be issued as “Reports” to help the cess is the relationship between these system to be protected from a major
work of those National Bodies which two distributions (Figure 1G). As a collapse (cascading effect) if a failure
are responsibie for the preparation of result of many discussions, calcula is triggered in a given component.
Specifications and Regulations. At a tions, reasoning, graph plotting etc. it Security is a deterministic concept as
Technicai Committee meeting in Fort has been agreed, for inclusion in IEC opposed to reuiability which is a proba
Worth (USA) in 1986 it was agreed documents, that the loads would be bilistic concept.
that there would be four parts in the related to a return period of an event Safety (structural) is the ability of a
new Publication. (50, 150 or 500 years) and that the system not to cause human injuries or
strength wouid be accepted as that ioss of lives. in these documents, safe
At a subsequent meeting of the which gives 90% probabiiity of being
Technical Committee in Buenos Aires ty relates mainly to protection of wor
achieved. Then, the area under the kers during construction and mainte
in December 1987, the decision was two overlapping curves is calculable,
made to publish them ail in a single nance operations. The safety of the
and the reliability can be assessed. public should be covered by National
document, with several chapters, each This is the probabilistic approach.
covering a particular aspect. Regulations.
However, it must be stated from the The purpose of this paper, together
it was felt within CIGRE 22.06 that a
outset, that the designer wili be given a with the 1EC documents is to acquaint
publication, by C1GRE, of additional deterministic value of load which he
explanations would heip in a national the reader with the tools which are
will compare with a deterministic value available and which can be used to
acceptance of the principles outlined in of strength, but the “specification wri
the documents. it must be noted that satisfy specific requirements.
ter” will have a fair knowledge of the
the paragraph numbering in this com reliability of the overail design, becau
mentary bears no relation to the num 1.3 ParaIlel activities
se he knows the coefficients of varia
bering in the IEC documents. The need for the publication of
tion of the load and of the strength.
documents explaining the probabilistic
1.2 Basic concepts This probabilistic approach to de approach to design is best illustrated
sign has other interlinked features: by the numerous papers which have
Most overhead line regulations been written on the subject of
1. ut s unfortunate that reliability which load/strength reiationships. Aiso, seve
define the loads to be applied to the is defined as probability of survival
elements of the une, they specify fac raI countries have already attempted to
is inseparable from risk (or probabi
tors of safety and refer to National prepare codes of practice incorpora
lity of failure). The area under the
Standards so far as materials are ting the concepts of reliability and
overlapping portions of the curves s
concerned. security.
a measure of the risk (P~~ and the
Figure lA, the socalied “determi complement to one is the reliability a) In the United Kingdom
nistic approach” iliustrates this state = 1—
The following documents are alrea
ment. It is assumed, for design pur 2. By designing components in such a dy available from the
poses, that the load has a unique, way that there is a probability that
defined, value and so has the strength British Standards institution:
some of them are stronger than the
of the component subjected to this other a coordination of strengths 1. BS 8100 : Part 1 1986 : Lattice
load. The separation between the load can be achieved, and the reiiability towers and masts Part 1 : Code of

and the strength was usuaily referred of the une wili be that of the weakest practice for loading
to, in the past, as “factor of safety”. it component. 2. BS 8100 : Part 2 1986 : Lattice

has now been agreed by Working towers and masts Part 2 : Guide to

Group members (CIGRE and IEC) that in order to avoid confusion, speciai the background and use of Part 1
the expression “factor of safety” would attention should be given to the iaws Code of practice for loading”
only be used when human ives are at used for the calculation of risk, in order 3. DD 133 : 1986 Draft for
stake. to differentiate between: deveiopment code of practice for
in real terms, when the une is — une failure and component (e.g. strength assessment of members of
constructed and subjected to meteoro tower) failure lattice towers and masts
iogical influences, the ioad will be a — climatic event causing a iocalised Essentially these documents are
variable quantity, conventionaliy failure and the extent of the failure applicable for the design or appraisal
expressed by a mean value (Q) and a (i.e. number of components affected of free standing tower structures of lat
standard deviation YQ (coefficient of by the event) tice construction. They include speciai
variation VQ)[VQ= (cr0/Q)] chapters on overhead unes. The Work
It may be appropriate at this stage ing Group members of CIGRE and IEC
Similariy, the strength in real life will
to recall the three basic definitions have taken cognisance of the British
not have a unique value, but it can be which have been accepted: documents.
described by a density (or cumulative)
strength distribution. Both distributions Reliability (structurai) is the probability b) In the United States
require a knowledge of statisticai data. that a system performs a given task,
Figure 1 B is an illustration of this state under a set of conditions, during a spe The American Society of Civil Engi
ment. At this stage, the separation bet cified time. Reuiabiuity is thus a measu neers (ASCE), through its “Committee
ween the mean values of load and re of the success of a system in on Eiectrical transmission Structures”
strength is immaterial. accomplishing its task. has prepared a document “Guideiines
N°129 ELECTRA 69
for transmission une structural loading” tion of probabilistic techniques, the or if we represent °T’ the general
which is available as a draft. design engineer will use discrete variable, by x, we can write generally:
Working Group 22.06 has carried values of loads and strengths.
out a comparison of the UK and USA F(x)=[1 —(1/7)) (1)
documents with IEC documents as to 2.2 Loads (Wind effects)
their contents and as to the resuits The knowledge of extreme loads is
obtained by the application. connected with the statistical know Where F(x) is the cumulative distri
They have the following aspects in ledge of meteorological phenomena bution function of Iargest (maximum)
common: giving rise to these load effects (tem extreme values. t is also referred to as
perature, wind, ice...). Fisher Tippet Type 1 (Gumbel) distri
—
i) they accept a distribution of loads bution and is expressed generally as:
(Gumbel type 1), Extreme loads are associated with
ii) they accept that the strength can be “return periods” and the greater the
represented by a distribution (usual return period, the higher the loads. F(x) = e_e_z exp. [—exp. (—z)] (2)
ly Gaussian), The return period, largely due to
iii) the resultant risk is calculated in the Gumbel, is defined as the time interval Where
same way, between phenomena whose intensity z=(C1/G)[x—.~+cY(C2/C1)]
exceeds a selected level. In fact, the
iv) the load applied to a component is = mean value of distribution
IEC documents accept that loading
function of the response of that o= standard deviation
effects are best described by Gumbel’s
component, C1, C2 = Gumbels law coefficients
first law of extremes (see Figure 2).
However, they differ in other V~ = coefficient of variation of variable x
aspects: It can be proved mathematically that
[Note that the distribution of smal
the probability that a level “X”, of an
i) the averaging times for wind speeds /est (minimum) extreme values is:
event X, will be exceeded in any one
are different e.g. year is given by the relationship F(x) = e~e~ = exp. [— exp. (z)] (3)]
IEC : 10 minutes The IEC documents contain several
8S8100:1 hour P(X>x)—(1 / 7) tables for the calculation of the maxi
ASCE : fastest mile mum value of a loading event as func
ii) the reference reliabilities are diffe where Tis the return period. tion of the average value of observed
rent, Corresponding to this return period data, for various levels of reliability. In
ot an event, there is a level QT of loa practice, the availability of a more
iii) there is another important difference general table of coefficients would be
between the objectives of IEC and ding.
useful.
ASCE. IEC proposes to control the In order to obtain the mathematical
reliability of an entire line, whereas From equations (1) and (2) by trans
laws expressing wind speeds (or wind
ASCE limits its goal to the control of loads, or any other type of weather formation, we obtain:
the reliability of structural compo related phenomenon) statistical tech
nents in the line, niques are used and it should be ~ (4)
iv) etc. remembered that a margin of confiden Ci C1~ L “ Tu
ce is bound to be the best estimate.
The conclusions of the Working
Group are that any document could be
For instance, in Canada, when evalua
ting the 50 year annual maximum wind
If2~=(_1n[—1n(1 _.~)])
used since the differences resulting speed, the range for 95% confidence
from their applications are unlikely to was of the order of 18 to 21%. Conver
be critical. However the main point to sely, for 95% confidence the range of (5)
emphasize is that each document is return periods corresponding to a cho ci
coherent in itself and the temptation to sen maximum annual wind speed was
use only parts of each should be resis also quite large.
ted, unless dealing with general
If a line component was subjected =11 Vx (6)
mathematical principles. cl
to a load Q7~ mentioned earlier, and if
this line component had a s~rength
If we denote by XT the value of x
exactly equal to QT (withstand up to corresponding to a return period 7
2. RELIABILITY LEVELS
~T, failure beyond QT), the annual pro then
bability of failure should be (1 /7).
2.1 Background
The IEC documents are based on Gumbel’s law giving the annual pro
bability that a maximum load QTwillbe v~ (7)
the assumption that neither the loads ci
applied to an overhead transmission exceeded, when the mean load is Q
line, nor the strength of its component and the standard deviation is o~ is C1 and C2 are coefficients whose
have single values, and are best des expressed specifically by: values depend on the number (N) of
cribed by distribution functions. Howe available observations. Typical values
ver, it is implicitly accepted, that after (1 /7)=P[Q>QT]=1 F(QT) applicable to overhead line design are
statistical analysis of data and applica i.e. F(QT)=[l—(1 /7)1 given in Table I.
N°129 ELECTRA 71
f (R) F (A)
0.5
AREA=0.9
AREA~0. I
0.1
(10%) R L (10%) R
TABLEAU III. Comparaison d’un vent de 150 ans avec une résistance ayant une ce. it s seen that the failure probability
probabilité de tenue de 90%. in this range is approximateiy equal to
TABLE III. Matching 150 year wind with 90% probabiity of strength being exceeded. (1/27).
in fact similar exercises were car
CDV  cov RAPPORT DES PROBABILITE ried out for different distribution of
CDV CHARGE RESISTANCE VALEURS MOYENNES D’AVARIE
COVLOAD STRENGTH RATIO 0F FA IL URE loads and of strengths and it was
MEAN VALUES PROBABJLI7Y found that the results were comparable
(see Figure 5).
0.300 0.025 2.1036 0.005129
0.300 0.050 2.1757 0.004096 lt is accepted that the combination
0.300 0.075 2.2528 0.003431 of load expressed by Gumbei Law and
0.300 0.100 2.3356 0.003050
0.300 0.125 2.4247 0.002920 of strength expressed as a normal law,
0.300 0.150 2.5208 0.003054 could be considered as the reference
0.300 0.175 2.6250 0.003513 format.
0.300 0.200 2.7381 0.004397
0.300 0.225 2.8613 0.005808 However, good approximations for
0.300 0.250 2.9962 0.007799 the risk of failure can be obtained by
0.300 0.275 3.1445 0.010344 the following methods:
0.300 0.300 3.3082 0.013318
Let Y, ~: mean value of load and
Valeur moyenne de la charge due au vent = 1 CDV = Coefficient de variation
mean value of strength
Meanvalueofwindforce = 1
UQ, o~R: standard deviation of Q and R
TABLEAU IV. Comparaison d’un vent de 500 ans avec une résistance ayant une V0 = (r~j~); VR = (o~/W) : coeff. of
probabilité de tenue de 90%. variation of Q and R
TABLE IV. Matching 500 year wind with 90% probability of strength being exceeded.
a) Format (RQ)
CDVCOV RAPPORTDES
VALEURS MOYENNES
PROBABILITE
D’AVARIE in order to use standard tables for
CDV CHARGE RESISTANCE RATIO 0F FA IL URE
COVLOAD STRENGTH
probability of failures, the assumption
MEAN VALUES PROBABILITY is made that the distribution of the dif
0.300 0.025 2.3951 0.001489 ference of load and strength is normal.
0.300 0.050 2.4771 0.001168 Then the risk of failure can be evalua
0.300 0.075 2.5649 0.000978 ted from the mean values of strength
0.300 0.100 2.6592 0.000888
0.300 0.125 2.7607 0.000894 R and load Q and the standard
0.300 0.150 2.8702 0.001016 deviation cYR and UQ without establi
0.300 0.175 2.9887 0.001317 shing the distribution in detail. Under
0.300 0.200 3.1175 0.001904 this assumption, differences to the
0.300 0.225 3.2578 0.002906
0.300 0.250 3.4114 0.004428 resuits obtained from the reference for
0.300 0.275 3.5802 0.006497 mat wiil be obtained for certain ranges
0.300 0.300 3.7666 0.009037 of standard deviation data.
Valeur moyenne de la charge due au vent =
Mean value of wind force
= 1
P~=ø(fi)
100 * PROBABILITY ANALYSIS
110 Y=50/ * RETURN PERIOD FOR WINO wherej3=
115 P 1—lIT, Z1=—r,0G(—LOG(p)) / Z FOR PROB~..BILITY~P
*
~Y1 YR2 + UQ2
120 F = 1.2816 / * FACTOR ON COV FOR 90% PROBABILITY
130 ;/;° MATCHING”Y” YEAR WIND WITH 90% 0F STRENGTH
BEING ZXCEEDED’ it was found by comparison with the
140 M0=1 / * MEAN 0F WIND FORCE DISTRIBUTION reference format that this approxima
150 C(1)=.12, C(2)=.2, C(3)=.5, ~ / * COV 0F WIND & NO. 0FF
160 ;/;‘ MEAN VALUE 0F NIND FORCE 1” /; tion was quite acceptable for ail values
170 ;‘ CCV COV MEAN FAILURE” of VQ, but only when
180 ;“ LOAD STRENGTH STRENGTH PROBABILITY”
190 l~.#i~# lI.l~## ~t##.it4# VR> 0.15 or 0.2
200 FOR 1=1 TO K/;/ S1=C(I)*t~O I ~ S1=SD FOR EACH WIND CCV
210 M1MO— .45005*51 / * WIND MODE b) Format In (RIO)
220 X1=Z1&S1/1.28255+M1 / & X FOR PROBABILITY P
230 FOR C~.O25 TO .31 STEP .025 / * COV FOR STRENGTH Here again, in order to use standard
240 M2=X1/(1_F*C), S2÷M2&C / MEAN & SD 0F STRENGTH tables, the assumption is made that
250 A1—1.28225/S]. / * CONSTANT
260 D—82/10 / * INTEGRATION INTERVAL
the distribution of the In (R I Q) is nor
270 X=MAX(O, f,12_10*S2), L=M2+5*S2 / * INTEGRATION LII~ITS mal. Differences to the resuits obtained
280 A—0 I * SET& INTEGRAL TO ZERO
290 * STAR’T INTEGRATION LOOP
300 T—(X—M2)/S2, Z=A1*(X_M1) I * NORMALIZED ORDINATES
310 A_A÷(1_EX2(_EXP(_Z)))*EXP(_T*T/2) Court programme sur ordinateur pour cal
320 x—X+O culer la probabilité d’avarie. Charge = loi
330 IF X<L THEN 290
340 ;USING 19O,C(I),C,M2,A&D/SQR(2*3.1415926)/S2
des valeurs extrêmes (Gumbel). Résistan
350 NEXT C
ce = distribution normale (Gauss).
360 NEXT I Short computer program to calculate fallu
370 ;/;/ END re probability. Load = Gumbel. Strength
>
Gaussian.
N° 129 ELECTRA 75
lations is the wind speed. However, for CCV = 0.2 (VT/VSo) 1 1.114 1.234
reliability assessment, it is the load
due to the wind which must be consi (VT”~50>2 1 1.240 1.523
dered in its interaction with the
strength of the loaded component.
Now, the load due to the wind is tivities in wind speeds and wind pres document for the probabilistic design
proportional to the square of wind velo sures. of overhead lines : “Guidelines for
city. t has been found that the distribu It will be seen that for a COV of transmission me structural loading”.
tion of wind loads does not follow the 16%, the selection of reliability level Il Instead of declaring reliability levels,
same law as the distribution of wind or III instead of I corresponds to the the ASCE document introduces the
speeds. selection of wind speeds greater by 10 concept of relative reliabilities by using
to 20% and pressures 20 to 44% a factor G. The following correspon
However, the reliability has been dence would apply:
calculated by accepting loads to follow greater.
Gumbel law. The errors in so doing are The impact on cost of the choice of IEC Level ASCE Values of G
negligible, since only the area under a different reliability level can therefore 1.0
the tails is of importance. Various be estimated. 3.0
attempts at increasing the mathemati Another way to consider Table V is III 10.0
cal accuracy 0f the models have to compare failure probabilities. Thus
shown that the reference format was level III has a yearly probability of failu 2.8 Une reliability versus compo
quite adequate, and had the advanta re of io—~ whereas level I has a yearly nents reliability
ge of easy hand ling. probability of failure of 10—2. This is The assessment of the reliability of
equivalent to stating that level III is a transmission line with respect to cli
2.7 Relative reliabilities 10—2 / iO—~ = 10 times more reliable matic loadings requires the availability
On the basis of the statements than level I. of suitable statistical models describing
made in the previous paragraphs, In fact, in the United States, the load and strength of the line as a
Table VA can be prepared to give rela ASCE has been working on a new whole.
N°129 ELECTRA 77
If load and strength distribution (calculated reliabiuity less than real remembers the randomness in the
functions are derived for single compo value) such as: quality of steel, in the quauity ot manu
nents then the output reliability applies — Assuming that use factor of compo facture, in the design expertise etc.
to this component (e.g. ASCE model); nents (specially the weakest) is Working Group 8 on Towers has this
while if the same functions are derived 100% point on its agenda.
for a whoie line, the output reliability is
Assuming that extreme wind speed Be it as t may, the opinion of the
a line reliability (IECITCII model). —
Load effect < strength or 4.3 Coordination of strengths components subjected to the cri
tical load.
In paragraph 2.5 it was stated that Çbq = factor related to the quality level
‘YLJQT < 4RR0 the GOy of strengths in the range 0.05 of the component during fabrica
to 0.125 were important, because, typi tion and construction.
i.e. load factor x limit load < strength cally this is the range of CCV for = factor related to coordination of
factor x characteristic strength towers (a GOy of 0.075 is fairly repre strength.
Thus, in order to quantify reliability sentative). Although it was stated that
levels, consideration must be given to the reliability of a line was [1 (1/27)],
—
limit values for loads and for strengths. in fact the assumption was already 5.1 Use factor coefficient applied
made, intentionally, that one compo to Ioad
4.1 Limit states for materials nent would be weaker than another. A There has been Iengthy discussions
careful inspection of Figure 5 will show both inside and outside the Working
Until now, most design offices use a that the probability of failure, hence the
single value of strength (or design Groups to decide whether a factor was
reliability, of different components sub needed on the loading side of the
stress) with little awareness 0f the dis jected to the same load is not the
persion of strength 0f the materials. equation.
same. The ambition of a designer is to
Depending on national codes or natio achieve a given reliability for a given Finally the opinion prevailed that,
nal practices, use is made of minimum ength of overhead line. The concept of since aIl towers were not fully and
yield stress, maximum permissible sequential failures achieved by a co equally loaded, it would be more logi
working stresses etc. Generally ordination of strength of the various cal to be able to influence the loading
stresses used for design are ilidefi components attempts to reach this part of the equation.
ned. objective.
The analysis of the performance 0f 5.1.1 Effect of span dispersion on
a line as a system, leads to the accep If it were possible to ensure that the reliability (Beta curves)
tance 0f two limit states for materials probability 0f failure of one component
In most transmission lines, towers
or components is very much less than that 0f another
are not custom designed for each loca
component, the reliability 0f a system
a lower damage limit tion and one or two tangent towers are
—
would be expressed by the reliability of
used to fit in a large array of weight
a higher failure limit the weakest component. (Table VI)
—
and wind spans, most of which are
Practically, it would be possible to On the basis 0f many considera shorter than the maximum design
use either limit value provided that tions, the following sequence has been spans.
adequate coefficients are judiciously accepted, starting from the weakest The fact that many towers throu
selected and applied. The suggestion component. ghout a transmission line are not used
in the IEC documents are reasonable at their maximum design spans has
The above strength coordination
and justifiable, but some time may definitely a positive effect on reliability
can be applied to most lines, but there
elapse before the damage and failure which has not been quantified in ear
can be some situations where different
limits are generally used. lier studies.
criteria could be applied and thus lead
In fact this practice has always been to another sequence of failures. The In order to assess the effect of span
introduced in the Standards of some principles in the IEC documents and in variation on reliability, the first step is
countries, where allowable stresses this commentary are sufficiently gene to find a suitable model to this variable
are used under normal Ioading condi raI for application for such different cri and then use statistical techniques
tions and yield stresses are used for teria. In the following paragraphs it previously described in order to calcu
abnormal loading conditions (allowable shall be assumed that the suspension late its influence on probability of failu
stress <yield stress). tower is the component to fail first. re.
N°129 ELECTRA 81
J
Span variation is best modelled by
FONCTION DE DENSITE DE FREQUENCE
using a non dimensional unit U, called DU TAUX D’UTILISATION DE
use factor, which equals the ratio of LA I’ORTEE DE VENT
the actual span to the maximum span. FREQUENCY DENS!TY FUNCTI0W 0F
W!ND SFAN USE FACTOR
When dealing specifically with wind
or weight spans, we can use the wind L) REAL UNE I3ErA FUNCTION
z
use factor equal to: DATA SIMULATION
w p  ~ 
— wind span w
wind —
max. wind span >
(r/t)=U;t=[U(1_LT)I~u2]_1 tion is
* Fora tuIler treatment 0f the Bela and Gamma functions reference can be made b “Probability, Statistics and Decision for Civil Engineers” by Jack R. Benja
min and C. AIIm Corneil (McGraw Hill Book Company).
N” 129 ELECTRA 83
transmission une is known because of TABLEAU VIII. Paramètres statistiques de la variation de la portée poids.
the following relation: TABLE VIII. Statistical parameters of weight span variation.
Average span = Average wind span =
Average weight span VALEURS DE —
VALUESOF ~
Thus, the average wind use factor
Uwind can be calculated from: Contraintes et terrain Al Bi, A2 Cl, 82 C2
~onstraints and terrain
— Averagowindspan
‘‘wtnd Nombre de types de
max design wind span pylônes d’alignement
Number of tangent
Average span tower o’pes
max design wind span 1 O.85;0.05 0.75;0.1O O.65;0.15 0.50;0.20
QU.cR
— Calculate statistical parameters of 5.1.2.3 Etfect 0f num ber of towers The numerator in the above equa
applied loads Q’ = W2U. This can Ofl lu tion will be disciissed in connection
be done by Montecarlo simulation with the factor çt~ related to the num
When the maximum intensity of a ber of components affected by the criti
or by numerical integration. For climatic load event covers a large
small values of o~ approximate cal load (chapter 5.2.2). The denomi
number of towers (~, the impact of nator can be obtained by the same
methods can be used. use factor dispersion is now changed: methods, either by simulation or by
— Calculate Q50 from Q = i4~2 (using
If strength dispersion is very low or numerical integration. Table IX outlines
Gumbel tables), this is the load
can be neglected, then the most criti some results obtained by MonteCarlo
without the effect of use factor.
cally oaded tower amongst the N simulation.
— Calculate Q’SO from new curve Q’ = towers will be the one having the lar
W2U this is the load including the From Table IX we notice that ~ s
gest span (or use factor). In order to almost constant (it increases very
influence of use factor. find Yu (with N towers), we have to slightly) for different values of N. This
— The use factor coefficient y~ is cal consider in previous calculation a new result can be explained by the fact that
culated from the ratio (Q’501Q50). curve : minNU. the tower having the largest span is
b) Numerical example When N is very large and strength not necessarily the one with the lowest
dispersion is assumed negligible, ru strength : strength dispersion and
Input: W~= 1.0; V~= 0.12 (Gumbel) closeto 1.0. spans are two independent random
U =0.85; o~=0.10(Beta) variables.
The physical explanation of Yu
Calculations: increasing up to 1.0 is simple : if the In order to better illustrate the calcu
w50= ItT(1 + 2.592 V~) maximum intensity of a loading event lations of this table, let us try an
(from para 2.2) covers a large space, then it will likely approximate method to derive yj~,
“see” the tower having the longest when:
Q50= Kx W2sc= 1,72 K
(where K= is a constant) span.
Calculate the probability density As indicated, the above conclusions W’=i.o, vR=lo%;U=O,95,
function of Q’= W2U. can be made only if the strength dis
persion of towers is neglected compa cru= 0.05 (Vu” 0.053)
From the new Q’ curve, we obtain red to that of spans. However in well N= 10
Q’50 = K x 1.48 (value obtained by optimized lines, the use factor disper
numerical integration). sion could be Iow and ii is important to
— (10%) minNR = 0.77 (by the method
establish the joint effect of use factor
later explained in 5. 2. 2).
Thus)~,=l.48I<=0.86 and strength dispersion.
1.72 K — In order to calculate (10%)
Two methods have been used for
minN(RIU), we have to calculate the
this purpose and were found to lead to statistical parameters of R’ = (RIU).
(Note This exercise was repeated for similar results : the first one using the
T= 150, 500 years as well as for V~,,= general relation QU < R and the
0.16 and 0.20. AIl resulting values of second Q < (Alu). We shail detail R’ = (AlU) = (1.0 I 0.95) = 1.053
were found to be in the range of hereafter the second method.
0.86 to 0.89). V~=(~R+ V~~)~2
The strength used in design equa = (0.102 + 0.0532)1/2
e) Approximate method : using the tion is (10%)R. With the influence of = 0.113
method of central moments use factor, this value becomes (10%)
Q’=kl’~x IT=1.02x0.85=0.85 (R/U). With N towers subjected to
maximum load intensity, the design With N = 10, the 10% exclusion limit
~f~Q= (2V~ + V2~= 0.0576 + 0.01 38 strength becomes (10%) minN (RIU). corresponds to 2.31 standard devia
VQ’= 0.267 tions (see chapter 5.2.2) therefore:
Thus the influence of use factor can
Q’50 (T = 50 years) = Q’ with a 2% be quantified by: (10%)minNR’= W’(l —2.31 V’R)
probability. It is equal to 2.054 stan 1.053(1 —2.31 x 0.113)
dard deviations from mean value of yu[(10~) minNRI[(10%) minN(RIU)]
0.78
normal curves.
Q’50 = 0.85 (1 ÷ 2.054 x 0.267) = 1.32
Q’50 (T= 50 years) without the effect TABLEAU IX. Valeurs de ~ en fonction de iT et N pour VA = 10%.
of U: TABLE IX. Values of’y~ as a function of (T and N for VA = 10%.
Q=W2
N
Q=1.0; VQ=2x0.l2=O.24 U,~ 1 2 5 10 20 50 100
The significance 0f this resuit is consider the total number of compo The product øc,øq.Rc which is equi
important : when the maximum intensi nents affected by the same loading valent to (10%) R can be obtained
ty of a ioad “sees” 10 components event, since we obtain virtually the directly from a statistical analysis of
each belonging to a normal density same overail probability of failure. the test results as foilows
function with a coefficient of variation From each test the ratio
of 20% the probability of failure is 5.2.3 Quality level factor
much higher than if this load were to 0f ail the factors affecting the cha test load
see only one component. In order to racteristic strengths of the compo theoretical strength
obtain the same probability of failure in nents, the factor related to quality s
both cases, the normal strength in the probably the most subjective. Whilst
case of 10 components has to be is computed and an estimate of m
most manufacturers and contractors (mean value) or 0m (standard devia
derated by 0.72 (or by 0.92 if VR = ensure a high quality for their products,
0.075) tion) can be made.
there are cases, especially when in an
The same principle applies to insu emergency situation, some compo Thus
lator strings : the mechanical rating of nents/activities cannot be obtained
øc~øq~Rc (10%)R= (m— 1,28 clm) X
an insulator string depends on the from reputable manufacturers! contrac
theoretical strength
number of insulators in the string and tors. In such cases, t s felt that a fac
on the strength dispersion, VR, of insu tor reflecting the quality of ‘services”
lator units. Assuming VR = 5%, a string rendered should affect the strength b2. Foundation strength from theo
of 100 insu lators each rated R0 has to side of the equation. For a proper eva retical formulae
be derated by 0.90 while a string of 7 luation of this factor, statistical data Formulae such as the inverted frus
of the same insulators has to be derat referred to the percentage of rejects tum for predicting uplift strength fall
ed by 0.95. If VR = 15% then the dera would be needed, but, because of the into this category. Many of these for
ting factors 0N becomes respectively emotive aspects of such analysis, both mulae were calibrated and validated
0.67 and 0.84 (difference of 25%) thus CIGRE and IEC Working Groups have by extensive testing. However, experi
highlighting the importance of 0N when agreed some typical values. mental conditions for foundation instal
strength dispersion is high. lation may differ significantiy from
a) Quality level factor for towers conditions during une construction
The question often arises as to
whether the components are in series t was found convenient to use Figu (e.g. difference in compaction, backfill,
composition etc.).
or in.parallel. In the above example the re 2.1 of BS 8100 : Part 1: 1986, from
insulator units of a string are in series, which the following factors could be In this case the product 00R0 can be
but in a multistring insulator set, they deduced. considered equai to the strength given
are in seriesparallel. by the theoretical formulae, while
TABLE Xii. Quality ievel factors related to can be taken from Table XIII.
It is suggested, for the purpose of towers.
design, that Table XI be used for the 5.2.4 Factor for strength coordina
total number of components subjected tion ø~
to the same load intensity, indepen Quality level 0q
dently of series or parallel arrange Transmission line components have
ments, for the following reason —verygood 1.0 different strength dispersion and res
—good 0.92 ponse to loading. In order to limit the
Let us consider 7 components each —average 0.85 consequences of a failure it was found
having a probability of failure of 104. desirable to design the system in such
In fact it is suggested that these a way that the reliability of one compo
a) Series system nent would define the reliability of the
values be used for ail overhead une
components except foundations. system, i.e. the reliability of aIl the
Pt=~ P11=7x10~
b) Quality level factor for founda
i.e. probability of failure = sum of pro tions TABLE XIII. Quality ievel factor for founda
babilities of failure. When dealing with foundations, two tions designed according to theoretical for
mulae related te construction quality.
cases may be considered,
b) Parallel system
I) when data is obtained by testing
Quality level 0q
2) when no testing is performed and
only theoretical considerations defi —verygood 1.0
Pf=1—t(1 —Pg) ne the strength of the foundations. —good 0.90
1 —(1 — 10—4)7 —average 0.80
bi. Tested foundations
=6.998x 10=7x 104
When specific tests are carried out Definifions:
i.e. probability of failure equals the on the line, and if it can be assumed very good compacted backfill and proper control
complement to one of the products of that these tests are representative of of installation
individual reliabilities. the foundations as installed, then the good: compacted backfiii with only casual
control 0f installation
For our purposes we can ignore the effect of 0q can be assumed to be average: compacted backfill but uncontrolled
arrangements of the components and included in the test resuits. installation
N~ 129 ELECTRA 91
TABLEAU XIV. Valeurs de a et de e~ pour que l’avarie du composant R2 ait lieu après the activities (meetings, discussions,
l’avarie du composant R1 avec une probabilité de 90%. papers etc.) of two Working Groups
TABLE XIV. Values of a and ø~ to ensure that component R2 will faf! after component one from CIGRE Study Committee 22
R1 with a 90% probability. and one from IEC Technical Commit
tee 11. A complete collaboration has
“P~i 0.05 0.10 0.20 been maintained, each Working Group
anaiysing criticaily the proposais of the
~ ~ ~ ,øs
other, and such collaboration was
0.05 1.1 0.91 1.15 0.81 1.26 0.63 extended to the activities of the Amen
can Society of Civil Engineers whose
0.10 1.16 0.92 1.20 0.83 1.30 0.66
input into such work must be recogni
0.20 1.36 0.93 1.37 0.86 1.46 0.69 sed.
0.30 1.63 0.93 1.63 0.87 1.70 0.71 When the work was started it was
anticipated that complex answers, with
compiicated tables would be needed
other components wouid be greater the towers. There must be some com for a rationai application of the ioa
than that of the first component to fail. peliing, economic reasons to alter this dingfstrength concept for overhead une
coordination of strength. This docu design.
Attempts at quantifying this extent
of the difference in reliability have led ment provides ail the necessary tools The documents (CIGRE, IEC, BSI,
to specific recommendations. Appen to enabie the designer to evaluate the ASCE) when read carefully, shouid
dix 1 “Derivation of Strength Factors” reiiabiiity of his une if a speciai scena provide any specification writer with ail
gives the reasoning, the details and rio of performance is accepted. the mathematicai toois needed for the
the methods for defining and quanti There is an interesting feature resui preparation of a document based on
fying ratios of strength to achieve a ting from a critical analysis of Table the probabilistic approach.
seiected probability that some compo XIV and this is the almost “Constant”
nents are stronger than others. value which couid be adopted for ø,~
Table XIV gives the values of ~ as applied to foundations in the typical ACKNOWLEDGEMENTS
(ratio of average strengths) and ø~ to range of values. Although this docu
ensure, with a 90% probability of suc ment accepts a 90% probability, there
cess, that component R2 (coefficient of has been a particular, practicai, case Tribute must be paid not oniy to the
variation = V1~) will be stronger than when a probability of 70% was accep members of WG2206 iisted at the begin
component R1 (COV = VR1). ted. ning, but aiso to the members who have
activeiy contributed, aithough they have
Thus, if in a baianced design it is A further calibration would iead now withdrawn from, or were not officiai
desired that the tangent towers (VR1 aimost inevitabiy to the suggestion that members of the Working Group; L. Paris
usualiy 0.075) shouid be weaker than a single value could be adopted for the (Itaiy), C. Manuzio (ltaly), M. Cojan
the foundations (V1~ say 0.2) the dera foundations, such as ø~ = 0.85 (say) (France), JG. Paradinas (Spain), H.B.
ting factor on foundation strength for ail foundation types. The point at White (Canada), P. Juul (Denmark), J.C.
should be 0.9. There is a 90% probabi issue being that the principle of dera Pohiman (United States), T.J. Pioeg
lity of success in such a design prin ting “Rc~” of the foundations with res (Netheriands), H. Dos Santos (Portugal)
ciple. pect to the towers is much more rele and many others.
(Note : Reference to the appendix and vant than the accuracy in the value of
in addition, tribute is due to members
to the figures in the appendix will show this derating factor as function of an of Working Group 08 of IEC TCii under
that, theoretically, a 100% confidence anticipated value of VR for the founda the abie chairmanship of E. Ghannoum
can neyer be achieved regarding tions. (Canada); K.V. Boos (Germany); D. de la
strength coordination). Houssaye (France); G. Neher (Switzer
land); G. Orawski (United States), P.
5.2.4.1 Strength factor Ø~ for foun 6. CONCLUSIONS Paoii (ltaly); J.C. Pohiman (United King
dations dom); P. Pincemaffie (France); B. Rhe
It is generaliy accepted that the An attempt has been made to sum bergen (Netheriands); K. Schjetne (Nor
foundations shouid be stronger than marize, in a few pages, the resuits of way); A.B. Wood (United Kingdom).
N° 129 ELECTRA 93
APPENDIX 1
COORDINATION 0F STRENGTHS — DERIVATION 0F STRENGTH FACTORS
t)
bld O
O~ =ç~O%)R1 (1O%)R2
(arh. scabe)
failure.
Figure A4.
This means ø~ = (1 2.054. V(R2)
—
For example:
If V(R2)
But
= 0.10; o~ = 0.91
EE1~ELZ7~%2 ~E° c~  ~ j // / /_ ~ V~ oO05
(R 21R 1) = a ~ 16V~ / —
= (i—i .28V(Ri)I(i2.054V(R2)) 12— ~
The strength R2 with respect to Ri J3sof = 1.28 or Plsof] = 0.90 as a too small probability (20%, see
etc. Fig.A6)that3failafter2.
The ø~ ratios are calculated (see
Assume that the density function of Table I) Because ø~(3.1) > ø~(2.i) x ø~(3.2)
the strength follows a normal distribu
tion. From mathematical tables the o~(2.i) = 0.86 a(2.i) = 1.37 ø~(3.i) = 0.86 x 0.63 0.54 (This is
—
reliability index for the preferred o~(3.1)=0.81 a(3.i)=i.15 the value to use n the calculations to
sequence of failure (f3sof) can be deri o~(3.2) = 0.63 0(3.2) = 1.26 achieve 90% probabilities in the coor
ved. o~(2. 1) = ø~ between components 2 and 1 dination of strength).
Let ø5(a, b) as coordination of strength NB: Because components have diffe
I3sof = ~ [P(sof)j rent response Ioadings, care has to be
factor between elements a and b.
e.g. for P[sof] = 0.90 : J3sof = 1.28 taken by comparing the actual strength
We see that a(3.1), by combination ot ot components. For example the dead
P(sof) = 0.98; 13(sof) = 2.054, etc.
coordinations (2.1) is equal to 1.37 x weight has a positive effect on the
1.26 = 1.73 > 1.15 which is the value resistance (strength) of a foundation
R1R2
f3sof= (1) of direct coordination (3.2). We cannot with uplift piles but not on the strength
accept this value because in this case, of a tower. Thus care and attention
we’ll obtain: must be paid when applying the
o(R1) = V(R1)~Ri a(3.2)= cx(3,1)/a(2.1)= 1.15/1.37=0.84 strength factors.
o(R2) = V(R2)~R2
~1, ~2 = average component
strengths TABLEAU I / TABLE!
Say a= ~2/ Ri (2)
“\~vR1 0 05 0 10 0 20
ai
~a2t’(R2)2+~4Ri)2 (3) V(R2N~\ O~
Dans es figures A3, A4, A5, A6 et A7: In Figures A3, A4, A5, A6, and A7
=(1 0%)R1 = Ri (1. — 1.28 ~4Ai)) RI corresponds ta S1 — 1
~ (i0%)R2 k2 .(i — 1.28 ~4A2)) S~_~> correspond à Ri
S correspond à A2 R2 corresponds toS~
V1 correspond à V~2 v~ corresponds ta V(R2)
V~j_i> correspond à VAl V~ corresponds to V(R 1)
_~ (11.28 ~4Ri)) (4) P[S1> Spi] correspond à PIR2> Ail P)S~ > S~ — ~] corresponds f0 PIR2> RI]
~a (1—1.28 ~‘(R2))
4. EXAMPLE