Professional Documents
Culture Documents
System Reliability Analysis of Transmission Line Towers: Pergamon 00457949 (94) E0206-H
System Reliability Analysis of Transmission Line Towers: Pergamon 00457949 (94) E0206-H
1994
Elsevier Science Ltd
Pergamon 00457949(94)E0206-H Pnnted in Great Britain.
0045-7949/94s7.w + 0.00
Abstract-Even though methods to estimate system reliability are being studied extensively, the practical
inclusion of system reliability at the level of complex structures has not received much attention. In this
paper, an attempt has been made to implement the system reliability theories and methods with respect
to the transmission line tower using second-order bounding and equivalent linear safety margin
approaches. The paper also discusses important issues related to the non-linear forcedeformation
behaviour of the tower members under buckling failure. The critical safety margins (failable members)
are identified by automatic failure mode generation by matrix formulation with the aid of the member
replacement technique. A complete computer program has been automated to compute the reliability of
transmission towers. Practical towers used to assess the reliability and illustrate the effectiveness of the
methods lead to the adaptation of a system-based reliability analysis procedure.
member using the IEC and ASCE recommended Finally, in many system reliability models the struc-
loads and loading conditions [8]. tural elements are replaced by forces at the failure of
the element. Often this implies that unloading of
2. STRUCTURAL SYSTEM RELIABILITY MODELS
failed elements is not accounted for. In the present
study, ideal plastic (ductile) force-deformation for
At the present stage of probability calculation tension failure members and semi-brittle non-linear
techniques, a formulation in terms of a functional force-deformation for compression failure under
relationship between the load history and the state of buckling are taken into consideration for the tower
the structural system is not practicable for life size elements. The value of the post-buckling reduction
structures. Simplifications are obtained if the system factor (q) depends on the slenderness ratio of the
state at each point in time depends only on the member and suggested values of q range from 0.5, for
present values of the physical variables. Further a ratio of 150, to 0.35, for a ratio of 75 [9] and are
simplifications are, of course, achieved if the re- incorporated in the present analysis while using the
liability problem can be described by a fixed surface member replacement technique.
in load space. Two important types of such models
are the limit state surface of a time-independent Limit state
elastic structural system with respect to initial failure The states of the structure are classified into safe
(first non-elastic behaviour) and the limit state sur- states and failure states. Different criteria for this can
face corresponding to the plastic collapse of a rigid be applied. Failure of a structure is characterized
ideal plastic structural system. according to (1) the type, (2) the extent, and (3) the
However, it is only for reliability models of the cause of the failure. Each of the criteria may be quite
random variable type that a practicable calculation of idealized. The term system reliability models refers
the reliability is presently possible for large reliability usually to models of failure involving several struc-
problems, i.e. problems described by, say, at least a tural elements. Of course, initial failure also has an
100 basic variables. Thus, in order to deal with large important system reliability aspect, as mentioned in
size structures, formulations in terms of random the introduction.
variables are preferred. The limitations on the capa-
bility of present reliability calculation methods intro- Load model
duce a number of idealizations in structural system As compared to element (component) level struc-
reliability formulations. Important idealizations are tural reliability where a worst case loading situation
as follows. can be established for each element, a corresponding
concept is not natural for a complete structural
Mechanical model for tower element system. Different load patterns, directions, etc., must
The structural system is described by a discrete be taken into account and for each failure mode a
model in which local failure may take place only in worst (extreme) loading situation may be set up.
a finite number of points. The structure is assembled
Response
by a number of members, each of which is assumed
to behave according to some load-deformation Only static system reliability analyses have been
model. In Fig. 1, typical loadcleformation behaviour performed as regards to the progressive failure of
for truss elements are shown. In most system re- structures with random properties. As regards initial
liability models the strengths are assumed random, failure, some dynamic analyses for deterministic sys-
whereas the geometry and stiffness are deterministic. tems have been carried out under special conditions,
V 7 FORCE (TENSION)
FORCE ITENSION) v FORCE (TENSION)
k) SEMI-
(a) OENERAL MODEL (b) WCTILE AN0 BRITTLE MODEL BRITTLE MODEL
such as broken wire conditions, to get the critical generally not considered since each leg mem-
responses of the system. ber/segment is assumed pinned between two joints.
Redundant members are not included in the analysis
System model since they have very little effect on the forces in the
The modelling used here is based on the assump- load-carrying members.
tion that the total reliability of the structural system The loadings on the transmission towers are con-
can be sufficiently accurately estimated by consider- sidered probabilistically under an extreme value dis-
ing a finite number of failure modes as dominating tribution of the wind. The load criteria which govern
and then combining them in complex system re- the critical member/component load effects are fully
liability theories. It is assumed that structures consist based on IEC recommendations and ASCE manu-
of a finite number of elements and that these struc- als [3, 5, lo]. Thus the transfer model from wind
tural elements are connected by a finite number of speed to wind load in addition to other loads has been
joints. For each of the structural elements a number automated in the computer program TLTOWER [8].
of different failure modes exist and system failure will
in general only occur when a number of element 4. ELEMENT LEVEL RELIABILITY ANALYSIS
is very large, but only a small fraction of these are can be also expressed by using the safety margins of
likely to occur and a crucial step in the process is to the failed members:
identify the most-likely-to-occur modes and se-
quences. Currently, matrix formation with member Z,(rl,rz ,..., r,_,)<O @=1,2...P,). (7)
replacement technique is available that can efficiently
help to identify these modes. The failure criteria of a The automatic generation of failure modes or
statically indeterminate truss structure are systemati- identification of the most failable members for the
cally generated by using a matrix method [12]. In the tower has been done using the SPACE TRUSS
case of a highly statically redundant structure, failure ANALYST program (STAT-I) [7]. This is a first-
in any one member does not necessarily result in order elastic space truss analysis program incorporat-
structural failure. Structural failure is defined to ing the member replacement technique.
occur when the structure is turned into a mechanism.
A failure mode is generated as in the following 6. SYSTEM LEVEL RELIABILITY OF TRANSMISSION
manner. When any one member fails, redistribution TOWER
of the internal forces arise among the members in The system failure event may be formulated as the
survival and the member next to fail is determined. union and/or intersection of individual events, the
After repeating the similar process, structural failure latter corresponding to the failure of a component or
results when the member upto some specified number a failure of the system. The union of events is used
Pk, e.g., members r,, r2, , and rpk are lost. The in a series system, where the system fails if any of its
formation of a mechanism is determined by investi- component fails or any of the potential failure modes
gating the singularity of the total structural system occurs. On the other hand, the intersection of events
stiffness matrix [K’Pk’]formed with (n - Pk) members is used in a parallel system, where the system fails
in survival. That is, a criterion for structural system only if all its constituent components fail. It is not
failure is given by uncommon for practical structures to exhibit the
behaviour of a combination of series and parallel. In
1[IP’] I = 0, (4) the present investigation, the tower system considered
as a series system and the two approximate methods
where I[. ]I = the determinant of a matrix [ .I. namely the equivalent hyperplane method and second
Now introduce the expression for the safety margins order Ditlevsen’s bounds method have been applied
for the members in survival after some members are to compute system reliability of the transmission
in failure. For instance, when members r,, rz.. , tower. Both the methods are based on a first-order
and r,, have failed, their stiffness matrices are put approach and the only difference is in the compu-
to zero and their post-failure strengths are applied to tations of the probability of a union of events. In the
the nodes as artificial forces, corresponding to the equivalent hyperplane method, an equivalent hyper-
type of failure. When a member of a ductile material plane approximating the system failure surface is
fails in tension, the yield strength of the member obtained while in the Ditlevsen’s bounds method, an
is taken as the residual or post failure strength while upper and a lower bound to the system failure
for a semi-brittle material failing in compression, the probability are obtained. The main features of the
post failure strength is some percentage of the ulti- methods are discussed below.
mate strength of the element. The post-buckling 6.1. Second-order Ditlevsen’s bounds method
reduction factor (r~) used here is as in [9]. Then, a
In this method, the method of second-order
stress analysis of the structure is carried out once
bounds, Ditlevsen’s bounds [13] are used to estimate
again by using the matrix method, and the internal
the system reliability. The bounds for the probability
forces of the members in survival are determined as
of the union of events, say Ai, i = 1, n, can be given
follows:
in terms of the probability of each event and the
pairwise intersections. The expressions for the lower
S,(r,,r2,...,rp)= i &jL,(P)+ C h,R,, (5)
and upper bounds are respectively:
/=I q = i.j
II (8)
members and the sequential order of failure. Conse-
quently. the safety margins are given by + i 0, P(A,) - 2 P(A,fl A,)
,=2 iQi
The depend on the order in which the indices and p is the correlation matrix of the com-
events are considered. It is recommended that the ponent safety margins. Again, we can write from eqn
events be arranged in the order of increasing re- (12) as in [14]:
liability to produce narrower bounds [14]. The pair-
wise intersections of events are calculated using the
bivariate normal cumulative density function (CDF)
using the correlations calculated as:
The following procedure is used in this method:
Pij = {ailTIaj)9 (10)
(i) The direction cosines and the reliability indices
of the most critical failable components are arranged
where {ui) is the vector of direction cosines for the in order of increasing reliability.
failure hyperplane for component i, aI that of com- (ii) Equivalent linear safety margins for the union
ponent j and p,, the correlation between the safety of pairs of most critical failable components are
margins of components i and j.
pfS=P
[i~,(hYW+-Bi~~”
1y (11) I T
Pr,=I--n(&Pk (12)
found. The pairs are: components 1-2, 3-4, 5-6, 7-8, those components are considered ((n - 1)/2 pairs + 1,
etc. The reliability index p’ of the union of the failures if n is odd), and n/2 equivalent linear safety margins
of critical components i and j failure is obtained as are determined. These n/2 hyperplanes can be paired
follows: again and combined with n/4 hyperplanes. The re-
sulting single linear safety margin represents the
B’ = -@-‘fl - @2(BirBj; Pij)l, (14) failure of the tower. The probability of failure of the
tower is then @[-/I’].
where /I, and 8, are the reliability indices of the The main computer program SYSRELIA uses the
components i andj and pij the correlation between the above two methods to perform a system reliability
safety margins of components i and j. analysis. The reliability indices and their direction
(iii) In the initial operation, if there are n critical cosines used in this program are important from the
failable components in the structure, n/2 pairs of output of COMRELIA. The schematic layout of the
program SYSRELIA with its auxillary programs and Insulator string length = 2500 mm
routines comprising the structural system reliability Insulator string weight =12OON
analysis package is presented in Fig. 2. No. of insulator discs = 13.