441

The dynamic analysis of piping systems using

pseudo-dynamic techniques

F J M Q de Melo¤ , J A O Carneiro, H R Lopes, J F Dias Rodrigues and J F Silva Gomes

New Materials and Experimental Methods, INEGI, Institute of Mechanical Engineering and Industrial Management, Porto,
Portugal

Abstract: This paper describes an application of pseudo-dynami c techniques to the dynamic analysis of
piping structures. Essentially it consists in coupling a direct time integration algorithm, such as the
Newmark method, to an experimental step. At each time step the integration algorithm generates a
displacement vector of the structure, which is prescribed for the test specimen. This is mounted in a rigid
test rig fitted with a set of displacement actuators and load cells at the level of the structure degrees of
freedom. The load cells allow the reading of the internal restoring force vector, which is fed back to the
direct time integration algorithm in an actual time step. Further calculations for the velocity and
acceleration vectors will define a new structure configuration by evaluating a new displacement vector
referred to the next time step. This procedure makes it possible to assess experimentally a realistic stress
distribution at sections of complex shape piping parts. The method is a precise tool in dynamic analysis
and, on being carried out in a quasi-static procedure, it operates with less expensive equipment than is
necessary in real dynamic test.

Keywords: dynamic analysis, piping systems, pseudo-dynami c techniques

oped seismic resistant structures after pseudo-dynami c tests

1 INTRODUCTION
The broad application areas for piping systems include in
the project the influence of time-dependent load actions,
following specifications in design codes. This is the case of
seismic forces or those resulting from the influence of
reciprocating machines included in the system. The case of
seismic forces has a leading importance in this type of
structure, as catastrophic effects may arise in the system,
given the high amplitude of the displacements. Damage
mechanisms occur in piping systems frequently at the level
of supports, releasing the structure from necessary restric-
tions and contributing to high-level stress fields at piping
details particularly vulnerable to overloads. This is the case
of pipe elbows or tee connections, where these accessories
undergo a considerable section distortion due to ovalization
under in-plane or out-of-plane bending effects.
Relevant work in the design of seismic proof construc-
tions has involved research techniques involving pseudo-
dynamic methods. Aktan [1], Aktan and Hashish [2], Beck
and Jayakumar [3] and Akao and Watabe [4] have devel-
The MS was received on 21 February 2000 and was accepted after revision
for publication on 19 April 2001.
¤ Correspondin g author: Departamento de Engenharia Mecânica e Gestão
Industrial, Faculdade Engenharia, Universidade do Porto, Rua Dr
Roberto Frias S/N, 4200-465 Porto, Portugal.
S01200 # IMechE 2001
performed on reduced-scale models. These researchers
have also studied the influence of several damping models
on the reduction in the vibration amplitude, having devel-
oped dissipation energy motion supports. In the research
area of buckling mechanisms of axially loaded columns
submitted to transverse seismic actions, contributions have
been made by Miki and co-workers [5–17] at the Daido
Institute of Technology, Nagoya, Japan. For this purpose,
pseudo-dynami c techniques were selected to define the
damage model of axially loaded columns and submitted to
time-variable transverse loads.
Significant contributions to the application of these
techniques to the study of seismic actions on structures
have been made by the European Laboratory for Structural
Assessment (ELSA), part of the Joint Research Centre
(JRC) located at Ispra (VA), Italy. In these research
facilities, pseudo-dynami c tests were carried out with full-
or reduced-scale models assembled in a huge reaction wall
test rig. The simulation of seismic actions has been
performed at ELSA, JRC. In these research facilities the
pseudo-dynami c tests were carried out with full- or
reduced-scale models, assembled in a huge reaction wall
test rig. At the JRC, Pinto and co-workers [18–23] have
presented contributions on the stress analysis and damage
mechanisms in civil constructions.
The application of pseudo-dynami c techniques is still
expanding, meeting increasing interests in the scope of
JOURNAL OF STRAIN ANALYSIS VOL 36 NO 5
442 F J M Q de MELO et al.

general structural design other than civil construction. The linear elastic behaviour or not, assuming that a tangent
development of fully equipped test rigs is a matter of stiffness matrix could be defined between two consecutive
challenge and imagination for designers. In fact, the time steps.
method leaves more scope for the design of test rigs for The iterative method used here deals with the accelera-
specific applications than further modifications in the tion as the iteration variable. Considering that viscous
method itself, which is actually quite efficient in the use damping may be neglected in the majority of dynamic
both of numerical algorithms and data processing methods. problems, as in the application under discussion here, the
This is the motivation to report in this work an incursion of iteration scheme is set up as follows:
pseudo-dynami c methods in the area of piping engineering. ³ ´
¢t 2
[K] ‡ [M] U t‡¢ t
4
³ ´
2 THE NUMERICAL SOLUTION ALGORITHM ¢t 2
ˆ Ft‡¢ t ¡ [K]U t ‡ ¢t [K] U_ t ‡ [K] U t
4
In structural dynamics, the behaviour of a structure is
defined from the dynamic equilibrium equation (2)

[K]U ‡ [C] U_ ‡ [M] U ˆ F(t) (1) At each time step t ‡ ¢t it is possible to define the
structure kinematic state from the previous iteration at
where [K], [C] and [M] are the stiffness, damping and instant t:
mass matrices respectively, while the right-hand side is the
external force vector. ¢t _
U t‡¢ t ˆ U t ‡ ( U t‡¢ t ‡ U_ t ) (3a)
The procedure described in this work refers to a joint 2
contribution between a numerical model involving equation
¢t 
(1) and an experimental step, which leads to the calculation U_ t‡¢ t ˆ U_ t ‡ ( U t‡¢ t ‡ U t ) (3b)
of a force vector called the internal structure force or 2
structural restoring force vector. This force vector corre-
sponds to the effective physical (or mechanical) behaviour Equation (2) can be written as the compact expression
of the structure at the level of the selected degrees of
freedom. This analysis may deal with a linear or non-linear ^ U t‡¢ t ˆ F t‡¢ t
[K] (4)
structural behaviour.
The dynamic behaviour of the structure is numerically where
approached by direct time integration of equation (1). A ³ ´
necessary condition to deal with pseudo-dynamic techni- ¢t 2
[K]
^ ˆ [K] ‡ [M] (5a)
ques demands that an implicit time integration algorithm 4
must be used, which in this case is the Newmark method.
In its usual version, as described for example by Bathe [24] and
or Adina Research and Development Incorporated [25], the
2
primary variable to be iterated is the structure global ¢t
F^t‡¢ t ˆ F t‡¢ t ¡ Rst ¡ ¢t [K] U_ t ¡ [K] U t (5b)
displacement vector, which allows further determination of 4
the velocity and acceleration vectors by numerical deriva-
tion. However, a Newmark version presenting slight where Rst is the internal restoring force vector (experimen-
differences was used in this analysis. Here, an updated tally measured).
structure state identified by the displacement vector The iterative procedure starts once the initial conditions
referred to a time step t ‡ ¢t depends on the evaluation of for displacement and velocity are defined. The acceleration
the structure internal restoring force vector, experimentally at t ˆ 0 may be calculated from U 0 ˆ [M]¡1 F0 . Therefore,
measured in a previous time step t. The analysis may deal the starting conditions for a pseudo-dynami c test need
with linear or non-linear behaviour, as mentioned. numerical definitions for the matrices and vectors included
In order to make it possible to couple the variables in equations (4) and (5). Once equation (5) is solved for the
included in the numerical iteration algorithm with experi- acceleration vector at time step t ‡ ¢t, the velocity and
mentally measured parameters, as described, an efficient displacement vectors are calculated from equations (3a)
version proposed by Liu [26] was used. The version and (3b). Also, it is necessary to solve [K]U t only for the
considers at time step t the experimental evaluation of the first iteration of the method, in order to start running the
structure internal restoring force vector at the level of process, defining the velocity and acceleration vectors for
degrees of freedom. This vector is given by the product the next time step.
[K]U t , where [K] is the global structure stiffness matrix Once the degrees of freedom from the structure dis-
and U t is the displacement vector. The procedure holds for cretization are defined, the displacement actuators and load
JOURNAL OF STRAIN ANALYSIS VOL 36 NO 5 S01200 # IMechE 2001
 THE DYNAMIC ANALYSIS OF PIPING SYSTEMS USING PSEUDO-DYNAMIC TECHNIQUES 443

cells allocated to that level allow the vector Rst to be Table 1 Dimension of structure parts
directly accessed; there is no need for further calculation of Elbow curvature angle (deg) 90
the product Rst ˆ [K]Ut . At any time step, the structure Transverse-sectio n mean radius r (mm) 50
will respond with its own physically real behaviour, Mean curvature radius R (mm) 200
Pipe thickness h (mm) 2 (overall)
whether it is linear elastic or not. Length of vertical and horizontal straight runs, L (mm) 8000
Length of elbow tangent terminations, Lt (mm) 1000

3 SUBSTRUCTURING IN PIPING SYSTEMS

When a structure to be analysed is of considerable size, it is
possible to test only some subsets, using substructuring
procedures. Pseudo-dynamic techniques are an attractive
tool for this purpose. The test specimen, assumed to be
detached from a global structure, now has compatibility
nodes or contact matching surfaces, which are assigned to
internal degrees of freedom. In the experimental steps of
the analysis, these nodes or surfaces are available to receive
the installation of displacement actuators, as previously
described. Furthermore, as only degrees of freedom along
contact surfaces or matching nodes will be selected in this
analysis, the test specimen is equivalent to a structure
where condensation of internal degrees of freedom was
carried out.
In the present analysis, a simple piping structural
arrangement and a test substructure, submitted to in-plane
dynamic forces, were considered, as presented in Fig. 1.
The overall dimensions are given in Table 1 and the
Table 2 Mechanical properties of structure parts
Material Stainless steel, ASTM 204 (overall)
Young’s modulus 200 GPa
Poisson’s ratio 0.3
Specific mass r 7850 kg=m3
mechanical properties in Table 2. A pseudo-dynamic analy-
sis of the substructure including a pipe elbow was
performed and is reported herein.
The test rig used for this purpose is schematically
represented in Fig. 2. It consists of a simple and rigid L-
shaped HEB beam arrangement with fixtures to receive
two pairs of position motors. The position motors consist
essentially of a precision worm gear-type speed reducer,
fitted with a backlash-free power screw. Each pair can be
driven simultaneously, therefore inducing a nodal displace-
ment to the substructure test specimen, or they can be
operated separately, generating an angular nodal rotation.

Fig. 1 Piping structure and substructure for pseudo-dynamic test

S01200 # IMechE 2001 JOURNAL OF STRAIN ANALYSIS VOL 36 NO 5
444 F J M Q de MELO et al.
Fig. 2 Test rig for pseudo-dynamic test in piping arrangements with substructuring
Both superimposed motions will complete the displacement
field, compatible with the kinematics state of the global
structure at each time step. At each position of a clamping
ring pair a load cell is inserted between the clamping rings
and the screw actuator in order to measure the internal
reaction force, prescribing a displacement set as shown in
Fig. 2.
According to the substructuring basis, a macro-element
is defined from the principal structure. The substructure is
subjected to a displacement field at the interface nodes,
generating the respective internal compatibility force field.
This displacement –force system must be in geometric
agreement and equilibrium with the corresponding data of
the remaining structure subsets via the contact nodes or
surfaces, as previously referred to. Figure 2 shows the
definition of internal degrees of freedom and reaction
forces necessary for analysis.
Considering an in-plane dynamic load in the test speci-
men, the substructure compatibility nodes N i and N j
JOURNAL OF STRAIN ANALYSIS VOL 36 NO 5
include a pair of displacements (vertical and horizontal
respectively, referred to a global basis {x0 , y0 }) and a nodal
rotation ö.
The next step consists in defining equation (1), but now
configured for a substructure test specimen. From Fig. 2, it
is considered that three macro-elements may be defined, as
follows:
(a) a horizontal run starting at left-hand side node 4 until
node N j ˆ 3;
(b) the test specimen, or substructure (detail in Fig. 2;
containing nodes N i ˆ 2 and N j ˆ 3);
(c) a vertical run joining node N i ˆ 2 to node 1 (built in to
a rigid fitting or rigid boundary).
Only the substructure will be subjected to the pseudo-
dynamic test. Stiffness and mass matrices for the other
piping sets (the horizontal and vertical runs) have elemental
calculation, available from basic solid mechanics.
The calculation of the stiffness matrix for the test
S01200 # IMechE 2001
 THE DYNAMIC ANALYSIS OF PIPING SYSTEMS USING PSEUDO-DYNAMIC TECHNIQUES 445

specimen needs a more elaborate analysis. This can be elastic field. The stiffness matrix to be defined is arranged
achieved by use of the finite element method, where it is as follows:
possible to use two methods, as in the case of conventional 2 3
finite shell elements or finite ring elements [27–30]. All k 11 k 12 k 13 k 14 k 15 k 16
these numerical tools have presented quite accurate results, 6 k 22 k 23 k 24 k 25 k 26 7
6 7
in spite of the need for more complicated meshes with 6 k 33 k 34 k 35 k 36 7
[K] ˆ 6
6
7
application of conventional shell elements. 6 k 44 k 45 k 46 7
7
In the above procedure, substructuring was considered to 4 (symmetric) k 55 k 56 5
simplify the analysis. As mentioned, substructuring techni- k 66
ques lead to a condensation process for the degrees of (6)
freedom of each structure subset. The stiffness matrix only
contains elements referring to degrees of freedom shared
with other structure subsets. With this procedure, factors k 11 to k 13 assigned to node
According to the degrees of freedom identified in the N j can be obtained. The imposition of the horizontal
pipe detail in Fig. 1, the stiffness matrix has a 6 3 6 displacement ¢l makes it possible to define the first line in
arrangement, once there are three parameters assigned to equation (6) by simply reading the load cells installed in
each node. The experimental calculation of the stiffness the driving screws of each position motor, as described.
matrix was performed according to the definition of each Such read force values must be normalized, after dividing
of the matrix factors. Figure 3 shows schematically how the them by the intensity of the prescribed displacement. The
calculation of the stiffness factors was carried out. first line of equation (6) is as follows:
The strategy to set up the experimental calculation of the
substructure stiffness matrix does not rely on any pre- k 11 ˆ Fhc1 ‡ Fhc2
defined arrangement. For convenience, related to a num-
bering adopted for the position motor location, the arrange- k 12 ˆ ¡(Fvc1 ‡ Fvc2 )
ment started by prescribing a horizontal displacement ¢ l at
(Fhc1 ¡ Fhc2 ) H
node j of the substructure in Fig. 1. At the same time, all k 13 ˆ
the other degrees of freedom were set to zero (this was 2
(7)
achieved by preventing any other translation or rotation at
nodes i or j). The procedure made it possible to define six where
coefficients, as a result of the structural elastic response to
the prescribed displacement ¢l. It is noted here that Fvc1 and Fvc2 ˆ forces read from the vertical load cells
virtually any displacement intensity can be prescribed, Fhc1 and Fhc2 ˆ forces read from the horizontal load
provided that the structure may be distorted within the cells
H ˆ distance between clamping rings applied to the test
specimen (see Figs 2 and 3)

(Data read from load cells are now assumed to be normal-

ized.)
Following the proposed notation for the stiffness matrix
arrangement, k 11 represents the force necessary to move
the degree of freedom ‘1’ horizontally by one unit of
length. An equivalent statement identifies k 12 as the
vertical structure reaction necessary to maintain a null
vertical displacement (v ˆ 0) for node N j . It is possible to
measure this reaction from vertical load cells with forces
Fvc1 and Fvc2, after having inverted the resultant signal.
Also, the reaction bending moment measured at node N j
defines k 13 . From the same prescribed value, other values
are available in the set-up of the first matrix line:

k 14 ˆ ¡k 11

k 15 ˆ ¡k 12

(Fvc1 ¡ Fvc2 ) H
k 16 ˆ
Fig. 3 Prescribed motions for experimental calculation of the 2
substructure stiffness matrix
(8)
S01200 # IMechE 2001 JOURNAL OF STRAIN ANALYSIS VOL 36 NO 5
446 F J M Q de MELO et al.

From equilibrium conditions, other factors on the leading

diagonal in equation (6) are also available:

k 22 ˆ k 55

k 44 ˆ k 11 (9)

The factors k 33 to k 66 (where this last factor is a ‘reflected’

force pair at remote node N i ) result from a prescribed
rotation at node N j , as shown in Fig. 4. Naturally, at this
step, all other degrees of freedom must be restrained:

(Fhc1 ¡ Fhc2 ) H
k 33 ˆ
2
(Fvc1 ¡ Fvc2 ) H
k 36 ˆ (10)
2

More matrix terms can be calculated, however redundant, Fig. 4 Prescribed rotation for determination of the remaining
by considering symmetry in the stiffness matrix. Neverthe- coefficients in the experimental stiffness concerning
less, such verification was carried out, as this action moments and reflected forces
showed that the expected symmetry was not presented very
accurately. Possible reasons for this are structural imperfec-
tions in the rig (lack of absolute rigidity) or some loss of The substructure mass matrix was set up from a lumping
precision in the position motors, which exhibited increasing distribution criterion proposed by Cook et al. [31]. Consid-
backlash in the positioning screw after a few tests under ering the inertial effects assigned to the degrees of freedom
load. Also, an absolutely still position for the pair of in equation (12), the mass matrix for the test specimen is
position motors could not be achieved in the cases of
remote bending moments, for example, and thus symmetry [M] ˆ
in the substructure stiffness matrix was not ensured. 2 3
In order to define an approach to the problem, matrix mLtotal
6 2 7
symmetry was enforced simply by taking the mean value 6 7
6 mLtotal 7
between each pair of symmetric factors: 6 (zeros) 7
6 2 7
6 mL2total 7
fk ij gmeasured ‡ fk ji gmeasured 6 7
Effective k ij ˆ (11) 6 7
6 24 7
2 6 mLtotal 7
6 (zeros) 7
6 2 7
6 mLtotal 7
Considering the parameter numbering in Fig. 1, the 6 7
6 7
following arrangement for the unknowns is assumed: 6 2 7
4 mL2total 5
äTe ˆ f u1 v1 õ1 u2 v2 õ2 u3 v3 õ3 u4 v4 õ4 g 24
(12) (14)

The experimental results for the stiffness matrix for the
test specimen (equivalent to the problem substructure),
having considered the previous approach for symmetry in
off-diagonal terms, is denoted [Ksubest ] and expressed in SI
units (N=m or N m=rad):
where Ltotal refers to the total developed length of the test
specimen (Fig. 1) and m ˆ 2ðrhr is the mass per length
unit.
To complete the problem definition, stiffness matrices
for straight pipe runs were taken from the Euler–Bernoulli

2 3
199 246 42 438 ¡116 990:62 ¡199 246:4 ¡42 438 ¡7875:3
6 42 438 213 084 18 847 ¡42 438 ¡213 084 135 202 7
6 7
6 ¡116 990:62 18 847 79 221:471 116 990:6 ¡18 847 9530:6 7
[Ksubest ] ˆ 6
6 ¡199 246:4
7 (13)
6 42 438 116 990:6 199 246 42 438 78 753 77
4 ¡42 438 ¡213 084 ¡188 47 42 438 213 084 ¡135 202:4 5
¡7875 135 202 9530:6 78 753 ¡135 202:4 101 801:8

JOURNAL OF STRAIN ANALYSIS VOL 36 NO 5 S01200 # IMechE 2001

 THE DYNAMIC ANALYSIS OF PIPING SYSTEMS USING PSEUDO-DYNAMIC TECHNIQUES 447

beam theory, available from structural mechanics text- matrices must be defined in advance. A mixed procedure is
books. In this work, results for stiffness matrix beam carried out with this method, where numerical data
elements from Reddy [32] were considered. Such matrices determine the generation of experimental values that are
referred to straight pipe runs are identified as [Ktvert ] and fed back again in the same iterative algorithm. The process
[Kthoriz ] respectively. ends when the user decides the maximum number of
The global stiffness matrix of the complete structure is iterations or the structure reaches a collapse configuration.
arranged, having assembled the stiffness matrices of each
substructure in banded location, as sketched in Fig. 5. The
assembled stiffness matrix contains superimposed k ij
parameters from adjacent or connected elements. This is
4 TEST PROCEDURE
the case of nodes 2 and 3 referring to parameters
{u2 , v2 , õ2 } and {u3 , v3 , õ3 } respectively. The final size of
A pseudo-dynami c test based on the previous mixed
the global stiffness matrix is a 6 3 6 arrangement, because
procedure is described. The experimental set-up used in the
node 1 has all displacements and rotation prescribed as
work and sketched in Fig. 2 is shown in Fig. 6, where a
zero (built-in node), while node 4 has a harmonically time-
detail shows the arrangement of the strain gauges. The
dependent vertically prescribed displacement, with zero for
strain gauges are of rosette type, allowing complete strain
the remaining unknowns .
definition for the piping accessory analysed here. Figures 6
As previously stated, in algorithm (5) the right-hand side
and 7 show the schematic diagrams of the data acquisition
term [K]U t is
system and the automated driving block to activate the set
8 9 of position motors.
>
> u2 >
> >
> >
> The external force is applied at node 4 (Fig. 1) along the
> v >
µ < 2>
¶> = vertical direction; this force depends on time t as depicted
(global õ2
(15) in Fig. 8. Its variation consists in uniformly increasing as a
stiffness matrix 6 3 6) >
> u3 >
>
>
> >
>
> v3 >
>
: >
;
õ3 time stepˆt

where the product

[Ksubest ]fu2 v2 õ2 ; u3 v3 õ3 gT (15a)

corresponds to the internal substructure force vector or

substructure restoring force vector as mentioned previously.
This force vector is available after data reading from load
cells installed at the level of the substructure degrees of
freedom, as described.
As reported previously, the calculation of the global
stiffness matrix is necessary only at the first time step for
the Newmark method start-up, where all substructure

Fig. 6 A global view of the pseudo-dynamic testing device for

Fig. 5 Band assembling arrangement for element stiffness sub- in-plane bending of a curved pipe with tangents and a
matrix detail of the arrangement of strain gauges
S01200 # IMechE 2001 JOURNAL OF STRAIN ANALYSIS VOL 36 NO 5
448
Fig. 7 Sketch of the position motor driving automated system and close view of a position motor. The picture shows
the cursor enclosure for the position feedback signal (on the left) and a cycle counter encoder (bottom right)
Fig. 8 External force acting vertically on node 4 of the pipework
in Fig. 1
ramp until t ˆ 0:6 s, reaching a top value of 10 N and then
suddenly vanishing after that time. The basic time step
lengths are ¢t ˆ 200 ms. It is noted that the external force
does not represent any standard action; the aim of this
shape function for the external load is only to investigate
the accuracy of the method, once slow and sudden load
changes are involved.
Figure 9 shows vertical and horizontal displacements at
nodes 2 and 3 (notation of Fig. 1) calculated after the
pseudo-dynami c technique and with a totally numerical
Newmark analysis. In this last procedure, the finite
elements in the structure mesh have stiffness matrices
based on the Euler–Bernoulli theory for the long straight
pipe runs, while the substructure of the elbow-tangent parts
has the previous experimentally calculated stiffness matrix.
Figure 10 depicts the longitudinal membrane stress ó xx
at the transverse section at equal distance from elbow ends.
As a consequence of the elbow ovalization in this section,
the curved pipe becomes more flexible than the equivalent
straight element. High stress values may be reached, in
spite of having loaded the accessory with relatively low-
level bending moments, as described by Thomson [33].
JOURNAL OF STRAIN ANALYSIS VOL 36 NO 5
F J M Q de MELO et al.
Graphical results for displacements and rotations in nodes
2 and 3 (Fig. 1) show that there is some lack of agreement
between the maximum values of the pseudo-dynami c and
the total numerical analysis. Nevertheless, the global time
variation of the investigated data matches fairly well. Some
reasons for the detected facts are as follows:
1. The global behaviour of a dynamically stressed structure
was estimated from substructuring techniques. Some
modelling difficulties impairing the procedure accuracy
were presented. This is the case when setting the degrees
of freedom at the level of nodes 2 and 3 in the test
specimen, having considered in the finite element
analysis that the substructure nodes were effectively
taken at equal distance between clamping ring pair (see
Fig. 2).
2. The overall dimensions of the substructure in the test
imposed too coarse a lumped mass matrix; this also
applies to the remaining straight elements, which might
have been assumed to be too long, for simplicity.
3. The clamping rings loading the test specimen really act
as rigid flanges, thereby inducing end effects in the
tangent parts, which was not considered in the numer-
ical model. Millard and Roche [34] presented a detailed
study about the propagation effect from end constraints
at thin-walled pipes, as in the case of rigid flanges.
4. A local buckling close to the clamping rings pair could
happen. It is noted that a reasonably high value for the
displacement at node 3 was calculated (about 5 mm).
5. A false in-plane bending (ideally assumed) could
happen, thereby inducing cumulative errors during the
experimental and numerical data transmission. This lack
of precision in the rig design may also be associated
with friction occurring at the level of spherical bearings
used to drive each clamping ring pair.
S01200 # IMechE 2001
 THE DYNAMIC ANALYSIS OF PIPING SYSTEMS USING PSEUDO-DYNAMIC TECHNIQUES 449

Fig. 9 Vertical and horizontal displacements at nodes 2 and 3

5 CONCLUSIONS
The test results suggest some revision in the rig design,
rather than in the method itself, which has proved to be
quite accurate whenever rigid and simple rigs were used.
This applies to civil engineering constructions, where the
displacement field to be prescribed for the structure in
S01200 # IMechE 2001
analysis is only horizontal, requiring simple design test
rigs, thereby contributing to minimizing global error analy-
sis. The present study has to deal with a more elaborate test
rig, where more error sources may be expected. Neverthe-
less, the application of pseudo-dynamic techniques pre-
sented in this work led to a result that converged and
JOURNAL OF STRAIN ANALYSIS VOL 36 NO 5
450 F J M Q de MELO et al.
Fig. 10 Longitudinal membrane stress at transverse section equidistant from elbow ends; time step t ˆ 0:8 s
(fourth step). Experimental values during the PSD test measured with strain gauges, as in Fig. 5
approached the solution obtained with a totally numerical
dynamic analysis.
ACKNOWLEDGEMENTS
This work was supported by FCT, The Portuguese Founda-
tion for Science and Technology, in a PRAXIS XXI
research programme. The authors also thank Dr Artur V.
Pinto (Reaction Wall Facility of ELSA, JRC, Ispra, Italy)
and Dr Joao G. M. Gonçalves, both working at the Joint
Research Centre (JRC) of the European Communities,
Ispra VA (Italy), for their technical assistance.
REFERENCES
1 Aktan, H. M. Pseudo-dynamic testing of structures. J. Engng
Mechanics, February 1986, 112(2), 183–197.
2 Aktan, H. M. and Hashish, A. A. Pseudo-dynamic testing of
structural components. In Dynamic Response of Structures,
1986, pp. 131–137 (American Society of Civil Engineers,
New York).
3 Beck, J. L. and Jayakumar, P. System identification applied
to pseudo-dynamic test data: a treatment of experimental
errors. In Dynamic Response of Structures, 1986, pp. 505–512
(American Society of Civil Engineers, New York).
4 Akao, Y. and Watabe, M. A proposal of pseudo-dynamic
aseismic design by statistical approach—estimation process of
seismic load and application to basements (in Japanese). J.
Structural Construction Engng (Trans. Archit. Inst. Japan),
1992, 436, 21– 30.
5 Miki, T. and Kotoguchi, H. Experimental studies on collapse
behaviors and deformation capacity of steel beam-to-column
connections. Proc. Structural Engng, Japan Soc. Civ. Engrs,
A, March 1991, 37, 121 –134.
6 Miki, T. Elasto-plastic behaviors of steel cantilever columns
with variable cross-section subjected to horizontally cyclic
JOURNAL OF STRAIN ANALYSIS VOL 36 NO 5
loads. J. Structural Mechanics Earthquake Engng, Japan Soc.
Civ. Engrs, April 1992, 446(I-19), 127–136.
7 Nakai, H., Miki, T. and Hashimoto, Y. On limit state design
method considering shear lag phenomenon of corner parts of
steel rigid frames. J. Structural Mechanics Earthquake Engng,
Japan Soc. Civ. Engrs, October 1992, 455(I-21), 95–104.
8 Miki, T., Kotoguchi, H. and Nishi, K. An experimental study
on elasto-plastic hysteretic behaviors of cantilever columns
with variable cross-section. Proc. Structural Engng, Japan
Soc. Civ. Engrs, A, March 1993, 39, 271– 284.
9 Miki, T. Elasto-plastic collapse mechanism of steel cantilever
columns with variable cross-section subjected to alternating
horizontal load. In Proceedings of the 43rd National Congress
on Theoretical and Applied Mechanics, January 1994, pp.
131– 134.
10 Miki, T. Elasto-plastic dynamic behaviors of steel cantilever
columns with variable cross-section subjected to horizontally
seismic ground motion. J. Structural Mechanics Earthquake
Engng, Japan Soc. Civ. Engrs, October 1994, 501(I-29),
75 –84.
11 Miki, T. Earthquake response analysis on cumulative strain
and hysteretic energy absorption of steel columns with
variable cross-section. In Proceedings of the 9th Earthquake
Engineering Symposium, December 1994, pp. 1639–1644
(Japan Society of Civil Engineers, Tokyo).
12 Miki, T., Kotoguchi, H. and Osada, Y. Inelastic earthquake
response and collapse modes of steel portal frames with
variable cross-section. Proc. Structural Engng, Japan Soc.
Civ. Engrs, A, March 1997, 43, 205 –216.
13 Miki, T. and Nethercot, D. A. Cyclic instability of columns
with variable cross-section due to combination of collapse
mechanisms. In Proceedings of the 5th International Collo-
quium on Stability and Ductility of Steel Structures, Nagoya,
Japan, July 1997, pp. 566–576; also in Stability and Ductility
of Steel Structures (Eds S. Usami and K. Itoh), 1998, pp. 337–
346 (Elsevier, Amsterdam).
14 Miki, T., Kotoguchi, H. and Osada, Y. Inelastic earthquake
response of steel portal frames with shear collapse at center of
beam. In Proceedings of the International Conference on
Behaviour of Steel Structures in Seismic Areas (STESSA’97),
Kyoto, Japan, August 1997, pp. 74 –81.
S01200 # IMechE 2001
 THE DYNAMIC ANALYSIS OF PIPING SYSTEMS USING PSEUDO-DYNAMIC TECHNIQUES
15 Miki, T. Inelastic earthquake response of steel portal frames
with shear collapse in middle of beam. Proc. Structural
Engng, Japan Soc. Civ. Engrs, A, March 1998, 44, 169– 178.
16 Miki, T., Kawada, S. and Iwamura T. Elasto plastic behav-
iour of steel columns with tapered base part subjected to
horizontal repeated load. In Proceedings of the 10th Earth-
quake Engineering Symposium, November 1998, pp. 2151–
2156 (Architectural Institute of Japan, Tokyo).
17 Miki, T. and Yamada, O. Inelastic earthquake response
analysis of steel two-story frames with shear collapse in
middle of beam. Proc. Structural Engng, Japan Soc. Civ.
Engrs, C, March 1999, 45, 227 –234.
18 Pinto, A., Verzeletti, G., Negro, P. and Guedes, J. Cyclic
testing of a squat bridge-pier. Report EUR 16247 EN, Joint
Research Centre, Ispra (VA), Italy, June 1995.
19 Guedes, J. and Pinto, A. Numerical simulation of the PSD
test on bridges. Report I-21020, Joint Research Centre, Ispra
(VA), Italy, June 1995.
20 Arede, A. and Pinto, A. Analysis of a 4-storey R/C building
to be tested in ELSA—the influence of global modelling
parameters. In Proceedings of the Conference on Earthquake
Resistant Construction and Design (ERCAD 94), Berlin, 15 –
17 June 1994.
21 Donea, J., Jones, P. M. and Pinto, A. Technical note I-92-
146; also presented at 2nd Meeting of European Association
of Structural Mechanics Laboratories, Joint Research Centre,
Ispra (VA), Italy, October 1992.
22 Guedes, J., Pegon, P. and Pinto, A. PSD testing of bridges—a
numerical validation study. In Proceedings of the Conference
on Earthquake Resistant Construction and Design (ERCAD
94), Berlin, 15 –17 June 1994.
23 Guedes, J., Pegon, P. and Pinto, A. PSD testing of bridges—a
numerical validation study. In Proceedings of the Conference
S01200 # IMechE 2001
451
on Earthquake Resistant Construction and Design (ERCAD
94), Berlin, 15 –17 June 1994.
24 Bathe, K.-J. Finite Element Procedures, 1992 (Prentice-Hall,
Englewood Cliffs, New Jersey).
25 ADINA R&D Inc. Advanced Dynamic Incremental Non
Linear Analysis, Application Manual, Version ARD 97, 1992
(Adina Research and Development Incorporated, Watertown,
Massachusetts).
26 Liu, H. Pseudo-dynamic tests of low-rise metal structures.
PhD thesis, University Microfilms International, Dissertation
Services, 1990.
27 Otsubo, H. and Watanabe, O. Stress analysis of pipe bends
by ring elements. Trans. ASME, February 1978, 10, 112– 121.
28 Bathe, K.-J. and Almeida, C. A. A simple and effective pipe
elbow element—linear analysis. Trans. ASME, J. Appl. Mech-
anics, 1980, 47, 93 –100.
29 Bathe, K.-J. and Almeida, C. A. A simple and effective pipe
elbow element—interaction effects. Trans. ASME, J. Appl.
Mechanics, 1982, 49, 165 –171.
30 Melo, F. J. M. Q. and Castro, P. M. S. T. The linear elastic
stress analysis of curved pipes under generalized loads using a
reduced integration finite ring element. J. Strain Analysis,
1997, 32(1), 47 –59.
31 Cook, R. D., Malkus, D. S. and Plesha, M. E. Concepts and
Applications of Finite Element Analysis, 3rd edition, 1989
(John Wiley, New York).
32 Reddy, J. N. An Introduction to Finite Element Method, 1993
(McGraw-Hill, New York).
33 Thomson, G. In-plane bending of smooth pipe bends. PhD
thesis, University of Strathclyde, Scotland, 1980.
34 Millard, A. and Roche, R. Elementary solutions for the
propagation of ovalization along straight pipes and elbows.
Int. J. Pressure Vessels Piping, 1984, 16(1), 101–129.
JOURNAL OF STRAIN ANALYSIS VOL 36 NO 5