Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

EARTHQUAKE ENGINEERING & STRUCTURAL DYNAMICS

Earthquake Engng Struct. Dyn. (2016)


Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/eqe.2738

A model of triple friction pendulum bearing for general geometric


and frictional parameters

Apostolos A. Sarlis1,2,‡ and Michael C. Constantinou1,*,†


1
Department of Civil, Structural and Environmental Engineering, University at Buffalo, State University of New York,
Buffalo, NY 14260, USA
2
ExxonMobil, Upstream Research Company, Houston, TX, 77389 USA

SUMMARY
Current models describing the behavior for the triple friction pendulum (TFP) bearing are based on the as-
sumption that the resultant force of the contact pressure acts at the center of each sliding surface.
Accordingly, these models only rely on equilibrium in the horizontal direction to arrive at the equations
describing its behavior. This is sufficient for most practical applications where certain constraints on the
friction coefficient values apply as a direct consequence of equilibrium.
This paper presents a revised model of behavior of the TFP bearing in which no assumptions are
made on the location of the resultant forces at each sliding surface and no constraints on the values
of the coefficient of friction are required, provided that all sliding surfaces are in full contact. To accom-
plish this, the number of degrees of freedom describing the behavior of the bearing is increased to
include the location of the resultant force at each sliding surface and equations of moment equilibrium
are introduced to relate these degrees of freedom to forces. Moreover, the inertia effects of each of
the moving parts of the bearing are accounted for in the derivation of the equations describing its
behavior.
The model explicitly calculates the motion of each of the components of friction pendulum bearings so
that any dependence of the coefficient of friction on the sliding velocity and temperature can be explicitly
accounted for and calculations of heat flux and temperature increase at each sliding surface can be made.
This paper presents (a) the development of the revised TFP bearing model, (b) the analytic solution for
the force–displacement relations of two configurations of the TFP bearing, (c) a model that incorporates in-
ertia effects of the TFP bearing components and other effects that are useful in advanced response history
analysis, and (d) examples of implementation of the features of the presented model. Copyright © 2016 John
Wiley & Sons, Ltd.

Received 16 March 2015; Revised 14 February 2016; Accepted 10 March 2016

KEY WORDS: seismic isolation; spherical sliding bearing; friction pendulum; friction; hysteresis

1. INTRODUCTION

The behavior of the triple friction pendulum (TFP) bearing has been originally described by Fenz and
Constantinou [1–5] and Morgan [6]. Figure 1 shows the geometry of a TFP bearing and its
parameters. Its behavior is characterized by radii R1, R2, R3, and R4 (typically R1 = R4 and R2 = R3);
heights h1, h2, h3, and h4 (typically h1 = h4 and h2 = h3); distances (related to displacement capacities)
d1, d2, d3, and d4 (typically d2 = d3 and d1 = d4); and friction coefficients μ1, μ2, μ3, and μ4(typically μ2 ¼

*Correspondence to: Michael C. Constantinou, Department of Civil, Structural and Environmental Engineering, Univer-
sity at Buffalo, The State University of New York, 132 Ketter Hall, Buffalo, NY 14260, USA.

E-mail: constan1@buffalo.edu

Research engineer, formerly graduate student, University at Buffalo.
§
SUNY Distinguished Professor.

Copyright © 2016 John Wiley & Sons, Ltd.


A. A. SARLIS AND M. C. CONSTANTINOU

Figure 1. Schematic of triple friction pendulum bearing and definition of parameters.

μ3 < μ1 ≤ μ4 ). Figure 2 [1–5] shows the force– displacement relation (consisting of five regimes) of a
TFP bearing subject to the conditions (a)Reff1 = Reff4 ≫ Reff2 = Reff3, (b) μ2 ¼ μ3 < μ1 < μ4 , (c) d1 >
ðμ4  μ1 ÞReff 1 , (d) d 2 > ðμ1  μ2 ÞReff 2 , and (e) d 3 > ðμ4  μ3 ÞReff 3 . In these equations,
Reffi = Ri  hiand d i ¼ di Ri =Reffi , where i = 1, 2, 3, and 4. Note also that in Figure 2, the effective
friction coefficients μi are used instead of the actual friction coefficients μi. The distinction between
μiand μi will be explained later in this paper. In Figure 2, Regime I starts at point I when sliding
initiates on surfaces 2 and 3 and terminates at point II when sliding stops on surface 2 and initiates on
surface 1. Between points II and III (Regime II), sliding occurs on surfaces 1 and 3. At point III,
sliding stops on surface 3 and initiates on surface 4. Between points III and IV, sliding occurs on
surfaces 1 and 4. At point IV, the displacement capacity of surface 1 is consumed and sliding starts on
surface 2. Between points IV and V, sliding occurs on surfaces 2 and 4 until the displacement
capacity of surface 4 is consumed. After point V, sliding continues on surfaces 2 and 3 until the
displacement capacities of surfaces 2 and 3 are simultaneously consumed.
The behavior shown in Figure 2 is derived on the assumption that the resultants of tractions at each
sliding surface act at the center of each sliding surface so that only horizontal equilibrium is needed for
the derivation of the force-displacement relation. In reality when a lateral force is applied at the top of the
bearing, the resultants of surface tractions develop at points away of the center so that equilibrium of
moments is satisfied. In effect, this requires that a force larger than the friction force at the sliding surface
is needed for sliding to initiate. In order to better explain this phenomenon and demonstrate the effect of
moment equilibrium on the behavior of the Triple FP bearing, consider the simple examples shown in

Figure 2. Force–displacement relation of triple friction pendulum bearing with R1 ¼ R4 ; μ2 ¼ μ3 ; R2 ¼


R3 ; h1 ¼ h4 and based on theory of Fenz and Constantinou.

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2016)
DOI: 10.1002/eqe
REVISED TRIPLE FRICTION PENDULUM BEARING MODEL

Figure 3. (a) Rigid block sliding on flat surface and (b) rigid block sliding on spherical surface.

Figure 3. Figure 3(a) shows a rigid block on a flat surface under a vertical load W. Consider a lateral force
F acting at the top of the block. Moment equilibrium requires that the contact forces at the sliding surface
apply at a distance r from the center of the sliding block. Sliding will initiate when the friction force S = μW
is equal to the applied force so that, F = S = μW where μ is the coefficient of friction. In this case the force
needed for sliding to initiate is independent of the location of the contact forces. Consider now the case of
Figure 3(b) where the block rests on a spherical surface. Moment equilibrium requires that the normal
force is inclined at an angle θs. Therefore, a force F equal to μW + Wθs is now needed for sliding to
initiate. Accounting for equilibrium of moments, the lateral force at initiation of motion is:

R
F¼μ W (1)
Rh

Force F in Equation (1) is larger than the friction forceμW. For the specific example of Figure 3(b),
the effective friction coefficient is defined asμ ¼ μR=ðR  hÞ. Note that the values of the coefficient of
friction in the model of Fenz and Constantinou [1–5] are determined from experiments of TFP bearings
(and not flat sliding interfaces) so that the values already contain the spherical surface effect (i.e., the
effective friction coefficient values μi in Figure 2 are directly obtained in experiments). Also, note that
in the free body diagrams of Figure 3, the normal force at the sliding surface is shown equal to the
vertical load W. This result is obtained from vertical equilibrium and in the limit of small angles
assuming thatμθ ≈ 0.
This paper investigates the behavior of the TFP bearing with due account for the effects of
moment equilibrium. This leads to the following new features: (i) no restrictions on the
geometric and frictional properties of the sliding surfaces; (ii) the sliding displacement and
velocities can be explicitly calculated at each surface; (iii) variations of the friction
coefficients due to the instantaneous temperature and velocity effects can be directly accounted
for; and (iv) concave plate misalignments can be included in the analysis. Also as a
consequence of feature (ii) as aforementioned, the TFP bearing model presented in this paper
can be extended to capture uplift behavior because the surface sliding displacements and
velocities are needed as initial conditions for each uplift episode. This extension is described
in [7].
It should be noted that models of the TFP bearing behavior have been also developed by
Becker et al. [8], Ray et al. [9] and Dao et al. [10]. These computational models are suitable
for response history analysis. Invariably, these models are restricted by the same basic
assumption of the Fenz and Constantinou [1–5] and Morgan [6] theories that the force
resultants on each sliding surface act at the center of that surface. However, the Becker et al.
[8] model does not place any restrictions on the frictional and geometric properties as the
other models do. In this paper, this theory will be called the ‘Fenz and Constantinou theory’,
but in reality, the aforementioned models produce the same results, provided that the
constraints of the theory are satisfied.

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2016)
DOI: 10.1002/eqe
A. A. SARLIS AND M. C. CONSTANTINOU

Figure 4. Free body diagrams of components of a triple friction pendulum bearing.

2. GENERAL EQUATIONS

Figure 4 shows free body diagrams of the individual components of a TFP bearing. Figure 4(a)
defines the contact forces, offset angles, and sliding angles, and Figure 4(b) shows the inertia
forces and moments, weights of components, and restrainer forces Fri for each surface i. The
two figures are shown separately for clarity. The free body diagrams of Figure 4(a) are identical
to those of Fenz and Constantinou [1] but with the resultant force at each sliding surface acting
at a point different than the center: the location being identified by the value of the offset angle
θsiof surface i. Specifically, angleθsi is formed by (i) the line extending between the point of
application of the resultant force and the center of curvature of sliding surface i (defined as the
one of larger diameter among the two in contact) and (ii) the line extending between the center
of contact surface i. The definition of the sliding angles is identical to the Fenz and
Constantinou [1] model.
The normal force at each surface, with due account for the weight of the TFP bearing components,
assuming small angles of rotation and μiθi ≈ 0is given by
W1 ¼W þ mTCP g þ mTSP g þ mRS g þ mBSP g
W2 ¼W þ mTCP g þ mTSP g þ mRS g
(2)
W3 ¼W þ mTCP g þ mTSP g
W4 ¼W þ mTCP g

Note that the vertical distance between the vectors of friction forces for surfaces 1 and 2 is
approximately equal (and correct for the small rotation theory utilized in the model) to h1-h2; for
surfaces 2 and 3, it is h2 + h3; and for surfaces 3 and 4, it is h4-h2. Also, mBSP, mRS, mTSP, mTCP are
the masses of the parts BSP, RS, TSP, and TCP, respectively, and g is the acceleration of gravity.
The subscripts denote the following: TCP stands for top concave plate, TSP is top slide plate, RS is
rigid slider, and BSP means bottom slide plate (see Figure 1 for definition).
In Figure 4(b), quantities FIBSP, FIRS, FITSP, and FITCP are the inertia forces of the TFP bearing
individual components. These forces act at the center of mass of each body, which is located at
distance z1 from contact surface 1 for the BSP, z4 from contact surface 4 for the TSP, and z2 from
contact surface 2 for the RS. The inertia forces are assumed to act in the tangential direction of the

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2016)
DOI: 10.1002/eqe
REVISED TRIPLE FRICTION PENDULUM BEARING MODEL

part’s trajectory, whereas the radial components of the inertia forces are ignored (Equation (2)).
Quantities MIBSP, MIRS, MITSP, and MITCP are the inertia moments acting at the center of mass of
each of parts BSP, RS, TSP, and TCP, respectively. Moreover, quantities IBSP, IRS, ITSP, and ITCP
are the mass moments of inertia of each of the parts. These quantities are related to the inertia
moments by M BSP ¼ I BSP θ€ BSP ; M RS ¼ I RS θ€RS ; M TSP ¼ I TSP θ€TSP ; M TCP ¼ I TCP θ€TCP . Quantities Fr1,
Fr2, Fr3, Fr4are, respectively, forces that result from contact between the moving parts and the
displacement restrainers of the bearing (hence, called restrainer forces).
The rotational acceleration of each body is calculated by double differentiation of rotations θ1 for
part BSP, θ1 + θ2 for part RS, θ1 + θ2  θ3for part TSP, and θ1 + θ2  θ3  θ4 for part TCP. When the
top plate is horizontal, quantityθ1 + θ2  θ3  θ4is zero, and the moment equilibrium equation for the
top concave plate can be substituted with the constraint equation θ1 + θ2  θ3  θ4 = 0 together with
MITCP = 0.
The equations of motion of the TFP bearing are derived for one-directional motion on the basis
of the free body diagrams of Figure 4. The equilibrium and kinematic equations are presented in
the set of Equation (4). For simplicity, the moment equilibrium equation of the TCP is
substituted by the constraint equation (4h). Equations (4a), (4b), (4c) and (4g) are the horizontal
equilibrium equations of parts BSP, TSP, TCP, and RS, respectively, and Equations (4d), (4e),
and (4f) are the moment equilibrium equations of parts RS, BSP, and TSP, respectively.
Furthermore, Equation (4i) relates the lateral displacement u to the rotations. Note that the
moment equilibrium equations are expressed with respect to the point where the resultant traction
force acts at a sliding interface (for example, for part RS is the point where forces W3 and S3
act (Figure 4 and Equation 4d)). It is noted that these equations are based on the assumption of
small angles of rotation. Also, in these equations, ügis the ground acceleration and quantity Reffi
is the effective radius of surface i given by

Reffi ¼ Ri  hi ; i ¼ 1; 2; 3; 4 (3)

Note that Equation (4) is presented in a specific sequence that facilitates partitioning of the matrix
form of the equations when the more general dynamic analysis model is presented in Section 4 of
this paper.

ðaÞ: W 2 ðθ1 þ θ2 þ θs2 Þ þ S2  S1 þ F r2  F r1  W 1 ðθ1 þ θs1 Þ  mBSP ðR1  z1 Þθ€ 1  mBSP u€g ¼ 0
 
ðbÞ: W 3 ðθ1 þ θ2 þ θs3 Þ þ S3 S4 þ F r3  F r4  W 4 ðθ1 þ θ2  θ3 þ θs4 Þ þ mTSP Reff 1  h4 þ z4 θ€1 þ
   
þ mTSP Reff 2  h4 þ z4 θ€ 2 þ mTSP Reff 3 þ h4  z4 θ€ 3 þ mTSP u€g ¼ 0
ðcÞ: F  W 4 ðθ1 þ θ2  θ3 þ θs4 Þ  S4  F r4  mTCP u€  mTCP u€g ¼ 0
ðdÞ: W 2 ðθs2 R2 þ θs3 R3 Þ  W 2 θs2 ðh2 þ h3 Þ  S2 ðh2 þ h3 Þ  F r2 ðh2 þ h3 Þ
   
 I RS þ mRS Reff 1 þ h2  z2 z3 θ€ 1  ðI RS þ mRS ðR2  z2 Þz3 Þθ€ 2
 mRS z3 u€g  mRS gðθ1 z3 þ θ2 z3 þ θs3 R3 Þ ¼ 0
(4)
ðeÞ: W 2 ðθs1 R1  θ2 R2  θs2 R2 Þ  ðh1  h2 Þ½W 2 ðθs2 þ θ2 Þ þ S2 þ F r2   I BSP θ€ 1 þ
 
mBSP ðR1  z1 Þθ€ 1  mBSP u€g z1 þ mBSP gðθs1 R1 þ θ1 z1 Þ ¼ 0
ðf Þ: W 3 ðθs4 R4  θs3 R3  θ3 R3 Þ  ðh4  h3 Þ½W 3 ðθs3 þ θ3 Þ þ S3 þ F r3 
 
 mTSP gðθs4 R4 þ ðθ1 þ θ2  θ3 Þz4 Þ  I TSP θ€ 1 þ θ€ 2  θ€ 3
      
 mTSP z4 Reff 1  h4 þ z4 θ€ 1 þ Reff 2  h4 þ z4 θ€ 2 þ Reff 3 þ h4  z4 θ€ 3 þ u€g ¼ 0
ðgÞ: W 2 ðθ1 þ θ2 þ θs2 Þ þ S2 þ F r2  W 3 ðθ1 þ θ2 þ θs3 Þ  S3  F r3 þ
 
mRS Reff 1 þ h2  z2 θ€ 1 þ mRS ðR2  z2 Þθ€ 2 þ mRS u€g ¼ 0
ðhÞ: θ1 þ θ2  θ3  θ4 ¼ 0
   
ðiÞ: u ¼ θ1 R1  ðh1 þ h4 Þθ1 þ Reff 2  h4 θ2 þ Reff 3 þ h4 θ3 þ R4 θ4

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2016)
DOI: 10.1002/eqe
A. A. SARLIS AND M. C. CONSTANTINOU

Table I. Triple friction pendulum configurations analyzed.


Configuration Constraints
A μ2R2/Reff2 < μ3R3/Reff3 < μ1R1/Reff1 < μ4R4/Reff4

B μ1R1/Reff1 < μ4R4/Reff4 < μ2R2/Reff2 < μ3R3/Reff3

Table II. Known and unknown variables of triple friction pendulum bearing of configuration A.
Transition point Known variables Unknown variables
I S2 = μ2W, S3 = μ3W, θ1 = θ2 = θ3 = θ4 = 0 F, u, θs1, θs2, θs3, θs4, S1, S4
II S1 = μ1W, S2 = μ2W, S3 = μ3W, θ1 = θ4 = 0 F, u, θs1, θs2, θs3, θs4, S4, θ2, θ3
III S4 = μ4W, S1 = μ1W, S2 = μ2W, S3 = μ3W, θ4 = 0 F, u, θs1, θs2, θs3, θs4, θ1, θ2, θ3
IV S4 = μ4W, S1 = μ1W, S2 = μ2W, S3 = μ3W, θ1 = d1/R1 F, u, θs1, θs2, θs3, θs4, θ2, θ3, θ4
V S4 ¼ μ4 W; S1 ¼μ1 W; S2 ¼μ2 W; S3 ¼μ3 W; θ1 ¼ Rd11 ; θ4 ¼ Rd44 F, u, θs1, θs2, θs3, θs4, Fr1, θ2, θ3

3. ALGEBRAIC SOLUTION

Using Equation (4), algebraic formulations for any combination of properties of the TFP bearing can be
derived when the weight and the inertia force effects of the moving parts are ignored. However, this is
cumbersome, and a different algebraic solution would be needed for each set of parameters of the
isolator. In this paper, two sets of parameters, as shown in Table I, are selected and used to arrive at
algebraic expressions for the force–displacement relation of the bearing. Note the difference in the
constraints of Table I, which involve products of the coefficient of friction and the related radii of
curvature, to those in the theory of Fenz and Constantinou [1–5] that only involve the coefficient of
friction of the four sliding surfaces. As explained earlier, the current formulation utilizes the actual
coefficient of friction at each sliding interface. The model of Fenz and Constantinou [1–5] utilizes
effective coefficients of friction (μi ) obtained in the testing of the bearings. The effective coefficient of
friction is larger than the actual coefficient of friction (μi) because of the requirement for equilibrium
of moments. The two values are related (see following presentation). It should be noted that it is the
effective coefficient of friction that determines the transition between the sliding regimes of the bearing,
which due to the spherical geometry and the requirement for equilibrium of moments, is an effective
coefficient of friction larger than the actual one. It is the effective friction coefficient that determines the
conditions of initiation of sliding and transition between regimes.
Equation (4) is solved for each transition point for the known and unknown quantities shown in
Table II. The weight and the inertia force effects of the moving parts are ignored so that Equation
(2) yields W1 = W2 = W3 = W4 = Wand all inertia forces and weights are zero. The restrainer forces Fri
of each surface are zero when |θι| < di/Riand are nonzero when|θι| = di/Ri.
Figure 5 presents the force–displacement relation of the bearing of configuration A. Expressions for
forces and displacements at each transition point are shown in the figure (details of the derivation may
be found in [7]). By comparing Figure 5 with Figure 2, the following transformation in the friction
coefficients produces identical results for the current theory (coefficient of frictionμi) and the theory
of Fenz and Constantinou (effective coefficient of friction μi ) for the special case of μ2 = μ3:

R2 μ R1  μ2 R2 μ R4  μ2 R2
μ2 ¼ μ3 ¼ μ2 ;μ ¼ 1 ;μ ¼ 4 (5)
Reff 2 1 Reff 1  Reff 2 4 Reff 1  Reff 2

That is, the current theory for the special case of μ2 = μ3predicts the same behavior for all regimes as
the one predicted by the Fenz and Constantinou model [1–5]. The offset angles obtained for each
transition point and for each surface are given by Equations (6) to (9).

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2016)
DOI: 10.1002/eqe
REVISED TRIPLE FRICTION PENDULUM BEARING MODEL

 
θIs2 ¼ ½2μ2 h2  ðμ2  μ3 ÞR3 = Reff 2 þ Reff 3
(6)
θVs2 ¼ θ IV
s2 ¼ θ s2 ¼ θ s2 ¼ θ s2
III II I
 
θIs3 ¼ ½2μ3 h2 þ ðμ2  μ3 ÞR3 = Reff 2 þ Reff 3
(7)
θVs3 ¼ θ IV
s3 ¼ θ s3 ¼ θ s3 ¼ θ s3
III II I
        
θIs1 ¼ μ3 R3 Reff 2 þ h1  μ2 R2 Reff 3  h1 = Reff 2 þ Reff 3 R1
     
θIIs1 ¼ μ1 Reff 2 þ h1  μ2 R2 = Reff 1  Reff 2 (8)
θIV
s1 ¼ θIII
s1 ¼ θIIs1
        
θIs4 ¼ μ2 R2 Reff 3 þ h4  μ3 R3 Reff 2  h4 = Reff 2 þ Reff 3 R4
        
θIIs4 ¼ ðμ1 R1  μ2 R2 Þ Reff 3 þ h4  μ3 R3 Reff 1  Reff 2 = Reff 1  Reff 2 R4
  (9)
θIII
s4 ¼ ½μ4 ðR3  h2 þ h4 Þ  μ3 R3 = Reff 4  Reff 3
θVs4 ¼ θIV
s4 ¼ θ s4
III

Figure 5. Force–displacement of triple friction pendulum bearing of configuration A with R1 = R4, μ2 = μ3,
R2 = R3, h1 = h4 and based on current theory.

Note that the offset angle of a surface changes only if sliding has not initiated or motion was forced
to stop by a restrainer on that particular surface. As a result, the stiffness of the regimes predicted by the
current theory and the Fenz and Constantinou theory [1–5] is the same.
The sliding angles calculated for each transition point and each surface are given by Equations (10)–(13):

θ I1 ¼ θII1 ¼ 0
   
μ4 Reff 1  Reff 2  μ1 R1 þ μ2 R2 μ4 Reff 3 þ h4  μ3 R3
θ1 ¼
III
þ (10)
Reff 1  Reff 2 Reff 4  Reff 3
θV1 ¼ θIV
1 ¼ d 1 =R1

θI2 ¼ 0
 
μ1 R1  μ3 Reff 1  Reff 2  μ2 R2 ðμ3  μ2 ÞR2  2h2 μ3
θ2 ¼
II
þ
Reff 1  Reff 2 Reff 2 þ Reff 3 (11)
θIV ¼ θIII
¼ θII2
2 2
   
θV2 ¼ ðμ2 R2 þ μ3 R3 Þ= Reff 2 þ Reff 3 þ ðμ4 R4  μ3 R3 Þ= Reff 4  Reff 3  d1 =R1 þ d4 =R4

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2016)
DOI: 10.1002/eqe
A. A. SARLIS AND M. C. CONSTANTINOU

θI3 ¼ 0
     
θII3 ¼ μ1 R1  μ2 Reff 1  μ2 h2 = Reff 1  Reff 2  ½2h2 μ2 þ ðμ3  μ2 ÞR3 = Reff 2 þ Reff 3
    (12)
θIII
3 ¼ ðμ4 R4  μ3 R3 Þ= Reff 4  Reff 3  ðμ2 R2 þ μ3 R3 Þ= Reff 2 þ Reff 3
θV3 ¼ θIV
3 ¼ θ3
III

θI4 ¼ θII4 ¼ θIII


4 ¼0
   
μ1 R1  μ4 Reff 1  Reff 2  μ2 R2 μ3 R3  μ4 h4 þ Reff 3 d1
θ4 ¼
IV
þ þ (13)
Reff 1  Reff 2 Reff 4  Reff 3 R1
θV4 ¼ d4 =R4

For the bearing of configuration B (Table I), the known and unknown variables for the solution at
each transition point are presented in Table III. Solution for the force–displacement relation requires
a different approach than for configuration A and is detailed in [7]. The force–displacement relation,
together with expressions for forces and displacements at each transition point and also
assumingd1 < d4, is shown in Figure 6.
The theory of Fenz and Constantinou [1–5] cannot correctly model the behavior of configuration B,
except for the special caseμ1 = μ4 < μ2 = μ3, d1 = d4, d2 = d3. For this special subcase of configuration B,
the force–displacement relation in Figure 6 becomes simpler so that the displacements at the transition
points II, III, IV, and V are equal (uII = uIII = uIV = uV). Then the two theories produce identical results,
provided that the following transformation is made:

R2 R3 R1 R4
μ2 ¼ μ3 ¼ μ 2 ¼ μ3 ; μ1 ¼ μ4 ¼ μ1 ¼ μ4 (14)
Reff 2 Reff 3 Reff 1 Reff 4

Results for the offset and sliding angles for this case are presented in [7].

4. DYNAMIC ANALYSIS MODEL

This section presents a more general formulation that can be used for response history analysis and that
requires the solution of systems of differential equations to describe the constitutive relation of the
bearing. The model is valid for any set of parameters and can capture more detailed aspects of the
TFP bearing behavior. For the presentation in this paper, a horizontal top and bottom concave plate
were assumed; details on how the theory changes when the plates rotate (either because of
misalignment when they have a constant value or because of the flexibility of the substructure and
superstructure when they are functions of time) are given in [7].
Solution of the set of combined differential and algebraic Equation (4) is complex so that the
two kinematic equations given by (4h) and (4i) are used to eliminate sliding angles θ3 and θ4.
Specifically,

     
1 Reff 1 þ Reff 3 Reff 2 þ Reff 3
θ4 ¼ u θ1  θ2
Reff 4  Reff 3 Reff 4  Reff 3 Reff 4  Reff 3
(15)
Reff 1 þ Reff 4 Reff 2 þ Reff 4 u
θ3 ¼ θ1 þ θ2 
Reff 4  Reff 3 Reff 4  Reff 3 Reff 4  Reff 3

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2016)
DOI: 10.1002/eqe
REVISED TRIPLE FRICTION PENDULUM BEARING MODEL

Table III. Known and unknown variables of triple friction pendulum bearing of configuration B.
Transition point Known variables Unknown variables
I S1 = μ1W, S4 = μ4W, θ1 = θ2 = θ3 = θ4 = 0 F, u, θs1, θs2, θs3, θs4, S2, S3
II S1 = μ1W, S4 = μ4W, θ1 = d1/R1 F, u, θs1, θs2, θs3, θs4, S2, S3, θ4
III S1 = μ1W, S4 = μ4W, S2 = μ2W, θ1 = d1/R1 F, u, θs1, θs2, θs3, θs4, Fr1, S3, θ4
IV S1 = μ1W, S4 = μ4W, S2 = μ2W, θ1 = d1/R1 , θ4 = d4/R4 F, u, θs1, θs2, θs3, θs4, Fr1, S3
S1 ¼ μ1 W; S4 ¼ μ4 W; S2 ¼μ2 W; S3 ¼μ3 W
V F, u, θs1, θs2, θs3, θs4, Fr1, Fr4
θ1 ¼ d1 =R1 ; θ4 ¼ d4 =R4

Figure 6. Force–displacement of triple friction pendulum bearing of configuration B with R1 = R4,


R2 = R3, μ1 < μ4 < μ2 < μ3, h1 = h4, d2 = d3, d1 < d4and based on current theory.

Substitution of Equation (15) into Equation (4a–g) gives the following:

ðaÞ: ðW 2  W 1 Þθ1 þ W 2 θ2  W 1 θs1 þ W 2 θs2 þ S2  S1 þ F r2  F r1  mBSP ðR1  z1 Þθ€ 1  mBSP u€g ¼ 0


W4
ðbÞ: ½W 3 þ W 4 C 4 θ1 þ ½W 3 þ W 4 C 5 θ2  u þ W 3 θs3  W 4 θs4 þ S3  S4
Reff 4  Reff 3

  C 0  z1
þ F r3  F r4 þ mTSP ðR4  z1 Þ C 4 θ€ 1 þ C 5 θ€ 2  mTSP u€ þ mTSP u€g ¼ 0
C1
 
1
ðcÞ: F  W 4 u  C 4 θ1  C5 θ2 þ θs4  S4  F r4  mTCP u€  mTCP u€g ¼ 0
C1
ðdÞ:  mRS gh2 ðθ1 þ θ2 Þ þ ðR2  2h2 ÞW 2 θs2 þ R3 ðW 2  mRS gÞθs3  2S2 h2
   
 2F r2 h2 þ I RS þ mRS Reff 1 h2 θ€ 1  I RS þ mRS Reff 2 h2 θ€ 2  mRS h2 u€g ¼ 0
   
ðeÞ: mBSP gz1 θ1  W 2 Reff 2 þ h1 θ2 þ ðW 2 þ mBSP gÞR1 θs1  W 2 Reff 2 þ h1 θs2
 ðh1  h2 ÞðS2 þ F r2 Þ þ ðmBSP z1 ðR1  z1 Þ  I BSP Þθ€1  mBSP u€g z1 ¼ 0 (16)
W 3 C 0  mTSP gz4
ðf Þ: ðmTSP gz4 C 4  W 3 C0 C 2 Þθ1 þ ðmTSP gz4 C 5  W 3 C 0 C 3 Þθ2 þ u
C1
 W 3 C 0 θs3  ðh4  h2 ÞðS3 þ F r3 Þ þ ðW 3  mTSP gÞR4 θs4  mTSP z4 u€g
 
  mTSP z1 ðC 0  z4 Þ I TSP
þ ½I TSP  mTSP z1 ðR4  z4 Þ C 4 θ€1 þ C5 θ€2 þ  u€ ¼ 0
C1 C1
ðgÞ: ðW 2  W 3 Þθ1 þ ðW 2  W 3 Þθ2 þ W 2 θs2  W 3 θs3 þ S2  S3 þ F r2  F r3 þ
þ mRS Reff 1 θ€1 þ mRS Reff 2 θ€2 þ mRS u€g ¼ 0
C 0 ¼ Reff 3 þ h4
Reff 1 þ Reff 4 Reff 2 þ Reff 4 Reff 1 þ Reff 3 Reff 2 þ Reff 3
C 1 ¼ Reff 4  Reff 3 ; C 2 ¼ ; C3 ¼ ; C4 ¼ ; C5 ¼
Reff 4  Reff 3 Reff 4  Reff 3 Reff 4  Reff 3 Reff 4  Reff 3

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2016)
DOI: 10.1002/eqe
A. A. SARLIS AND M. C. CONSTANTINOU

Equation (15) can be written in matrix form as


Mθ€ t þKθt þSþFg þFe ¼0 (17)

In Equation (17), θt is a vector containing the sliding and offset angles and the displacement u of top
the bearing, S is a vector containing terms that involve the friction forces, Fg is a vector containing
terms that involve the restrainer forces, and Fe is a vector containing the excitation (terms involving
the ground acceleration history üg). Vector θt is given in the following together with sub-vectors θs
(offset angle vector) and θ arising from partitioning of vector θt:
( )
T
θ
θt ¼ b θ1 θ2 u θs1 θs2 θs3 θs4 c ¼
θs (18)
θ ¼ b θ1 θ2 u cT ; θs ¼ b θs1 θs2 θs3 θs4 cT

Equation (17) can then be partitioned as follows:


 ( ) 




Maa Mab θ€ Kaa Kab θ Sa Fga Fea
þ þ þ þ ¼0 (19)
Mba Mbb θ€ s Kba Kbb θs Sb Fgb Feb
Note that matrices Mab, Mbb are nil so that Equation (19) can be written as
Maa θ€ þ Kaa θ þ Kab θs þ Sa þ Fga þ Fea ¼ 0
(20)
Mba θ€ þ Kba θ þ Kbb θs þ Sb þ Fgb þ Feb ¼ 0

The second equation in (20) can be solved with respect to the offset angle vector:
 
θs ¼ Kbb1 Feb  Fgb  Sb  Kba θ  Mba θ€ (21)

Substituting Equation (21) in the first equation in (20), the condensed equations of motion are
obtained:

M e þe
e θ€ þ Kθ eg þ F
SþF ee ¼ 0 (22)

In Equation (22), the condensed vector and matrices are given by


e ¼ Maa  Kab Kbb 1 Mba ; K
M e ¼ Kaa  Kab Kbb 1 Kba
(23)
e eg ¼ Fga  Kab Kbb 1 Fgb ;
S ¼ Sa  Kab Kbb 1 Sb ; F ee ¼ Fea  Kab Kbb 1 Feb
F

The friction force Si at sliding surface i may be modeled using the following formulation, which is
based on a modification of the Bouc–Wen model and originally implemented in program 3D-BASIS
[11] (with parameters A = 1, β = 0, γ = 1, and η = 2):
(
  1; θ_ i Z i > 0
Si ¼ μι W i Z i ; Z_ i ¼ ðRi =Y Þ 1  ai Z i θ_ i ; ai ¼
2
(24)
0; θ_ i Z i ≤ 0

Note that parameter Y (typically with a value less than 1 mm) is a ‘yield displacement’ for use in the
visco-plastic representation of the friction force.
 T
The equations may be written in state-space after defining vector Q ¼ θ θ_ Z , where Z is a
vector containing parameters Zi (Equation (24)). Equations (22) and (24) take the form:
8 9 8 θ_
9
< θ= > < >=
dQ d e 1 Kθe þe eg þ Fee
¼ θ_ ¼ M SþF (25)
dt dt : ; > : >
;
Z Z_

Note that the formulation is based on the assumption that a force is applied at the top of the bearing
and a ground motion is applied at the bottom of the bearing. When the element is used in the dynamic

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2016)
DOI: 10.1002/eqe
REVISED TRIPLE FRICTION PENDULUM BEARING MODEL

analysis of a seismically isolated structure, force F = 0(in Equation (16)) and Equation (25) need to be
combined with the equations of motion of the superstructure. In a case when the element is used for the
analysis of a single bearing under the action of a known lateral force F, the ground acceleration ügis set
equal to zero.
Another interesting case is when the bearing is subjected to a known history of vertical load and lateral
displacement u of the top of the bearing with respect to its bottom. In this case, Equations (16), (17), and
(19–25) are valid but for the partitioning of vector θt, which is now given by Equation (26) instead of
Equation (18):
( )
T
θ
θt ¼ b θ1 θ2 F θs1 θs2 θs3 θs4 c ¼
θs (26)
θ ¼ b θ1 θ2 cT ; θs ¼ b F θs1 θs2 θs3 θs4 cT

Note that in Equation (25), all terms involving the displacement u and its derivatives are now known
and are included in the excitation vector Fe.
The coefficient of friction of surface i can be considered velocity-dependent and to follow the
relation [12]:

μi ¼ μf i  μf i  μsi eai jvi j (27)

The coefficient of friction can also be modeled to be temperature-dependent. Such model requires
test data for calibration, but in general, the following relation may be used:
 
μi ¼ μmin;i þ μmax;i  μmin;i eah T i (28)

In this equation, Ti is rise of the temperature of surface i, and the friction coefficient is equal to μmax,i
when temperature rise is zero (start of motion) and decreases exponentially with rising temperature to a
minimum value μmin,i. Quantity ah is a heating rate parameter. Furthermore, the calculation of the
temperature rise requires a theory of how heat is generated and partitioned at each sliding interface.
Herein, the theory presented in Constantinou et al. [13] is utilized, where it is assumed that the liner
material at each sliding interface is an insulator so that heat is only conducted towards the metal part
of the interface. The theory has been validated in experiments [13]. For the calculation of the
temperature rise, the heat flux is assumed to be supplied at the center of the sliding surface i (i = 1,
2, 3, and 4) and given by
8
< 0 ; θi Ri > bi =2
_
Q i ðt Þ ¼ μi W i θ i R i (29)
: ; θi Ri < bi =2
Ai

Ai is the contact area of interface i (Ai = πb2i /4). Note that quantity μi W i θ_ i Ri =Aiis the heat flux at the
sliding interface i. The temperature rise at the center of the each surface is calculated by the
convolution integral [13]:
rffiffiffiffi t
D Q i ðt Þ
T i ðt Þ ¼
1
k
∫ pffiffiffiffiffiffiffiffiffiffi dτ
π0 t  τ
(30)

In Equation (30), D and k are the thermal conductivity and thermal diffusivity of the material
conducting heat (appropriate values to use are those of stainless steel: D = 0.444 × 105 m2/s,
k = 16.3 W/(m °C)), respectively. The significance of frictional heating and its effects on response
have been studied in Kumar et al. [14] for the single FP bearing. Results demonstrating the heating
effects on the behavior of triple FP bearings based on the formulation of this paper are presented in [7].

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2016)
DOI: 10.1002/eqe
A. A. SARLIS AND M. C. CONSTANTINOU

The restrainer force Fri of surface i may be modeled as


0 ; jθi j ≤ di =Ri
F ri ¼ (31)
k ri Ri ðθi  sgnðθi Þd i =Ri Þ ; jθi j > di =Ri

where kri is the shear stiffness of the restrainer, approximately given by


πsi t ri F ry
k ri ¼ (32)
6Y r

In Equation (32) tri is the thickness of restrainer i shown in Figure 1, si is the diameter of the sliding
surface (si = bi + 2di), and Fry is the shear yield strength of the material. The stiffness value kri as
aforementioned is based on the shear strength of a 60° wedge of the restrainer ring divided by an
arbitrary yield displacement Yr (2.5 mm is a representative value). A value of 172 MPa for Fry is
representative of the ductile iron typically used in large TFP bearings.
The model of Equation (31) assumes infinite restrainer strength. A more appropriate model is to
consider shear failure of the restrainer when the strength is reached. Equation (33) describes this
more complex restrainer behavior. Note that Equation (33) distinguishes between left and right
restrainers so that the failure of the restrainer on one side does not affect the failure of the restrainer
on the opposite side:
8
> 0; jθi j ≤ d i =Ri
>
>
>
> k R ðθ  d =R Þ þ c θ_ ; θ > d =R and jmax½θ ð0 ≤ t ≤ t Þj < d =R þ Y r
>
< ri i i i i ri i i i i i n i i
Ri
F ri ¼ (33)
>
> _ Yr
>
> k R ð θ þ d =R Þ þ c θ ; θ < d =R and j min ½ θ ð 0 ≤ t ≤ t Þ  j < d =R þ
>
>
ri i i i i ri i i i i i n i i
Ri
:
0 otherwise
In Equation (33), tn is the time of the nth integration step and |max θi(0 ≤ t ≤ tn)|denotes the absolute
value of the maximum value of the sliding angle from time zero up to the n-th integration step.
Sliding angles θ3 and θ4 in Equations (29), (31), and (33) are calculated using Equation (15) and
using the known values ofθ1andθ2. Also, velocities θ_ 3 and θ_ 4 of surfaces 3 and 4 are required in
Equations (24), (27), and (29). These quantities are obtained by numerical differentiation of
Equation (15) using the known values of θ_ 1 , θ_ 2 , and u_ ˙ (these variables are obtained by numerical
integration per Equation (25)).

5. EXAMPLE OF ANALYSIS AND COMPARISONS TO OTHER ANALYTICAL OR


EXPERIMENTAL RESULTS

Figure 7 presents a comparison of results obtained by the dynamic analysis model and the algebraic
solution for the case of the bearing of case 1 in Table IV. Note that the behavior of this bearing
cannot be predicted by the Fenz and Constantinou model asμ2 ≠ μ3. Figure 7 shows the force–
displacement loop (force normalized by the constant vertical load W), offset angle-displacement
loops (normalized by the offset angle capacitiesθsi,max = bi/2Ri), and sliding angle-displacement loops
(normalized by the sliding angle capacitiesθi,max = di/Ri). Note that quantity umax in Figure 7 used to
normalize the displacement is the displacement capacity of the bearing. For the comparison, any
velocity dependence of the coefficients of friction is ignored and the mass of the moving parts of the
bearing is artificially made very small so that their inertia and weight effects are negated, conditions
for which the algebraic solution is valid. The results of the numerical and algebraic solutions are
essentially the same.
In another example, a rigid mass of 1500 kN weight supported by bearings having the geometric and
frictional properties of the full-scale bearings (either cases 1 or 2) in Table IV was analyzed when
subjected to the November 1999 Duzce, Turkey earthquake record at the Duzce station, FN
component (with PGA scaled to 0.68 g). Results are presented in Figure 8. Analysis was performed
using the developed dynamic analysis model without consideration of bearing weight and velocity-

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2016)
DOI: 10.1002/eqe
REVISED TRIPLE FRICTION PENDULUM BEARING MODEL

Figure 7. Comparison of results obtained using the algebraic solution and the numerical solution of the dy-
namic analysis model for full-scale bearing case 1 with μ1 = 0.075, μ2 = 0.05, μ3 = 0.01, μ4 = 0.125.

dependent friction effects. The same system was also analyzed using the series model implemented in
program SAP2000 [3, 15, 16]. Note that the series model can accurately model the behavior of the
bearing of case 2 (case of equal friction at sliding interfaces 2 and 3 and velocity independent
friction). However, for case 1, where friction is unequal at sliding surfaces 2 and 3, the series model
cannot be directly used. In this analysis, the series model was implemented with the assumption that
μ2 ¼ μ3 ¼ 0:0339, which corresponds to μ2 = μ3 = 0.03 in the current model (Equation (5)). That is,
the average of the two friction values was used in the approximate series model representation. It is
noted that in the case of the series model, based on the theory of Fenz and Constantinou that
neglects moment equilibrium, the friction coefficient at interface i is denoted as μi , whereas in the
current model, the friction coefficient is denoted as μiwith the two values related by Equation (5).
The results of Figure 8(b) for the case of the bearing with μ2 = μ3are nearly identical for the two
analyses, thus demonstrating the validity of the Fenz and Constantinou model. The loops for the
case shown in Figure 8(a) (whereμ2 ≠ μ3) have small differences at small amplitudes of vibration
and also small differences during unloading and reloading at large amplitudes. These differences
were observed to be larger when velocity effects on friction were considered [7]. Nevertheless, the

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2016)
DOI: 10.1002/eqe
A. A. SARLIS AND M. C. CONSTANTINOU

Table IV. Properties of full-scale triple friction pendulum bearings analyzed.


Geometric and frictional properties Case 1: μ2 ≠ μ3 Case 2: μ2 = μ3
μ1 0.075 0.075
μ4 0.125 0.125
μ2 0.05 0.05
μ3 0.01 0.05
R1 = R4 (mm) 3962
R2 = R3 (mm) 991
h2 = h3 (mm) 115
h1 = h4 (mm) 165
d1 = d4 (mm) 406
d2 = d3 (mm) 152
b1 = b4 (mm) 762
b2 = b3 (mm) 406.4

Friction coefficients μ1, μ2, μ3, μ4are used in the current model μ1 ; μ2 ; μ3 ; μ4 in the Fenz and Constantinou model
are given by Equation (5). Displacement capacities in the Fenz and Constantinou model are given by di ¼
d i ðRi  hi Þ=Ri .

Figure 8. Comparison of results obtained by the series model in program SAP2000 and the dynamic analysis
model for full-scale bearings in Table IV.

example demonstrates that consideration of moment equilibrium does not result in any important
differences in behavior and that the simpler model of Fenz and Constantinou is sufficiently accurate
for most practical cases.
Figure 9 presents results of analysis of the rigid structure of the example in Figure 8 with bearing
case 2 in Table IV (case withμ2 = μ3 = 0.05) under load of 1500 kN and then again under load of
300 kN and with or without consideration of inertia effects of the moving bearing parts. The results in
Figure 9 are for the loop of the base shear force (normalized by vertical load) versus displacement in
the two cases of vertical load. Evidently, there are small effects of the inertia of the moving parts that
tend to increase as the load of the bearing is reduced. This indicates the significance of the inertia of the
bearing parts when uplift occurs or is imminent.
Experiments were conducted with a small-scale bearing having the following geometric and unusual
frictional properties: R1 = R4 = 473 mm, R2 = R3 = 76 mm, h1 = h4 = 38 mm, h2 = h3 = 23 mm,
d1 = d4 = 64 mm, d2 = d3 = 19 mm, μ1 = μ4 = 0.064, μ2 = μ3 = 0.168. The frictional properties were
determined in quasi-static motion so that heating effects were not important. The bearing meets the
constraints of configuration B in Table I and cannot be analyzed with the model of Fenz and
Constantinou as μ2 = μ3 > μ1 = μ4. Figure 10 presents a close-up of the force–displacement relation

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2016)
DOI: 10.1002/eqe
REVISED TRIPLE FRICTION PENDULUM BEARING MODEL

Figure 9. Comparison of results of dynamic response history analysis of structure with bearing case 2 per
Table IV without and with due consideration of inertia effects of bearing parts.

Figure 10. Comparison of analytical and experimental results for tested unusual bearing with μ2 = μ3 > μ1 = μ4.

of the bearing at large displacement. Note that sliding first occurs on surfaces 1 and 4 until the lateral
displacement reaches about 120 mm, where it exhibits a stop until the lateral force becomes equal to
about 0.37 times the vertical force when it then starts sliding on surfaces 2 and 3. The analytical
model (denoted 3pleANI in the figure) predicts this unusual behavior very well. For the analytical
prediction, the dynamic analysis model was utilized with due account for material weights and
initial offsets of the parts of the bearing. More details and results of this test can be found in [7].
The interested reader is referred to [7] for additional studies and results. Particularly, the model
presented in this paper without the effects of initial offsets and inertia of the bearing parts has been
compared with the model of Becker et al. [8] in four cases of unusual geometric and frictional
parameters that could not be analyzed by the model of Fenz and Constantinou. The two models
exhibited differences that were small and presume\d to be the result of moment equilibrium, which
constitutes the main difference in the two models. It was concluded that the two models produce
results that are essentially the same for most practical configurations of TFP bearings.

6. SUMMARY AND CONCLUSIONS

The Fenz and Constantinou [1–5], Morgan [6], Becker et al. [8], Ray et al. [9] and Dao et al. [10]
models of behavior of the TFP bearing are based on the assumption that forces act at the center of
each contact surface, whereas equilibrium of moments requires that the force acts at a non-fixed
location during motion of the bearing.

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2016)
DOI: 10.1002/eqe
A. A. SARLIS AND M. C. CONSTANTINOU

This paper presented a new derivation of the behavior of TFP bearing that relaxes the constraints of
past theories and can thus describe its behavior for any random geometric and frictional parameters.
Key feature of the model is that it considers forces acting away from the center of each contact
surface as required by moment equilibrium. Analytic solutions in terms of algebraic equations for
the force–displacement relations of two configurations have been presented. These exact solutions
are valid for friction values that are constant (independent of velocity or temperature). The utility of
these solutions is in checking the validity of numerical solutions of more complex models.
Moreover, a more complex model suitable for dynamic analysis has been presented, which again
accounts for moment equilibrium and the effects of the weight and mass of the bearings parts,
obtains the motion at each sliding interface, can directly account for instantaneous velocity and
temperature effects on friction at each sliding surface, and can account for bearing top and bottom
rotations. The model may be extended to capture uplift behavior as shown in [7]. Results obtained
by this model were in good agreement with experimental results in a case of unusual combination of
frictional properties that produced a behavior that could not be predicted by the Fenz and
Constantinou model.
Results obtained with the presented model in a case of dynamic analysis of a simple isolated
structure demonstrated the significance of the inertia effects of the moving bearing parts as load on
the bearing is diminished. Moreover, other studies in [7] using the presented model demonstrated
the importance of the instantaneous heating effects on the prediction of the isolator displacement
demand.
Comparison of results produced by the presented model with due consideration of the moment
equilibrium effects produced results that are for practical purposes the same as those produced by
the simpler Fenz and Constantinou model, provided that the following basic conditions of the latter
model apply: (i)Reff1 = Reff4 ≫ Reff2 = Reff3; (ii)μ2 ¼ μ3 < μ1 < μ4 ; (iii) d 1 > ðμ4  μ1 ÞReff 1 ; (iv) d2 >
ðμ1  μ2 ÞReff 2 ; and (v) d3 > ðμ4  μ3 ÞReff 3 and that friction is interpreted in a particular way that is
clarified in this paper. It should be noted that the conditions for validity of the Fenz and
Constantinou model are met for the commonly used bearings and that the presented more complex
model is only useful in a few special cases. Moreover, studies in [7] compared results produced by
the presented model with results produced by the Becker et al. [8] model for cases of unusual
combinations of geometric and frictional properties that cannot be analyzed by other models. Small
differences were observed and presumed to be the result of moment equilibrium, which constitutes
the main difference in the two models.

ACKNOWLEDGEMENTS
Partial financial support and hardware tested for this project were provided by Earthquake Protection Sys-
tems, Inc., Vallejo, CA. The work was conducted at the University at Buffalo by the first author as part
of his doctoral qualifying examination under the direction of the second author.

REFERENCES
1. Fenz DM, Constantinou MC. Mechanical behavior of multi-spherical sliding bearings Report No. MCEER-08-0007,
Multidisciplinary Center for Earthquake Engineering Research, Buffalo, NY, 2008.
2. Fenz DM, Constantinou MC. Spherical sliding isolation bearings with adaptive behavior: theory. Earthquake Engi-
neering and Structural Dynamics 2008; 37(2):163–183.
3. Fenz DM, Constantinou MC. Spherical sliding ixsolation bearings with adaptive behavior: experimental verification.
Earthquake Engineering and Structural Dynamics 2008; 37(2):185–205.
4. Fenz DM, Constantinou MC. Modeling triple friction pendulum bearings for response-history analysis. Earthquake
Spectra 2008; 24(4):1011–1028.
5. Fenz DM, Constantinou MC. Development, implementation and verification of dynamic analysis models for multi-
spherical sliding. Report No. MCEER-08-0018, Multidisciplinary Center for Earthquake Engineering Research, Buf-
falo, NY, 2008.
6. Morgan TA. The use of innovative base isolation systems to achieve complex seismic performance objectives. Ph.D.
Dissertation, Department of Civil and Environmental Engineering, University of California, Berkeley, 2007.
7. Sarlis AA, Constantinou MC. Model of triple friction pendulum isolators for general geometric and frictional param-
eters and for uplift conditions, Report No. MCEER-13-0010, Multidisciplinary Center for Earthquake Engineering
Research, Buffalo, NY, 2013.

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2016)
DOI: 10.1002/eqe
REVISED TRIPLE FRICTION PENDULUM BEARING MODEL

8. Becker TC, Mahin SA. Experimental and analytical study of the bidirectional behavior of the triple friction pendulum
isolator. Earthquake Engineering & Structural Dynamics 2011; 41(3):355–373.
9. Ray T, Sarlis AA, Reinhorn AM, Constantinou MC. Hysteretic models for sliding bearings with varying frictional
force. Earthquake Engineering & Structural Dynamics 2013; 42(15):2341–2360.
10. Dao ND, Ryan KL, Sato E, Sasaki T. Predicting the displacement of triple pendulum™ bearings in a full-scale shak-
ing experiment using a three-dimensional element. Earthquake Engineering & Structural Dynamics 2013; 42
(11):1677–1695.
11. Nagarajaiah S, Reinhorn AM, Constantinou MC. Nonlinear dynamic analysis of three dimensional base isolated
structures (3D-basis), Report NCEER-89-0019, National Center for Earthquake Engineering Research, State Univer-
sity of New York, Buffalo, NY, 1989.
12. Constantinou MC, Mokha A, Reinhorn AM. Teflon bearings in base isolation. II: Modeling, ASCE Journal of Struc-
tural Engineering 1990; 116(2):455–474.
13. Constantinou MC, Whittaker AS, Kalpakidis Y, Fenz DM, Warn GP. Performance of seismic isolation hardware un-
der service and seismic loading, Report No. MCEER-07-0012, Multidisciplinary Center for Earthquake Engineering
Research, Buffalo, NY, 2007.
14. Kumar M, Whittaker AS, Constantinou MC. Characterizing friction in sliding isolation bearings. Earthquake Engi-
neering & Structural Dynamics 2015; 44(9):1409–1425.
15. Computers and Structures Inc. SAP2000: integrated finite element analysis and design of structures, Version 11.0.8,
Berkeley, CA, 2007.
16. Sarlis AA, Constantinou MC. Modeling of triple friction pendulum isolators in program SAP2000, supplement to
MCEER Report 05-009, document distributed to the engineering community together with executable version of pro-
gram and example files, University at Buffalo, 2010.

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2016)
DOI: 10.1002/eqe

You might also like