COMPDYN 2011

3rd ECCOMAS Thematic Conference on

Computational Methods in Structural Dynamics and Earthquake Engineering
M. Papadrakakis, M. Fragiadakis, V. Plevris (eds.)
Corfu, Greece, 25–28 May 2011

NEGATIVE STIFFNESS DEVICE FOR SEISMIC PROTECTION OF

STRUCTURES – AN ANALYTICAL AND EXPERIMENTAL STUDY
Apostolos A. Sarlis1, Dharma Theja R. Pasala2, Michael C. Constantinou1, Andrei M.
Reinhorn1, Satish Nagarajaiah2 and Douglas Taylor3
1
State University of New York at Buffalo, Buffalo, NY-14260, U.S.A
{aasarlis,constan1,reinhorn}@buffalo.edu
2
Rice University, Houston, TX-77005, U.S.A
{drp1, Satish.Nagarajaiah}@rice.edu
3
Taylor Devices Inc., North Tonawanda, NY-14120, U.S.A
[email protected]

Keywords: Negative Stiffness Device, Apparent yielding, Passive, True Negative Stiffness

Abstract. Structural weakening and damping is an approach previously proposed for the
reduction of seismic forces and drifts in structures. While this approach is very efficient, it
requires a reduction in strength, which will result in inelastic excursions and permanent
deformation of the structural system during the seismic event. This paper describes a true
Negative Stiffness Device (NSD) that can emulate weakening of the structural system without
inelastic excursions and permanent deformations. The Negative Stiffness Device (NSD) is a
self contained device that produces a force which is in the same direction as the imposed
displacement, thus the name “negative stiffness”. The device is inherently unstable but when
installed in a structure transfers a force at its installation level which in turn reduces all
forces, drifts and accelerations above that level. The NSD is therefore capable of limiting the
forces developed in a structure without any yielding which could lead to permanent
deformations. The NSD was developed at the University at Buffalo (UB), Rice University and
Taylor Devices Inc. and tested on a shake table by authors in a three storied structural model
isolated with elastomeric bearings. The NSD consists of a highly compressed spring in a
double negative stiffness magnification mechanism. In order to ensure a considerable amount
of positive stiffness of the global system around small displacements to minimize wind
vibrations a so called gap spring assembly (GSA) mechanism is implemented which delays the
engagement of the device by a prescribed displacement. Moreover, the device employs double
containment chevron braces that contain the large vertical forces needed for the development
of negative stiffness without transferring these forces to the structure. This paper describes
the development of analytical and computational tools that describe the behavior of the device
and presents results for the experimental verification of these tools.
Apostolos A. Sarlis, Dharma Theja R Pasala, Michael C. Constantinou, Andrei M. Reinhorn, Satish Nagarajaiah
and Douglas Taylor

1 INTRODUCTION
Designing structures for ductile behavior and allowing for significant inelastic action in
strong earthquakes results in reduction of inertia forces. However, this is accompanied by
large drifts, permanent deformations and eventually loss of functionality of the structure after
a seismic event. Reinhorn et al. (2005) [1] and Viti et al. (2006) [2] introduced the concept of
weakening structures (reducing strength and implicitly stiffness), while introducing
supplementary viscous damping to reduce simultaneously total accelerations and inter-story
drifts. However, weakening also reduces strength and results in some permanent
displacements. An alternative approach is to “simulate yielding” by introducing negative
stiffness at predescribed displacements leading to the idea of “apparent weakening”.
The concept of true negative stiffness for structural applications was first introduced by
Nagarajaiah et al [3]. True negative stiffness means the force must assist motion, not oppose
it as it is in the case of a positive stiffness spring. Pseudo negative stiffness [4] can be
accomplished using active or semiactive hydraulic device. True negative stiffness needs no
external power supply. The Negative Stiffness Device (NSD) presented herein is entirely
composed of springs and generates an elastic nonlinear true negative stiffness.
By engaging the NSD at an appropriate displacement (simulated yield displacement),
which is well below the actual yield displacement of the structural system, the composite
structure-device assembly, behaves like a yielding structure. The NSD has a re-centering
mechanism thereby avoiding permanent deformation in the composite structure-device
assembly unless, the main structure itself yields. However, the combined structural system
with just the NSD develops increased structural deformations. Addition of passive dampers
reduces and controls these deformations without any considerable increase in the base shear.
The addition of dampers combined with the reduction of stiffness also lead to an increased
damping ratio which presents an additional advantage in controlling the displacements.
In order to visualize the effect of adding true negative stiffness to a structure, consider the
force displacement plots shown in Figure 1(a) (green line is structure, magenta is viscous
damper and red is negative stiffness device). By adding NSD to the structure, schematically
shown in Figure 1(b), the assembly stiffness reduces to = − beyond the
displacement (shown as blue line in Figure 1(b)). If, and are the maximum restoring
force and maximum displacement of a perfectly-linear system (green line in Figure 1(b)) then
for the same load the maximum restoring force and maximum displacement of the assembly
are and (blue line in Figure 1(b)), respectively. is designed to achieve the desired
reduction in base shear. Force exerted by the NSD is shown as red line in Figure 1(b).
Although the reduction in base shear is achieved the maximum deformation of the adaptive
system is substantially increased in the process when compared with an elastic system.
Deformation of this assembly can be reduced by adding a passive damping device in parallel
to the NSD, schematically shown in Figure 1(c).

2
Apostolos A. Sarlis, Dharma Theja R Pasala, Michael C. Constantinou, Andrei M. Reinhorn, Satish Nagarajaiah
and Douglas Taylor

Figure 1: Working Principle of NSD (Nagarajaiah et al 2010)

Photographs of the NSD are shown in Figure 2a (un-deformed) and Figure 2b (deformed).
The device itself is inherently unstable. It can be installed in an isolated structure between the
ground and the isolation level or in between the floors of any fixed and/ or isolated structure.
Its effect is that it reduces the forces that are transferred above its installation level.
In order to ensure a considerable amount of positive stiffness around small displacements
in order to minimize wind vibrations and account for some variability in structural and NSD
properties, a so called gap spring assembly (GSA) mechanism is implemented which
essentially makes the device engage at a larger displacement (displacement x1 in Figure 1)
rather than at the initial position of the structure.
The current paper focuses on the analytical framework and experimental validation of the
NSD which uses only a part of the experimental NSD testing program. Complete details and
additional analytical and computational issues, further experimental validation together with
shake table testing and the verification of the true Negative Stiffness (NS) concept through
shake table testing can be found in [5, 6].

2 NSD DESCRIPTION

2.1 NSD Operation

The NSD shown in Figure 2a is composed of a pre-compressed spring shown in the center
of the device as well the gap spring assemblies on the bottom. A combination of frame
elements and plates hold these pieces together. When the device deforms, the pre-compressed
spring is the one that creates the force that assists the motion or the negative force and thus the
name negative stiffness for the device. The bottom spring assemblies (gap spring assembly
mechanism) provide the device inherently with a bilinear elastic positive stiffness in order to
make the device engage at larger displacements (displacement x1 in Figure 1). More
specifically, around equilibrium, the positive stiffness caused by the gap spring assembly
mechanism, cancels out the negative stiffness caused by the pre-compressed spring so that
essentially the force/stiffness generated by the device is close to zero. After a prescribed

3
Apostolos A. Sarlis, Dharma Theja R Pasala, Michael C. Constantinou, Andrei M. Reinhorn, Satish Nagarajaiah
and Douglas Taylor

displacement, the gap spring assembly softens drastically so that the pre-compressed spring
acts essentially on its own creating the negative stiffness. It is noted that the operation of the
gap spring assembly is achieved without any yielding so that there is no inherent permanent
deformation in the device.

a) Un-deformed NSD b) Deformed NSD

Figure 2: View of Negative Stiffness Device tested at SUNY Buffalo

Schematic diagram of the device in its deformed configuration is shown in Figure 3 after
imposing a displacement on the top of the device. The lever imposes a displacement on the
top of the pivot plate (point B) making the pivot plate to rotate about point C. Due to the axial
rigidity of the lever and its negligible rigid body rotation, the imposed displacement and the
displacement of Point B are approximately equal. Since the pivot plate rotates about C, point
D moves in the opposite direction from the imposed displacement. Point E is rigidly
connected to the top of the device and therefore has a displacement equal to the one imposed.
Due to the kinematics of points D and E, the pre-compressed spring rotates and its pre-
compression force facilitates the motion rather than opposing it. Moreover since the “negative
force” is generated by a spring whose deformation depends nonlinearly to the NSD imposed
displacement, the NSD exhibits nonlinear elastic behavior.
The spring exhibits its minimum length when the device is un-deformed. As the device
deforms, the spring extends thus its pre-compression force reduces while its inclination angle
increases and the stiffness magnitude generated by the device reduces. This gradual stiffness
reduction will eventually lead to positive stiffness at larger displacements and this is termed as
“stiffening” throughout this paper. This stiffening part can be very important in the case the
maximum considered earthquake has been exceeded since it limits the displacements of the
structure or the isolators.

4
Apostolos A. Sarlis, Dharma Theja R Pasala, Michael C. Constantinou, Andrei M. Reinhorn, Satish Nagarajaiah
and Douglas Taylor

Figure 3: Terminology used for the elements of the Negative Stiffness Device

2.2 Advantages of NSD Tested at SUNY Buffalo

Negative Stiffness in applications other than earthquake has been proposed and used at
small scale for the vibration isolation of equipment by Molyneaux (1957) [7], Platus [8-16],
Carrella et al [17] and for car suspension systems [18]. It has also been proposed in base
isolation [19] with the application of the convex pendulum bearings in parallel with
elastomeric bearings. The NSD described in the current paper is a large scale novel design
that passively generates negative stiffness and it is not restricted to only base isolation. The
NSD presents the following advantages:

 The device changes the apparent global lateral stiffness of the structure, without
changing the actual stiffness of the structure. The apparent stiffness is reduced to a
very low level simulating global lateral yielding without actual yielding in the main
structure.
 The device produces true horizontal negative stiffness by passively generating a
force that assists the imposed displacement. No external power supply is needed since
all the elements comprising the device are passive.
 The device is self contained and therefore when installed affects only the horizontal
stiffness of the system while leaving the vertical stiffness intact. Stability and buckling
limits of the structure are not affected. The NSD does not participate in transferring
the vertical loads.
 There is no significant hysteresis in the device. The NSD is essentially elastic.
 The device provides variable stiffness which becomes positive at large deformations,
therefore its global behavior is “elastic nonlinear”. This is a desired feature as it
promotes stability.
 The device employs a double magnification mechanism that allows for easy
adjustment of the negative stiffness value. The gap spring assembly (GSA) mechanism
allows for adjustable gap opening.

5
Apostolos A. Sarlis, Dharma Theja R Pasala, Michael C. Constantinou, Andrei M. Reinhorn, Satish Nagarajaiah
and Douglas Taylor

3 OUTLINE OF THE EXPERIMENTAL PROGRAM

The experimental setup is shown in Figure 4a. The 3 story structure –used before as a six
story structure [e.g., 20] - is sitting on top of four low damping elastomeric bearings used
before in [20]. The elastomeric bearings were tested individually at the single bearing testing
machine at UB described in [21]. Two NSD’s were installed on the side of the shake table and
were connected to the base of the superstructure with rollers in order to accommodate the
height loss that they exhibit during deformation. The tests conducted at UB were divided in to
two major categories: 1) Displacement controlled tests with and without the gap spring
assembly and 2) Shake table tests involving various configurations of the NSD with and
without viscous dampers. The displacement controlled tests were achieved by externally
bracing the superstructure at the base-mat level through a reaction frame that was constructed
on the side of the shake table and shown in Figure 4a on the right while the bracing system is
shown in Figure 4b.

a) 3story structure tested at UB b) Bracing of superstructure

Figure 4: Experimental Setup

Each NSD is sitting on top of four load cells [22] and instrumented at points B,C,D,E with
string pots and accelerometers. NSD East is also heavily instrumented with Krypton LED’s.
Krypton is an advanced camera coordinate tracking system described in [23].

a) Position of LED’s on NSD East b) Conventional Instrumentation of NSD East

Figure 5: Instrumentation for NSD East

6
Apostolos A. Sarlis, Dharma Theja R Pasala, Michael C. Constantinou, Andrei M. Reinhorn, Satish Nagarajaiah
and Douglas Taylor

4 ANALYTICAL EQUATIONS OF NEGATIVE STIFFNESS DEVICE

In order to derive the force displacement equations of the device, equilibrium at the
deformed shape needs to be considered since essentially the operation of the device is based
on geometric nonlinearity. Therefore one needs to consider the deformed shape and the device
being supported horizontally on top by the application of a force which is essentially the
negative force generated by the device. The forces acting on the pivot plate caused by the pre-
compressed spring and the lever are shown in Figure 6. From geometry and considering the
fact that point C is fixed, the displacements of other points of the device are:

Figure 6: Forces acting on pivot plate

u B  u E  u D l2 u (1)
l1
In Figure 6,  s is the angle of the spring  p the angle of the pivot plate with respect to
vertical.  s and  p are obtained from equation (2):

1  l1 
 s  arcsin u 1  
l s  l 2  (2)
u
 p (t )  arcsin
l2
The spring length at the deformed configuration is given by:
2
  u  
2
 l 
2

l s   l p  l1  l1 1      u 2 1  1  (3)

  l 2    l2 

7
Apostolos A. Sarlis, Dharma Theja R Pasala, Michael C. Constantinou, Andrei M. Reinhorn, Satish Nagarajaiah
and Douglas Taylor

l p is the length of the spring when the NSD is un-deformed and Fs is the force of the pre-
compressed spring:
Fs  Pin  K s l s  l p  (4)
Pin : is the pre-compression force of the spring and should have a positive value and K s is
the stiffness of the pre-compressed spring.
Writing the equilibrium equations for the pivot plate for the forces of Figure 6 and using
equations (1)-(4) yields the horizontal force at point C. Then from Figure 3 one can see that
the total NSD force is simply FNSD  FC  Fg . The final expression for the force generated by
the NSD as a function of the imposed displacement is given in equation (5):

 Pin  K s l p  l  l l p  l1 
FNSD    K s  1  2  2  u  F
(5)
ls   g
  l 2  l1 2
l2  u 2 
The force displacement of the gap spring assembly for monotonic positive displacement is
given by:
k g1u, 0  u  d gap 
Fg    (6)
 k g1d gap  k g 2 u  d gap  u  d gap 

d gap is a prescribed displacement at which the gap spring assembly softens and H is the
Heaviside function

Quantity Value
Distance from spring pin to fixed pin l1  10in
Distance from lever pin to fixed pin l 2  5in
Spring length l p  30 in

Spring rate K s  0.8 kips

in
Preload Pin  3.7kips
Gap opening d gap  0.65in
Gap Spring Assembly Stiffness for u  d gap K g ,1  2.81kips / in
Secondary spring assembly stiffness u  d gap K g , 2  0.17kips / in

Table 1: NSD Properties

The force displacement loops for the derivations above are generated in figure 7 for the
properties shown in Table 1.

8
Apostolos A. Sarlis, Dharma Theja R Pasala, Michael C. Constantinou, Andrei M. Reinhorn, Satish Nagarajaiah
and Douglas Taylor

4 4
NSD w/out gap assembly NSD w/ gap assembly

2 2
Force (kips)

Force (kips)
0 0

-2 -2

-4 -4
-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3
Displacement (in) Displacement (in)
Figure 7: Force versus NSD imposed displacement plots for NSD with and without the gap spring assembly

5 MODELING OF NSD IN SAP2000

Modeling the NSD in user subroutines is straightforward by implementing equations (5)
and (6). In SAP2000 [24], the NSD behavior can be modeled using the multi-linear elastic
spring element. One can directly import the force displacement loop of the NSD into
SAP2000. There is no restriction in SAP2000 for the sign of the stiffness and therefore
negative springs can be implemented as well. The gap spring assembly can be also modeled
using the same multi-linear element. The NSD model in SAP2000 is therefore a parallel
arrangement of two elements 1) a multi- linear elastic element representing the negative
nonlinear elastic component from equation (5) with Fg=0 and 2) a multi- linear elastic
element representing the positive bilinear elastic component generated by the gap spring
assembly and given by equation (6).
The configuration as implemented in SAP2000 is shown in Figure 8 on the left. Both
elements ML1 and ML2 share the same joints on top and bottom. Comparisons of the results
obtained in the SAP2000 model for a displacement controlled test are compared with the
analytical equations in Figure 8 on the right. Results match identically.

2
SAP2000
1 Analytical
Force(kips)

0
-1
-2
-3 -1 1 3
Displacement (in)
Figure 8: SAP2000 NSD elements for the cases with gap and without gap and comparisons of SAP2000 results
with analytical results

9
Apostolos A. Sarlis, Dharma Theja R Pasala, Michael C. Constantinou, Andrei M. Reinhorn, Satish Nagarajaiah
and Douglas Taylor

6 ANALYTICAL INVESTIGATIONS AND MODELING OF THE DETAILED NSD

BEHAVIOR
For most practical applications, the analytical equations and the modeling of sections 4 and
5 is sufficient to provide highly accurate results. The derivations of this section are intended to
provide the reader with complete understanding of the detailed mechanics of the NSD.
Understanding the complete mechanics of a system is essential in its practical implementation
and later on use of the simplified models.

6.1 Analytical Derivations of NSD Dynamics

The rigid body dynamics of the Negative Stiffness are derived here using a Lagrangian
formulation. The masses and moments of inertia in the actual device are shown in Figure 9.
All frame elements connected to the top channel and the double hinged columns, undergo an
inverted pendulum motion around the base and their moment of inertia is lumped at F and H
respectively. At point C the mass moment of inertia of the pivot plate is lumped. The center of
mass of the spring undergoes both translation and rotation. Its mass m s is lumped at its center
of mass located at point S. Its moment of inertia varies with displacement u.

Figure 9: Discretization of the device

Considering the fact that the rotations of the individual components are coupled to the
imposed displacement there is only one degree of freedom. The rotation of the pivot plate and
spring is given by equation (2), the horizontal displacement of the center of mass of the spring
is the average of the displacements of points D and E from equation (1) while the vertical
displacement is half the vertical displacement of point D and can be easily calculated from
geometry of Figures 3, 6 and 9 while angle θ can be calculated from the geometry of figure 9.
All quantities above are functions of the imposed NSD displacement alone.
The Euler-Lagrange equation for the NSD is:
d  L  L
  0
dt  u  u (7)
L  K V
Where K is the total kinetic energy of the system and V is the total potential energy of the
system with respect to a reference coordinate system given as the sums of the kinetic and

10
Apostolos A. Sarlis, Dharma Theja R Pasala, Michael C. Constantinou, Andrei M. Reinhorn, Satish Nagarajaiah
and Douglas Taylor

potential energies of the individual moving parts of the device. The total kinetic and potential
energies (with reference to C) of the system are given from:
1  2 1 1 1 1 1
I p p  ms u s  ms v s  I ss  I h 2  I f  2
2 2 2
K
2 2 2 2 2 2
(8)
 v   h 
V  m f g l p  l1  y cos   ms g  l1  DE   mc g  h  v LB  cos   Vel , s  W
 2   2 
The potential energy coming from the deformation of the spring in equation (8) is:
2
1   P 
Vel , s  K s l s   l p  in  (9)
2   K s 

The external work done by a restraining force in (8) has to be expressed as a potential so:
W    Fdu   Fu  C (10)
It is noted that the force above is the force needed to be applied at the top of device in order
to hold the device in place and make it stable. Equivalently this is the force generated by the
device.
Using equations (7)-(10), (1)- (4) and after performing some tedious algebra which
involves the calculation of the derivatives involved in equations (7) and (8) we can get the
equation of motion of the NSD which is given by:
M a u  M v u 2  M u u  F (11)
The terms Ma, Mv and Mu, are given by complex expressions and final results are omitted
for simplicity. The force displacement loops of the device for various driving frequencies are
plotted in Figure 10 for the device properties of table 1. The effect of the inherent dynamics is
that of increasing the negative stiffness and maximum force generated by the NSD while
slightly delaying the initiation of stiffening. The driving frequencies of Figure 10 are only for
demonstration purposes. The accelerations and velocities for the higher frequencies shown
there are unrealistic and therefore the effect of the inherent dynamics is negligible.
f=0.1Hz
6 f=1Hz
f=1Hz
4
f=1.2Hz
f=1.4Hz
Force (kips)

2
f=1.6Hz
0 f=1.8Hz
f=2Hz
-2 f=2.2Hz
f=2.4Hz
-4
f=2.6Hz
-6 f=2.8Hz
-4 -3 -2 -1 0 1 2 3 4 f=3Hz
Displacement (in)
Figure 10: Force Displacement Loops of NSD for various excitation frequencies

11
Apostolos A. Sarlis, Dharma Theja R Pasala, Michael C. Constantinou, Andrei M. Reinhorn, Satish Nagarajaiah
and Douglas Taylor

6.2 Analytical Derivations of NSD Imperfections

Although the pins are typically assumed to be frictionless and dimensionless, in reality they
are not. When a pin is connecting two members, in order for the member to start rotating
relatively one to the other, a friction force must be overcome either between one member and
the pin, or the other member and the pin, or both. For example consider the assembly shown
in Figure 11a:

a) Pin Friction modeling b) Pivot Plate Free body Diagram c) Double Hinged Column
FBD
Figure 11: Pin friction model

In Figure 11a, it is assumed that rotation of the pin occurs because connecting part 1 rotates
relatively to the pin while the pin and part 2 are in sticking conditions. In order to have a
relative rotation between one of the connecting parts and the pin, we need to apply a torque on
the pin of magnitude:
M p  r p Pp (12)

r is the radius of the pin, Pp is the resultant force acting on the pin and  p is the friction
coefficient.
Figure 11b shows the free body diagram of the pivot plate and 11c of the doubled hinged
columns considering friction in the pins. Based on Figure 11b, due to the moment at points B
and A, the lever now has a shear force equal to:
2M B
FBy  (13)
l lv
Writing the equilibrium equations for Figure 11b the unknown forces can be computed:

12
Apostolos A. Sarlis, Dharma Theja R Pasala, Michael C. Constantinou, Andrei M. Reinhorn, Satish Nagarajaiah
and Douglas Taylor

2M B 
 l 2  l 2  u 2 
2
MD  MB  MC 
F l  l p  l1  llv  
FBx   s 1 1  u 
ls l2  2
l2  u 2  2
l2  u 2
 
F  l 
FCx  FBx  s u1  1 
ls  l2  (14)
l 2
l p  l1  1 l 2  u 2
l2
FCy   Fs  FBy
ls
From Figure 11c, the end moments caused by the friction in the pins cause a shear force
that is constant throughout the height of the double hinged columns:
2M L
FLx  (15)
h
The total force of the NSD becomes:
FNSD  FCx  2 FLx  Fg (16)

The equation above can be decomposed into a nonlinear elastic part and a hysteretic part,
FNSD  Fel  Fpl . We approximate the hysteretic moment by using the Sivaselvan-Reinhorn
[25] hysteretic model:
2
 M 
 
dM  k o 1  sign M sign d d (17)
 M y 

d is the rotation increment of the pin being different for each pin, du is the increment of
the imposed displacement while ko elastic stiffness of the pin prior to slipping occurring.
According to equation (17), in order to calculate the moments of the pins, we need to
calculate the resultant forces as well as the rotation increments of all the pins. The resultant
forces can be calculated from the free body diagrams of Figure 11b, c. The pin rotation
increments can be calculated by differentiating the pin rotations with respect to the imposed
NSD displacement and then solving for the rotation differentials. The pin rotations are given
from equation (2) for point C, similarly for point B if the negligible rotation of the lever is
ignored, as  D   C   s for point D with the help of equation (2) for the spring angle and
from the geometry of Figure 9 for the rotation θ of the double hinged columns.
The set of equations (12)-(15) and (17) written for each pin with its corresponding resultant
forces and rotation increments represent a system of equations on the unknowns. Consider the
unknown vector at time step n  1 :
T
x(n1)  MB n1 MC n1 MD n1 ML n1 F(n1) FBx(n1) FBy(n1) FCx(n1) FCy(n1) FLx(n1)  (18)

Collecting all terms of the equations on the left hand side and define f x n 1   0 , using
( 0)
x n 1  x n as the first iteration, then Newton Raphson iteration follows as:

13
Apostolos A. Sarlis, Dharma Theja R Pasala, Michael C. Constantinou, Andrei M. Reinhorn, Satish Nagarajaiah
and Douglas Taylor

x n 1 ( j 1)  x n 1
( j)

 Df 1 x n 1
( j)
 f x n 1
( j)
 (19)
Note that the initial conditions at the first step is x 0  0 0 0 0 0 0 0 0 Pin 0
Results are plotted in Figure 12 for the properties listed in table 1 for various friction
coefficients. Hysteresis in the NSD is proportional to the horizontal force exerted by the NSD.
At large displacements when the NSD force becomes zero so does the hysteretic force while
after that it maintains the same sign.

1 m=0.05
m=0.08
Hysteretic Force (kips)

0.5 m=0.1
m=0.15
m=0.2
0 m=0.25
m=0.3
m=0.35
-0.5
m=0.4
m=0.45
-1 m=0.5
-4 -3 -2 -1 0 1 2 3 4
5 m=0.05
m=0.08
Total NSD Force (kips)

m=0.1
m=0.15
m=0.2
0 m=0.25
m=0.3
m=0.35
m=0.4
m=0.45
-5 m=0.5
-4 -3 -2 -1 0 1 2 3 4
Displacement (in)
Figure 12: Force Displacement Loops of NSD for various friction coefficients

6.3 Analytical Derivations of NSD Inherent Flexibilities

All the equations up to this point have been derived based on the assumption that point C is
fixed and point E undergoes only rigid body motion driven by the top channel of the device.
This is only true if the connecting frame elements, namely, the top supporting and the bottom
supporting frames are considered rigid. This in most cases is only approximately true.
In order to visualize the flexibility effects, consider an equivalent NSD model shown in
Figure 13 in equivalence to the points of Figure 3. Now, at point C a spring kb is added to
approximately represent the flexibilities of the bottom frame together with the bottom
supporting channel. Similarly for the top supporting frame, the rigid frame element is no
longer assumed to be continuous since a spring kt is added at its midpoint in order to
approximate the flexibilities of the top supporting frame together with the top NSD channel.

14
Apostolos A. Sarlis, Dharma Theja R Pasala, Michael C. Constantinou, Andrei M. Reinhorn, Satish Nagarajaiah
and Douglas Taylor

Figure 13: Flexibilities considered in the equivalent NSD model

Equilibrium equations and geometric equations need to be rewritten using equation (20)
instead of equation (1):
u f ,E  u E  u

l1 (20)
u D  u  u C   uC
l2
u fE : Displacement of point E due to flexibilities of the top supporting frame
The spring length is now given from:
2
  
2   
2
 2 l  l1
l s   l p  l1  l1   u  u C  1    u  u C  l  u C  u E  (21)
  l2 
   2 

The horizontal reaction at point C can be calculated from moment and force equilibrium of
the pivot plate for the new geometry:

 Pin  K s l p l 
 l p  l1    l 
FCx    K s  1   1 u  u C   u E  u C 1  2  (22)
 ls  l2  l 2  u  u C 
2 2   l1 

Moreover from equilibrium of forces at point C one gets:
FCx  k b u C (23)
While from equilibrium of point E after substituting for the spring inclination angle:
 Pin  K s l p  l  
   K s  u  u C  1   u C  u E   Fg  k t (u  u E ) (24)
 ls   l2  
The force displacement of the GSA is related to the total displacement of point E and is
given for positive displacements from equation (25):

15
Apostolos A. Sarlis, Dharma Theja R Pasala, Michael C. Constantinou, Andrei M. Reinhorn, Satish Nagarajaiah
and Douglas Taylor

k g1u E , 0  u E  d gap 
Fg    (25)
 k g1d gap  k g 2 u E  d gap  u E  d gap 

Then the total NSD force is given from:

FNSD  FCx  Fg (26)
Equations (22)-(26) together with the help of (21) and (4) are a system of 5 nonlinear
equations on 5 unknowns which are collected in the following vector
T
x  FCx Fg FNSD u C u fE  . Then similarly to the hysteresis derivations, all equations
are brought on the left hand side, written at step n  1 and iterate based on equation (19).

2 Kb/Ko=1.1
Force (kips)

Kb/Ko=1.3
0 Kb/Ko=1.7
Kb/Ko=1.9
Kb/Ko=2.1
-2 Kb/Ko=2.3
Kb/Ko=2.7
Kb/Ko=3.1
-4
-3 -2 -1 0 1 2 3 Kb/Ko=3.8
a) Force Displacement of NSD without Gap Spring Assembly
Kb/Ko=4.6
Kb/Ko=5.4
4
Kb/Ko=5.7
Kb/Ko=7.6
2 Kb/Ko=9.6
Force (kips)

Kb/Ko=17.2

-2

-4
-3 -2 -1 0 1 2 3
Displacement (in)
b) Force Displacement of NSD with Gap Spring Assembly
Figure 14: NSD force displacement loops assuming rigid top frame and flexible bottom supporting frame for
various stiffness values of the bottom supporting frame

It is noted, that the flexibility of the bottom frame is equivalent to arranging a negative
spring in series with a positive spring. Figure 14 shows the results assuming the top frame is
rigid and considering various stiffness values for the bottom supporting frame. Legend values
are given as ratios of the stiffness of the bottom supporting frame to the Negative NSD
stiffness at zero displacement. Flexibility of the bottom frame leads to increased negative
stiffness generated by the NSD while the peak force is unchanged and therefore stiffening of
the NSD occurs faster than if the bottom frame was rigid. An additional effect of increasing

16
Apostolos A. Sarlis, Dharma Theja R Pasala, Michael C. Constantinou, Andrei M. Reinhorn, Satish Nagarajaiah
and Douglas Taylor

the Negative Stiffness is the fact the gap spring assembly becomes ineffective since the
imposed NSD displacements together with the NSD force becomes inconsistent as shown in
Figure 14b.
Results assuming a rigid bottom frame and a flexible top frame are shown in Figure 15a.
Due to this flexibility, this case affects the results only when the gap spring assembly is
present since point E cannot follow the imposed NSD displacement. This is obvious in Figure
15b which is the force versus NSD imposed displacement of the gap spring assembly.

2
Force (kips)

Kt/Ko=0.2
0
Kt/Ko=0.4
Kt/Ko=0.6
-2 Kt/Ko=0.8
Kt/Ko=1.0
Kt/Ko=1.1
-4 Kt/Ko=1.3
-3 -2 -1 0 1 2 3
a) Force Displacement of NSD with Gap Spring Assembly Kt/Ko=1.5
Kt/Ko=1.9
4
Kt/Ko=2.7
Kt/Ko=3.4
2 Kt/Ko=5.7
Force (kips)

Kt/Ko=7.6
Kt/Ko=9.6
0 Kt/Ko=17.2

-2

-4
-3 -2 -1 0 1 2 3
Displacement (in)
b) Force Displacement of NSD with Gap Spring Assembly
Figure 15: NSD and Gap Spring Assembly force displacement loops assuming rigid bottom and flexible top
supporting frame for various stiffness values of the top supporting frame

7 EXPERIMENTAL RESULTS
Results from the experimental study conducted at UB are presented here for the
displacement controlled tests. These results are only a minor portion of the complete
experimental program. Complete details, additional analytical derivations and results are
presented in [5 and 6]. Figure 16 shows the NSD force displacement loops obtained for
displacement controlled tests conducted at 0.02Hz as well as comparisons of the forces
obtained from the elastomeric bearings, the NSD’s and the total base shear. Installing the
Negative stiffness device indeed reduces the total base shear and generates a “negative force”
if one compares the red line with the blue line. The blue line represents what the base shear
would have been in the absence of the NSD.

17
Apostolos A. Sarlis, Dharma Theja R Pasala, Michael C. Constantinou, Andrei M. Reinhorn, Satish Nagarajaiah
and Douglas Taylor

2 4
ENA-DHD1-EB DHD1-EC

NSD East Force (kips)

2
NSD East Force(kips)

0 0

-1 -2

-2 -4
-3 -2 -1 0 1 2 3 -4 -2 0 2 4
NSD East Displacement (in) NSD East Displacement (in)
20 20
Total Base Shear Total Base Shear
Elastomeric Bearings Elastomeric Bearings
10 Total NSD 10 Total NSD
Force (kips)
Force (kips)

0 0

-10 -10

-20 -20
-2 -1 0 1 2 -2 -1 0 1 2
Displacement (in) Displacement (in)

a) With Gap Spring Assembly b) Without Gap Spring Assembly

Figure 16: Experimental Results from Displacement Controlled Tests of NSD

Figure 17 shows the results obtained from Krypton for the most significant points of the
NSD. As expected, point A which is the displacement imposed on the NSD, point B and point
E have approximately equal displacements while point D is double in magnitude and opposite
in sign from the displacement of point B. All displacements below are given as the relative
displacements between their absolute displacements and the displacement of the shake table.

18
Apostolos A. Sarlis, Dharma Theja R Pasala, Michael C. Constantinou, Andrei M. Reinhorn, Satish Nagarajaiah
and Douglas Taylor

4 4
Point-A Point-B
Displacement (in)

Displacement (in)
2 2

0 0

-2 umax=-2.49in -2 umax=-2.51in

-4 -4
0 100 200 300 0 100 200 300
10 4
Point-D Point-E
Displacement (in)

Displacement (in)
5 umax=5.05in 2

0 0

-5 -2 umax=-2.45in

-10 -4
0 100 200 300 0 100 200 300
Time(sec) Time(sec)

Figure 17: Experimental results from Krypton data acquisition system for displacement controlled tests

8 VERIFICATION OF ANALYTICAL EQUATIONS AND COMPUTATIONAL

MODELS
Figure 18 compares all the analytical results described earlier with the results obtained
from the experiments. With the exception of the dynamic model comparisons, all other results
are shown for static tests conducted at 0.02Hz frequency. Analytical results are plotted for the
nominal properties of table 1 in order to demonstrate the additional advantage that the NSD
exhibits of small variability of its properties. The absence of friction elements and the absolute
dependence of the NSD behavior on the pre-loaded spring and geometry make the NSD
behavior to be predictable at the design level. Note that inclusion of the flexibilities of the top
frame explains the smaller stiffness in the gap spring assembly active region- observed in the
experimental results- the dynamics explain the higher “negative force observed during the
high frequency tests while the hysteretic element accurately follows the hysteresis trend
observed in experiments. Based on these comparisons one can though conclude that the basic
analytical equations derived in section 3 are sufficient for accurately capturing the NSD
behavior assuming proper design was followed.

19
Apostolos A. Sarlis, Dharma Theja R Pasala, Michael C. Constantinou, Andrei M. Reinhorn, Satish Nagarajaiah
and Douglas Taylor

2 4
Experimental Experimental
Analytical-Basic Analytical-Basic
1 2
Force (kips)

Force (kips)
0 0

-1 -2

-2 -4
-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3
Displacement (in) Displacement (in)
2 4
Experimental Experimental
Analytical with Hysteresis Analytical with Hysteresis
1 2
Force (kips)

Force (kips)

0 0

-1 -2

-2 -4
-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3
Displacement (in) Displacement (in)
2 4
Analytical with flexibilities Analytical with flexibilities
Experiemental Experiemental
1 2
Force (kips)

Force (kips)

0 0

-1 -2

-2 -4
-4 -2 0 2 4 -4 -2 0 2 4
Displacement (in) Displacement (in)

20
Apostolos A. Sarlis, Dharma Theja R Pasala, Michael C. Constantinou, Andrei M. Reinhorn, Satish Nagarajaiah
and Douglas Taylor

3 4
Analytical w/ Dynam Analytical w/ Dynam
2 Experimental Experimental
2
1
Force (kips)

Force (kips)
0 0

-1
-2
-2

-3 -4
-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3
Displacement (in) Displacement (in)
Figure 18: Verification of analytical derivations

9 CONCLUSIONS
Based on the material presented above the following may be concluded:
 A novel Negative Stiffness Device (NSD) has been developed. Proof of concept testing
was performed at the NEES facility at the University at Buffalo. The NSD was shown to
generate negative stiffness and reduce the base shear of a structure.
 Installing a Negative Stiffness Device is an efficient method of reducing the "apparent
strength/stiffness" of a structure’s floor/story without affecting the vertical stiffness or
stability limits of the structure.
 NSD operates with only minor energy dissipation and can be considered to be nonlinear
"elastic". Any minor permanent deformations that the NSD might exhibit are eliminated
by the gap-spring assembly.
 Basic NSD properties that determine its behavior are adjustable and therefore the NSD can
generate a wide range of behaviors by simple adjustments that can be performed even
after the installation of the device.
 NSD generates reliable and predictable behavior which can be accurately predicted. This
is due to the mechanical nature of the device, consisting of links and springs of well
known geometry and of predictable mechanical properties. Note that the comparisons of
the analytical versus experimental results are done using the nominal spring properties in
order to specifically demonstrate this advantage. No testing was conducted on the
individual components (springs).
 Although the NSD behavior might be affected by many factors, the influence of those
factors is minor. The NSD behavior can be very accurately predicted by the equations of
section 3 and modeling of section 4. Additional complex derivations presented herein help
in obtaining understanding of the complete mechanics of the system and provide the
designer with additional confidence in using the NSD in practical applications.
 Complete details on the experimental program as well as further analytical and
computational work will be enclosed in [5] and [6].

21
Apostolos A. Sarlis, Dharma Theja R Pasala, Michael C. Constantinou, Andrei M. Reinhorn, Satish Nagarajaiah
and Douglas Taylor

REFERENCES
[1] Reinhorn A.M., Viti S., and Cimellaro G.P. (2005), “Retrofit of Structures: Strength
Reduction with damping enhancement.”, Proceeding of the 37th UJNR Panel Meeting on
Wind and Seismic Effects, Tsukuba, Japan.
[2] Viti S., Cimellaro G. P., and Reinhorn A. M. (2006), Retrofit of a Hospital through
Strength Reduction and Enhanced Damping. Smart Structures and Systems, 2(4), 339-
355.
[3] Nagarajaiah S., Reinhorn A. M., Constantinou M. C., Taylor D., Pasala, D. T. R. and
Sarlis, A. A. (2010), “True Adaptive Negative Stiffness: A New Structural Modification
Approach for Seismic Protection”. 5th World Conference on Structural Control and
Monitoring, Tokyo, Japan, July 12-14.
[4] Iemura H. and Pradono M.H. (2009), “Advances in the development of pseudo-
negative-stiffness dampers for seismic response control,” Structural Control and Health
Monitoring, 16(7-8), 1545-2255.
[5] Sarlis A.A., Constantinou M.C, Reinhorn A.M, Pasala, D.T.R, Nagarajaiah S. and
Taylor D. (2011), “Negative Stiffness Device- Part 1: Analytical Framework and
Experimental Validation” - MCEER-11-XXXX, Multidisciplinary Center for
Earthquake Engineering Research, Buffalo, NY.
[6] Sarlis A.A., Constantinou M.C, Reinhorn A.M, Pasala, D.T.R, Nagarajaiah S. and
Taylor D. (2011) , Negative Stiffness Device- Part 2: Shake table testing and Proof of
Negative Stiffness Concept” MCEER-11-XXXX, Multidisciplinary Center for
Earthquake Engineering Research, Buffalo, NY.
[7] Molyneaux W. G. (1957), “Supports for vibration isolation”, ARC/CP-322,
Aeronautical Research Council, Great Britain.
[8] Platus D.L., Negative-stiffness-mechanism vibration isolation systems, Proceedings of
SPIE—The International Society for Optical Engineering, Vibration Control in
Microelectronics, Optics, and Metrology, San Jose, CA, USA, Vol. 1619, 1992, pp. 44–
54.
[9] Platus D.L., Negative-stiffness-mechanism vibration isolation systems, Proceedings of
the Opto-mechanical Engineering and Vibration Control Conference, SPIE—The
International Society for Optical Engineering, Denver, CO, USA, Vol. 3786, 1999, pp.
98–105.
[10] Platus D.L. (1993), “Vibration Isolation System”, US Patent No. 5178357, Washington
DC: US Patent and Trademark Office.
[11] Platus D.L. (1994), “Vibration Isolation System”, US Patent No. 5310157, Washington
DC: US Patent and Trademark Office.
[12] Platus D.L. (1994), “Damped Vibration Isolation System”, US Patent No. 5370352,
Washington DC: US Patent and Trademark Office.
[13] Platus D.L. (1995), “Vibration Isolation System”, US Patent No. 5390892, Washington
DC: US Patent and Trademark Office.
[14] Platus D.L. and Cunningham P. J (1996), “Vibration Isolation System”, US Patent No.
5549270, Washington DC: US Patent and Trademark Office.

22
Apostolos A. Sarlis, Dharma Theja R Pasala, Michael C. Constantinou, Andrei M. Reinhorn, Satish Nagarajaiah
and Douglas Taylor

[15] Platus D.L. and Durran D.A (1997), “Vibration Isolation System”, US Patent No.
5669594, Washington DC: US Patent and Trademark Office.
[16] Platus D.L. (2004), “Vibration Isolation System”, US Patent No. 6676101B2,
Washington DC: US Patent and Trademark Office.
[17] Carrella A., Brennan M.J., Waters T.P. and Shin K. (2008) On the design of a high-
static-low-dynamic stiffness isolator using linear mechanical springs and magnets,
Journal of Sound and Vibration, 315(3), 712-720
[18] Lee C. M., V. Goverdovskiy and A. Temnikov (2007), “Design of springs with
‘negative’ stiffness to improve vehicle driver vibration isolation”, Journal of Sound and
Vibration 302 (4-5) (2007) 865–874. URL http://dx.doi.org/10.1016/j.jsv.2006.12.024
[19] H. Iemura, O. Kouchiyama, A. Toyooka and I. Shimoda (2008). “Development of the
Friction-Based Passive Negative Stiffness Damper and its Verification Tests using
Shaking Table”.The 14th World Conference on Earthquake Engineering, October 12-17,
2008, Beijing, China
[20] Wolff E.D. and Constantinou M.C. (2004). “Experimental Study of Seismic Isolation
Systems with Emphasis on Secondary System Response and Verification of Accuracy
of Dynamic Response History Analysis Methods”, Technical Report MCEER-04-0001,
Multidisciplinary Center for Earthquake Engineering Research, State University of New
York at Buffalo, Buffalo, NY, USA.
[21] Kasalanati A. and Constantinou M.C. (1999). “Experimental Study of Bridge
Elastomeric and Other Isolation and Energy Dissipation Systems with Emphasis on
Uplift Prevention and High Velocity Near Source Seismic Excitation”, Technical Report
MCEER-99-0004, Multidisciplinary Center for Earthquake Engineering Research, State
University of New York at Buffalo, Buffalo, NY, USA
[22] Bracci J.M., Reinhorn A.M. and Mander J.B. (1992). “ Seismic Resistance of
Reinforced Concrete Frame Structures Designed Only for Gravity Loads: Part I- Design
and Properties of a One- Third Scale Model Structure”, Technical Report NCEER-92-
0027, National Center for Earthquake Engineering Research, State University of New
York at Buffalo, Buffalo, NY, USA
[23] Structural Engineering and Earthquake Simulation Laboratory (2004). “Laboratory
Manual” SEESL, Buffalo, NY USA. Available online:
htto://nees.buffalo.edu/docs/labmanual/html/
[24] Computers and Structures Inc. (2007).”SAP2000: Static and Dynamic Finite Element
Analysis of Structures (Version 11.0.2) Analysis Reference Manual”, Computers and
Structures, Inc., Berkeley
[25] Sivaselvan M.V. and Reinhorn A.M. (2001). “Hysteretic Models for Deteriorating
Inelastic Structures”, ASCE Journal of Structural Engineering, 126(6) 633-640.

23