You are on page 1of 30

Evaluation of Assumptions Used in Engineering

Practice to Model Buildings Isolated with Triple


Pendulum(TM) Isolators in SAP2000

1
NYA Isolation Designs
NYA has designed 12 isolated
buildings in the last 15 years

2
Several efforts to study and promote isolation

3
Why should we study this?
Recent Friction Pendulum buildings in the Bay Area alone:
Stanford Hospital
SF General Hospital
Mills Peninsula Hospital
UCSF Regenerative Medicine
Washington Hospital

=> Several billion dollars of construction


Many others around the world
All riding on a few boxes checked in the analysis program!

No recent earthquakes to verify (US) design assumptions

Because we can
E-defense is a great data set for this

4
No Clear Design Direction
No clear basis for design assumptions being
used:
Limited references/design guides
Limited discussion in code
Based on Engineering judgment
Carried from past projects by on convention
and expectations of regulatory agencies

5
Prior Work
Largely from a research perspective
Uses research analysis programs
(OpenSees, 3DBASIS, etc.)
Focus on bearing models
Focus on best fit rather than
design assumptions
Does not reflect what designers
are doing now
Small scale

6
Blind Analysis Contest
Pretty simple predict a few response parameters
using only basic building info and input excitation

We were very interested in this test:


Similar to Stanford Hospital design situation -
only prototype test data and building info given
Similar bearings
Similar 2-way moment frame system
Similar level of excitation
Already had an ETABS model setup

Strong competition:
~15 teams from research and industry
1st prize gets trip to Japan (during ski season!)

Different from design predict exact value

7
Blind Analysis Contest - Results
NYA landed in 4th place
Parameter Error From Test (%)
Not bad
Used typical design Isolator Displacement 11%
assumptions
Base Shear 14%

Drift 1 to 15%

Floor Acceleration 10 to 50%

8
Blind Analysis Contest What Now?
We were in the middle of design for Stanford Hospital

Does it matter?
Not really, still conservative with Phi factors and OSHPD requirements

Should we change anything?


What to change? We thought we did it right

Why were we wrong?


Not sure
Started looking at all of the assumptions that we take for granted in design

Research paper was born

9
Format of the Paper
Pretty basic :
Re-analyze the E-defense building using methods and
assumptions from common practice
consider different assumptions and compare to actual test data
Consider multiple ground motions instead of just one for B.A.
Contest
Compare to more detailed OpenSees model for a reality check
and to see if it is even possible to do any better

10
What is the State of Practice?
Seems to be minor variations of the same overall process

SAP or ETABS most common platform

Common Assumptions/Methods:
NLRHA for large projects
X-Y (possibly Z) excitation
Linear Elastic frame and shell elements
Damping varies: 2% and 5%, 2% all, 1% and 2%, etc
Fast Nonlinear Analysis (FNA)
Isolator model using built in SAP elements - Bilinear, Parallel, or Series
Rigid Diaphragms
Isolator P-delta moments superimposed on NLRHA

11
State of Practice Applied to this Study
Item Modeling Assumption Notes
Fast Nonlinear Analysis See section Selection of
Analysis Method
(FNA) Analysis Method
Acceleration Input and
X-Y-Z
Analysis Directions
Linear-elastic frame and shell
Test building was known to
Model Elements elements and nonlinear link
remain linear elastic
elements
2% in all modes, 0% in first 3
See section Selection of
Damping (isolated) modes using Modal
Damping
Overwrites
Determined through a
Number of Modes used in sensitivity analysis - see
200
Analysis section Selection of Number
of Modes
Bilinear model (Sarlis and See section Isolator
Isolator Analytical Model
Constantinou 2010) Modeling
11.0% on outer surfaces, Determined from shake table
Isolator Friction 1.7% on inner surfaces (Dao sinusoidal test see section
et. al. 2013) Isolator Modeling

Global P- Effects Not included See section Limitations


Only Vh component captured
Isolator Local P- Moments Partially represented - see section Isolator
Modeling
To ensure vertical load on
Taken as From Loads option isolators is compatible with
Mass Source
in program horizontal mass of
superstructure

Base reactions adjusted to


Applied to the structure using reflect measured reactions
Inital Dead Load in Analysis
ramp function see section Development of
Model

Structure Weight Calibrated to test results

Rigid Diaphragms Used at each floor level See section Results

12
FNA vs. DI
Fast Nonlinear Analysis FNA and Direct Integration (DI) can give very
different results in SAP/ETABS for isolated structures
NYA discovered this simultaneously with the release of a Constantinou paper
Issue is Damping Leakage that introduces artificial damping to the isolators
Perception in practice is that DI is like broccoli and exercise if is more
difficult it must be better for you.
FNA allows direct specification of isolator contribution to damping matrix and
is much faster making it the clear choice for isolated structures in SAP/ETABS

Isolator Displacement: MCE Hazard, X-direction , ~2% damping


16

12 FNA
Displacement [in]

8
DIM
4

-4

-8

-12

-16
0 5 10 15 20 25 30 35 40
Time [sec]

13
Isolator Modeling
Several Isolator Models Published: Isolator Friction:
Bilinear using single isolator Typically taken from isolator
embedded in SAP prototype tests.
Parallel model using 2 SAP We took them from sinusoidal
isolator elements tests on the table more accurate
Series model using multiple SAP Used first cycle zero displacement
isolator elements with gap friction
elements
0.2
FP1
FP2 0.15
Total FP
. 0.1

0.05
V/W

0
-30 -20 -10 0 10 20 30
-0.05

-0.1

-0.15

-0.2
Displacement (in)

14
Number of Modes in the Analysis
Important but can be overlooked
FNA requires a significant number of modes to capture dynamics
More modes needed for more detailed parameters
Not easy to calculate, recommend doing a sensitivity study

15
Damping
FNA allows two methods of damping:
Mass and Stiffness Proportional
Direct specification of each mode

Issues with Mass and Stiffness Proportional damping identified in previous


work

Direct specification is easy to implement and intuitive:


Less damping in first few isolated modes, more damping in remaining
modes of superstructure

We found that 0% in first 3 (isolated) modes and 2% in superstructure modes


gave the best fit (for floor spectra)
This are probably lower damping than normal:
Few nonstructural components
No moat covers
Still likely less than 2% and 5% used in some designs

16
OpenSees Model
Sap results were also compared to an OpenSees model

OpenSees model:
More detailed
Had more work put into it (PHD thesis vs. Design Engineer after hours)
Used totally different methods and assumptions
Shared little with SAP model
Provided a second check beyond the test data

17
Results Modal Properties
Very good agreement without diaphragm constraints in SAP (~1-3% error)
Reduced accuracy with diaphragm constraints (~10% error)
Diaphragm constraints required for realistic floor spectra

Mode Test Building SAP2000 with SAP2000 without OpenSees


diaphragm diaphragm
constraints constraints

1 Translation Y 0.678 sec 0.612 sec (-9.7%) 0.657 sec (-3.1%) 0.687 sec (1.3%)

2 Translation X 0.652 sec 0.598 sec (-8.3%) 0.644 sec (-1.2%) 0.666 sec (2.1%)

18
Results Isolator Displacement
Good agreement (~10% error)
OpenSees only slightly better (likely due to more detailed friction model)
Peak Isolator Displacements (cm)
Record Test SAP2000 OpenSees
100IWA 37 (-2.7%) (5.4%)
80TCU 53 (-11.3%) (1.9%)
70LGP 48 (-8.3%) (2.1%)
100TAK 53 (0.0%) (11.3%)

19
Residual Displacements
Residual displacements had no effect on peak displacements
Has been a topic in OSHPD review

10cm
initial
offset

20
Results Base Shear
SAP had good agreement (~10-15% Error)

Test Base Shear SAP2000 Error OpenSees Error


Record
X Y X Y X Y
100IWA [0.12W] [0.14W] (7.4%) (7.9%) (-5.0%) (10.4%)
80TCU [0.16W] [0.13W] (-3.3%) (11.0%) (-5.4%) (-2.8%)
70LGP [0.08W] [0.17W] (10.7%) (17.8%) (2.6%) (-0.8%)

100TAK [0.19W] [0.14W] (17.1%) (11.3%) (-6.2%) (-12.5%)

21
Results Hysteresis Loops
Not used for design but helpful to verify model is performing as expected
Accuracy of both models was degraded with vertical excitation:
Without Vertical Excitation:

With Vertical Excitation:

22
Results Drifts
Building was very stiff so drifts were small (0.2%) and ~1/10th of code drift limit
Small drifts more affected by measurement errors
Not really affected by vertical excitation

23
Results Floor Accelerations
Not usually used for design now, but gaining popularity with PBD
Records without Vertical Excitation:

Records with Vertical Excitation:

24
Results Floor Accelerations
Records without Vertical Excitation:

Records with Vertical Excitation:

25
Compare to Blind Analysis Contest
Some improvement for blind analysis ground motion:

Blind This Study


Analysis
Parameter Error From Error From
Test (%) Test (%)
Isolator 11% 5%
Displacement
Base Shear 14% 9%

Displacements 3-11% 2-13%

Floor 1 to 50%* 3 to 19%


Acceleration

26
Conclusions
SAP results compared well with test data and the OpenSees Model

A few minor changes recommended

A few potential issues identified

Would not expect a more detailed model to perform significantly better

Appropriate for design

Accelerations still difficult to predict

Likely significantly more accurate than state of practice for fixed base design

27
Recommendations For Design
SAP or ETABS
Use FNA
Modal damping:
Small damping in isolated modes (~0-1%)
Slightly larger damping in superstructure modes (~2-3%)
Bilinear isolator model
Perform sensitivity study to determine number of modes
Use first cycle isolator friction (for nominal properties)
Do not include global P-Delta in analysis (SAP only)
Be careful with rigid diaphragms
Local isolator moments can be superimposed on NLRHA results

Independently verify isolation analysis! (BISPEC, 3D-BASIS, etc.)

28
Future Work
Investigate ETABS Triple Pendulum element in ETABS
Look at effects of friction property modification factors
Look at how the code minimum equations relate to this

29
Thank You

Questions/Comments

30

You might also like