seismic performance in urban regions (spur) and prototype neesgrid implementation

Seismic Performance in Urban Regions (SPUR) and

Prototype NEESgrid Implementation

Roger L. King, Gregory L. Fenves, Jacobo Bielak, Tomasz Haupt

Bozidar Stojadinovic, Joerg Meyer

Antonio Fernandez, Jaesung Park, Prashant Chopra, Koel Das, Aysegul Askan, George Petropoulos, A. Kalayanasundaram,

M. Khotournenko, A. Chilkuri, N. Ammari

NSF NEES Awardees MeetingSeptember 11-12, 2003

Outline of Presentation

• What is SPUR?• Value of simulation and integration• Structural performance• Earthquake ground motion modeling• Site response and soil-structure interaction• Visualization• SPURport – web portal and demo

An NSF ERC Problem-Centered Research Flow Process

SPUR, a Distributed Simulation Framework for Seismic Performance for Urban Regions

• Advance the state-of-the-art in simulating the effects of a major earthquake on an urban region.

• Integration of earthquake ground motion modeling with modeling of structural and infrastructure systems using advanced computational and visualization methods.

• The goal is to develop tools that will ultimately permit damage estimates based on best available information.

• Generalize discipline specific approach towards system-level research.

PBEE

SSFI

Basin Effects

Large Scale Viz

Fundamental Research• Computational seismology• SSI with DRM• PBEE Performance measures• Spatial distribution of performance• Large scale viz.

Enabling Technologies• Archimides• OpenSees• NEESgrid• System integration

Middleware

NEESHPC Resource

Viz

Systems-level Applications• Loss estimation• Tools for decision makers• Scenarios for planning

MicrozonationPlanning

Emerg. Resp.Education

Loss Estimate

SPUR Strategic Vision

Leveraging Existing NSF Research for System-Level Problems

ground motion data

(CMU)precomputed simulations

ground motion data

(CMU)precomputed simulations

rendering + portal(UCI) (MSU)

immersive rendering &web-based portal

rendering + portal(UCI) (MSU)

immersive rendering &web-based portal

structural performance

(PEER - Berkeley)precomputed or online

simulations

structural performance

(PEER - Berkeley)precomputed or online

simulations

SPURport: The Grid-Based Portal for SPUR

• Provide earthquake community with collaborative environment for research on Seismic Performance of Urban Regions and training of future earthquake engineers

• Develop NEESgrid application of databases, computation, visualization, using distributed Grid-accessible resources; demonstrate ability to use NEESgrid resources at any location

• Opportunity to apply NEESgrid software releases to a substantial application and provide NEESgrid developers feedback

• Add to simulation capability of NEES

Strike-Slip Fault Simulation ModelPeak Ground Displacement

Peak Ground Velocity

Thrust Fault Simulation ModelPeak Ground Displacement

Peak Ground Velocity

Structural Models for Regional Simulation

ObjectiveEvaluate regional distribution of Engineering Demand Parameters (EPD)SDF Model u

u

u

uu

u

u

u

Grid point

- “Hydra” model- Multiple parameters and multiple orientations- Constant strength analysis- Constant ductility analysis

Simulation Tool – OpenSees (PEER software framework for simulation)

Building Models

ShearBeamModel

GeneralizedFrameModel

BuildingModels

Constant Ductility Analysis for Strike Slip Fault

SAC 9-story OpenSees Model

Column Fiber Section

Beam Fiber Section

Leaning columnsfor P- effects

Distributed Plasticity Beam-Column

LA 10%/50 year

Regional Distribution of SAC 9-story EDP

Roof drift ratio

Max Plastic Rotation

Max Story Drift Ratio

PGV

Regional Distribution of SAC 20-story EDP

Max Plastic Rotation

Roof drift ratio

Max Story Drift Ratio

Story 22

Story 15

PGV

Calibration of Shear Beam Model

Pushover Analysis of Frame Model Shear Beam Model

Story Force-Deform.

Comparison of Floor Displacement

Generic Frame Model

Floor 1 Floor 2 Floor 3

Floor 1 Floor 2 Floor 3

Shear Beam Model

Regional Distribution of EDP, 3-Story

Roof Displacement

SAC Frame Model Shear Beam Model Generic Frame Model

SAC Frame Model Shear Beam Model Generic Frame Model

Max Story Drift

Computational Challenge

SAC 9-Story Simulation– 306 DOFS, 1800 time steps.

– Approx 4 min. per grid point.

– 25,281 grid points.

– 70 days in single processors.

Constant Ductility Analysis– 28 parameter combinations.– Considering 8 orientations.– 25,281 grid points.– 5 iterations in average.– One million non-linear analysis

of SDF system per parameter.

OpenSees Applications for Parallel Computing

Parallel simulation applications built with OpenSees API and MPI API can be implemented using NEESgrid resources.

Model Domain

Element Material Pattern

Analysis Solution Procedures

OpenSeesFramework

wake

Packaging Send output

wait

Un-packaging

Receive Ground Motion

Structure parameter

Slave Node

Job Queue

Wait Queue

Ground Motion Data

Recording output

Receive outputUn-packaging

SendGround Motion

Structure parameter

pack

agin

g

ProcessManage

Master Node

Producer-Consumer Approach

Northridge Earthquake Mainshock

(USGS)(USGS)

Rupture Model

Dep

th (

km

)

Wald et al. (1996)Wald et al. (1996)Strike=122 (S58E), Dip=40 (S32W), Rake=101Strike=122 (S58E), Dip=40 (S32W), Rake=101

USGSUSGS

Snapshots of Surface Velocity

Peak ground velocity Peak ground velocity ((USGS)USGS)

Wavelength-Adaptive Tetrahedral Mesh

(~ 100 million elements; 3 hrs)

Site Response - SSICross Section

Reduced Domain

Cascading (3 models)

Domain Reduction Method

Run on different platforms

Free-Field ResponsePeak values of displacement and velocity

Analysis Region

Observation Point(1280, 3000)

Observation Point

Velocity (m/s)

|FT|

Peak Ground Velocity in Region of Interest

Vm

ax

800 m/s

100 m/s

200 m/s 40 m

Spatial variability of ground motion

3D vs 1D

Random City Model

Alternatively, populate with sets of uniform buildings

Structural Models and Distributions

Influence of Different Structural Distributions On Maximum Ground

Velocity (EW)

Free-field 3 1.0-Hz Buildings R-City

Low period structures

Random City Simulation

High-end VisualizationNew Algorithms for Rendering:• Efficient 3-D mesh decimation• More efficient than edge-

collapse• Various levels of detail

TetFusion QTetFusion

SPURport Architecture

Tele-presence

NEESpop (middleware)

MSU extensions(Enterprise Computational System)

ApacheTomcat

JetSpeedChef

NEESgrid services

DBMS(postgress)

EJB container(JBoss)

OGSI(globus 3.0)

ECSapplicationstreaming

device driver

Data Controller

Data streaming andchannel management

Authentication andauthorization

SPURapplet

Request

Data

NSDS

Data andMetadata

Collaboration

ERCat Mississippi State PSC NCSA

OpenSees OpenSeesGround Motion

DataStruct. Resp.

Data

Ground MotionData

Front End

Back End

SPURport Functionality

EarthquakeModel

Inventoryof Structures

view data

extract data

StructureModel

select or definea structure

(set parameters)

select location

run simulation

select or definean inventory

run simulation (future)

view data view data

Earthquake Model

Structural Model

Inventory Model

Spatial Response Data

Individual Structure Response