Hybrid Testing 2.0Quake Summit 2012

Thomas Frankie frankie2@illinois.edu

Michael Bletzinger mbletzin@illinois.edu

Hybrid Simulation @ MUST-SIM

Project Analytical Model

NEESR-SG: Seismic Simulation and Design of Bridge Columns under Combined Actions and Implications On System Response, David Sanders (University of Nevada, Reno), Abdeldjelil Belarbi (University of Missouri –Rolla) , Amr Elnashai (University of Illinois), Jian Zhang (University of California, Los Angeles), Shirley Dyke (Washington University)

NEESR-SG: Controlled Rocking of Steel-Framed Buildings with Replaceable Energy Dissipating Fuses, Gregory G. Deierlein, Sarah L. Billington, Helmut Krawinkler (Stanford University), Jerome F. Hajjar (University of Illinois)

NEES (Shared Use): Performance Evaluation of Semi-rigid Steel Frames Using Hybrid Simulation, Amr Elnashai (University of Illinois), Zifa Wang (Institute of Engineering Mechanics, China Earthquake Administration)

NEESR: Framework for Development of Hybrid Simulation in an Earthquake Impact Assessment Context, Bill Spencer, (University of Illinois), et. Al.

Hybrid Simulation Case Examined

4-span curved bridge RC Piers of various lengths Two scales of experimental

specimens Multi-directional earthquake

loading 6DOF control

NEESR-SG-0530737

Video

Hybrid Simulation Procedure

Computational component

{u}

{F}

Measure forces

gxSimulation Coordinator

Experimental componentFz

My

MyFy

FxMx

Target Disp.

Measured

forces

Calc Forces

Preparing for the CABER Three Pier Test

General Solution Strategy CABER Testing Evolution

– Analytical– Analytical Modularized– Analytical and 1 small scale– All small scale– Large Scale

Coordinate Transformation and Scaling

Coordinate systems, notation, and units vary by site and program Experimental: Cartesian to actuator space Scaling and similitude relationships vary for different parameters Currently:

– Plugin loads and uses transformation functions– Calculations occur at full scale– (Commands) x (Displacement scale factors) sent to modules – (Measured forces and moments) / (Scale factors) used in the PSD algorithm.

X

ZY

LBCB (experimental)

X

Y

Z UI-SimCor (overall model)

[displacement , rotation , force , moment]Analytical: [1 , 1 , 1 , 1]

Experimental: (1/3 scale, N-mm to lb-in)[1/3/25.4 , 1 , 1/9/4.448222e3 , 1/27/4.448222e3/25.4]

Mapping from Model Space to Experiment Space

Z

Y

X

Z

YX

Control points passed through SimCor Displacement-based 6DOF, derivation of transformation challenge At times, use more control points than locations with readings/data

Substructure Failure

Stroke and force capacity of equipment Failure of experimental specimen Diverging solution in numerical model Many hybrid tests don’t truly “finish”- Use updated stiffness data to provide simulated force feedback

- Trigger “end of life” scenario

- Back up analytical models to replace experimental specimens

Data synchronization

Capturing control (commands), instrumentation (sensors) and visual (photo) data

Necessary for– Viewing and understanding of behavior during a test– Post-processing and development of meaningful

complementary data and media Solution for PSD tests

– PSD hybrid tests enables data collection at brief hold periods

– Data is inherently synchronized because specimen is not moving

A triggering service is needed to inform all

acquisition devices

Hybrid Tests Will Pause

Problem Example Recovery

Internet link failure

Response from substructure module times out.

Re-establish the link. Continue test from failure point

Control system limit fault

Actuator force exceeds limits

Pause simulation.Fix the conditions of the fault. Continue test from fault point

Acquisition system failure

Camera freezes. Pause simulation.Fix or replace the system.Continue test from failure point

Application failure

Krypton DMM crashes Pause simulation.Restart applicationContinue test from failure point

Test takes to long

Operators can only work until midnight

Store simulation state.Continue simulation the next day.

Why Hybrid Testing 2.0

Hybrid testing experiences show the need for a general upgrade to UI-SimCor

LbcbPlugin is a prototype for this upgrade– MATLAB/Java application developed at the MUST-

SIM facility.– Site specific features.

• Robust Communications• Test specific customizations• Trigger server• Menu driven configuration management.

– Derived from UI-SimCor

LbcbPlugin Communications

Multi-Threaded– Each communication link has

its own Java thread Pause and resume logic

– Communication errors detected through timeouts• triggers test pause• Test resumes once link is re-

established

– Limit faults and other errors handled through decline processing• Tripped limits and user decline

directives trigger test pauses

LbcbPlugin Customization

Customization Plugin Type

Purpose

Command & Response Transformation

Custom transformations for commands and responses exchanged with UI-SimCor

Derived DOF Customize the control loop by adding additional commands based on current responses

LbcbPlugin Triggering

Runs a triggering server– Acquisition devices connect to

the server Response monitoring detect

problems Two levels of triggering Piggyback aggregated Data on

trigger message

Upgrading UI-SimCor

Requirements– Scalability– Flexible Customization– Robust Communications

New version will be a Java application.– Optional MATLAB components– LbcbPlugin features will be merged with UI-SimCor– Wizards and menus will serve as configuration

management GUIs.– Client implemented in other languages

• LabVIEW, Visual Basic

TC

P/I

P N

etw

ork

Original UI-SimCor Architecture

Stiffness Evaluation

Static Equilibrium

Dynamic Equilibrium

Main Routine

Object n of MDL_RF class

MDL n

Simulation Monitor

Clie

nt

Object 1 of MDL_RF class

MDL 1

Simulation Monitor

Clie

nt

AP

I

Component n

Ser

ver

Component 1

Ser

ver

AP

I

FEDEASLab

Vector 2

DO

F M

appi

ng

Disp.

Force

Objects of MDL_AUX class

AUX

Clie

nt

Equipment

AP

IDAQ

CameraSer

ver

Simulation Control

Command

Measurement

Command

Measurement

Scalability Example – CABER Simulation

Scalability Example – Caber Small Scale

TCP Msgs

TCP Msgs

Scalability Example – CABER Large Scale

TCP Msgs

TC

P

Msg s

TCP Msgs

TCP Msgs

Scalability Example – CABER Large Scale with Backup Modules

TCP Msgs

TCP Msg

s

TCP Msgs

TCP Msgs

TC

P M

sgs

Summary

Challenge Current Workarounds Upcoming Features

Model tailoring for site and specimen issues

LbcbPlugin customizationUI-SimCor Transformation & Scaling

Wizards and menus in new UI-SimCor. MATLAB code customization plugins

Tests will pause LbcbPlugin robust links & decline processingUI-SimCor restart feature

Robust communications in new UI-SimCor

First substructure failure ends test

Simplify analytical model Model switching during simulations. Additional research needed

Data synchronization and aggregation

LbcbPlugin Trigger Server

Trigger Server

Participation Needed

Need stake holders to tryout upcoming versions– Help test and document development versions– Implement clients in other languages– Become informal advisory committee– RC Frames project has already been “volunteered”

Informal Advisory Committee

Member University

Oh-Sung Kwon University of Toronto

Amr Elnashai University of Illinois

Bill Spencer University of Illinois

Questions?