Structural Health Monitoring Using Statistical Pattern Recognition

Presented by Charles R. Farrar, Ph.D., P.E. and Hoon Sohn, Ph. D.

Los Alamos Dynamics Structural Dynamics and Mechanical Vibration Consultants

Structural Health Monitoring using Statistical Pattern Recognition

Overview of the Course Summarize the rapidly evolving field of structural health monitoring.
Summarize the historical developments of this technology. Provide overview of current methods. Show real world application of this technology. Identify the limitations of the current technology. Present cutting edge statistical tools for diagnosis. Discuss current and future research directions.

Course Theme: Structural Health Monitoring is a problem in statistical pattern recognition.

Los Alamos Dynamics Structural Dynamics and Mechanical Vibration Consultants

Course Demo 2

Structural Health Monitoring using Statistical Pattern Recognition

Where Does Structural Health Monitoring Fit In The Big Picture

Usage Monitoring Structural Health Monitoring System assessment model Damage Prognosis

Actual loading and operating conditions

System

Predictive model

Response measurement Input measurement 1. Instrumentation 2. Data management 1. Modeling & simulation 2. Data interrogation

1. Modeling & simulation 2. Data interrogation 3. Embedded processing

Future Loading Estimation Predictive loading model

1. Data interrogation

Los Alamos Dynamics Structural Dynamics and Mechanical Vibration Consultants

Course Demo 3

Structural Health Monitoring using Statistical Pattern Recognition

Process of Structural Health Monitoring Vibration-based damage detection is part of the more general process of Structural Health Monitoring The Structural Health Monitoring process includes: 1. Operational evaluation of the structure 2. Data acquisition and cleansing 3. Feature extraction and information condensation 4. Statistical model development

Los Alamos Dynamics Structural Dynamics and Mechanical Vibration Consultants

Course Demo 4

Structural Health Monitoring using Statistical Pattern Recognition

Rotating Machinery Application

ax Amp .38

0 0 0 0 0 18-A

01-APR

21-MAR-96

21-MAR-96

21-MAR-96

20-MAR-96 0 1000 2000 3000 Frequency in Hz 4000 5000

Before BeforeBearing BearingReplacement Replacement Spectral response of machine vibrations before (bottom trace) and after bearing replacement Engineers at Intels Fab-11 plant measure vibrations on a vacuum blower motor

Los Alamos Dynamics Structural Dynamics and Mechanical Vibration Consultants

Course Demo 5

Structural Health Monitoring using Statistical Pattern Recognition

Early Work on Offshore Structures

Offshore Industry spent millions of dollars during the 70s and 80s in an effort to launch practical damage detection and health monitoring of offshore platforms Numerous examples in the literature of numerical modeling efforts as well as scale-model and full-scale experiments Many practical problems were encountered:
Machine noise Non-uniform inputs Hostile environment for instrumentation Marine growth Changes in foundation with time

Modal frequencies can be insensitive to many of the damage types the offshore industry is interested in
Course Demo 6

Los Alamos Dynamics Structural Dynamics and Mechanical Vibration Consultants

Structural Health Monitoring using Statistical Pattern Recognition

Overview of Aerospace Applications Aging aircraft Rotorcraft Reusable launch vehicles:

Space shuttle X-33 DC-XA

International space station & related truss test beds MIR space station

Damage to 1988 Aloha Airlines flight motivated the development of an FAA Aging Aircraft Center at Sandia National Laboratory

Los Alamos Dynamics Structural Dynamics and Mechanical Vibration Consultants

Course Demo 7

Structural Health Monitoring using Statistical Pattern Recognition

1. Operational Evaluation Operational evaluation begins to answer questions regarding implementation issues for a structural health monitoring system.
Provide economic and/or life-safety justifications for performing the monitoring. Define system-specific damage including types of damage and expected locations. Define the operational and environmental conditions under which the system functions. Define the limitations on data acquisition in the operational environment.

Operational evaluation will require input from many different sources (designers, operators, maintenance people, financial analysts, regulatory officials)
Los Alamos Dynamics Structural Dynamics and Mechanical Vibration Consultants
Course Demo 8

Structural Health Monitoring using Statistical Pattern Recognition

2. Data Acquisition: Conventional Monitoring vs. WiMMS

Sensors Sensors Sensors Cables Data Acq. Bus Unix Box

Micro- Wireless Processor Modem Sensors Batteries A-to-D

Wireless Modem

PC

Sensors

Cabling SU Centralized Data Acquisition SU SU

SU

Wireless Communication Sensor Units SU Centralized Data Storage SU

SM

From Straser, 1998

Los Alamos Dynamics Structural Dynamics and Mechanical Vibration Consultants

Course Demo 9

Structural Health Monitoring using Statistical Pattern Recognition

2. Data Acquisition: Commercial Wireless Monitoring Systems

2.5 cm

Developed at UC Berkeley EE Dept. Marketed through Crossbow, San Jose See www.xbow.com

Local processor and transmitter Photo Detector and thermal sensor Analog Devices two axis accelerometer

Los Alamos Dynamics Structural Dynamics and Mechanical Vibration Consultants

Course Demo 10

Structural Health Monitoring using Statistical Pattern Recognition

2. Data Acquisition: Demonstration of the Mote System

A portal test structure

The preload in the bolt is varied by a PZT actuator

The loosening of the bolt is detected and reported by the LDE lights in the sensor

Correlation reading between two sensors

Course Demo 11

Los Alamos Dynamics Structural Dynamics and Mechanical Vibration Consultants

Structural Health Monitoring using Statistical Pattern Recognition

3. Feature Extraction: Flowchart of Model Update-Based Damage Identification

Damage is identified by comparing two finite element models: one correlated with undamaged data; one correlated with damaged data

Modal Frequencies and Mode Shapes from Test of Damaged Structure

Update FEM with Data FEM Correlated with Damaged Data

FEM Correlated with Undamaged Data

DAMAGE?

Los Alamos Dynamics Structural Dynamics and Mechanical Vibration Consultants

Course Demo 12

Structural Health Monitoring using Statistical Pattern Recognition

3. Feature Extraction: Uncracked vs. Cracked Beam Response: Wigner-Ville Transform

Uncracked

Cracked

Los Alamos Dynamics Structural Dynamics and Mechanical Vibration Consultants

Course Demo 13

Structural Health Monitoring using Statistical Pattern Recognition

4. Statistical Modeling: Outline of the Statistical Process

1. DATA ACQUISITION
From healthy structure
x (t )

2. PREDICTION MODELING
AR CVA

3. FEATURE EXTRACTION
AR Coefficients Residual Error Modal Parameters

time From damaged structure

x (t )

Kalman Filter

Neural Network time

Flexibility Matrices

4. CONTROL CHART CONSTRUCTION

X-bar Chart S Chart CUSUM
i or e

5. MONITORING
* * * * ***** * * * **

6. DECISION MAKING

* *

UCL CL LCL

Los Alamos Dynamics Structural Dynamics and Mechanical Vibration Consultants

Course Demo 14

Structural Health Monitoring using Statistical Pattern Recognition

How to Get Started in Structural Health Monitoring We will be happy to help you get your program going:
Consult for you on various aspects of your project: Program and resource planning Experiment design Feature selection and identification Statistical methods Conduct an in-house short course tailored to your application

Please contact us for any further information:

Email Hoon or Chuck directly: sohn@la-dynamics.com farrar@la-dynamics.com http://www.la-dynamics.com (435) 603-0375 Los Alamos Dynamics Structural Dynamics and Mechanical Vibration Consultants
Course Demo 15