1

PrognosisTMS 2003 Annual Conference

San Diego, California

L. Christodoulou

Defense Science Office

Defense Advanced Research Projects Agency

J. M. Larsen

Materials Directorate

Air Force Research Laboratory

2

AcknowledgementsAcknowledgements

TMS for sponsoring the symposium

Session organizing committee

Numerous contributors who have helped shape our thinking on the subject

Sessions contributors

DEDICATED TO THE LATE ARTHUR DINESS --A GREAT FRIEND AND INSPIRATION

3

Goals of the Presentation Goals of the Presentation

Define “Prognosis”

Establish the context for “Prognosis”

Encourage collaboration across disciplines (materials, sensor technology, data processing and fusion, feature extraction, etc.).

Provide an example of an approach

4

DefinitionDefinition

PROGNOSIS: -- Knowledge of future performance based on reliable prediction capability of individual platforms.

Context: -- Materials and Structures

5

Prognosis - Power is Knowing the FuturePrognosis - Power is Knowing the Future

Delphi Oracle

6

Prognosis-based Asset Management Approach Prognosis-based Asset Management Approach

• Management, deployment and use of assets based on PROGNOSIS -- knowledge of future performance based on reliable prediction capability of individual platforms.

• Managing according to knowledge of the individual and actual remaining performance

• Managing uncertainty by reliable (physics-based?) predictive capability

• Enabling material “state awareness”

Prognosis Translates Knowledge and Information Richness to Physical Capability

7

Presently, “Fear of Failure” controls our design, management, deployment and use of all critical elements of complex mechanical systems (aircraft, helicopters, space vehicles, submarines, ships, UAVs, etc.).

• Forces undue conservatism (large safety factors) reducing performance.

• Severely impacts system availability and readiness.

• Forces non-optimal use of available assets.

• Results in high cost.

Present ParadigmPresent Paradigm

8

Fear is Justified: Materials Failure Matters!Fear is Justified: Materials Failure Matters!

9

Fear is Justified: Materials Failure Matters!Fear is Justified: Materials Failure Matters!

10

Fear is Justified: Materials Failure Matters!Fear is Justified: Materials Failure Matters!

PROBABLE CAUSE: "The National Transportation Safety Board determines that the probable cause of this accident was the inadequate consideration given to human factors limitations in the inspectionand quality control procedures used by United Airlines' engine overhaul facility which resulted in the failure to detect a fatigue crack originating from a previously undetected metallurgical defect located in a critical area of the stage 1 fan disk……... The subsequent catastrophic disintegration of the disk result in the liberation of debris in a pattern of distribution and with energy levels that exceeded the level of protection provided by design features of the hydraulic systems that operate the DC-10's flight controls." (NTSB/AAR-90/06)

Date: 19 JUL 1989Type: McDonnell Douglas DC-10-10Operator: United Air Flight 232Registration: N1819UYear built: 1973Total airframe hrs: 43401 hours Cycles: 16997 cyclesTotal: 111 fatalities / 296 on board Location: Sioux City-Gateway, IA

11

The PresentThe Present

Database:Mission History,

Maintenance, Life Extension and Design.

Insp

ect

Decision

Yes

NO

Empirical Criteria

Management, deployment and use of high value systems is dominated by our fear of failure

Failure Occurrences

Usage (D

uty Cycles)

“Book” Life: 99.9 % Unfailed

12

Impact of UncertaintyImpact of Uncertainty

Conde

mned

Engine

Disks

13

The Prognosis VisionThe Prognosis Vision

Database:Mission History,

Maintenance, Life Extension and Design.

Yes

NO

Prognosis

Failure physics, damage evolution,predictive models

Capability Profile, e.g

speed

altit

ude

Stat

e A

war

enes

s

Interrogation

Prognosis Translates Knowledge and Information Richness to Physical Capability

14

Life Management Through PrognosisLife Management Through Prognosis

log Life (e.g. Cycles or TACs)

Failu

re

8000(Design Life)

log Life (e.g. Cycles or TACs)

Failu

re

(Near Actual Life)

Improved Life Prediction

15

Technical ApproachTechnical Approach

What are the enabling elements of a Prognosis paradigm?

How would one practice Prognosis?

What would one actually do?

How do you go from “dislocations” to throttle settings?

16

Interrogation and State Awareness Interrogation and State Awareness

Conceptual:• Not inspection• Allows the material and structure to communicate

its statePractical:

• Local (embedded/in-situ) or global information• Multi-spectral, -spatial, temporal• May require external perturbation or pre-defined

maneuver(s)• Benchmarked (initially and subsequently?)• MAY demand inspection (last resort)

Computational:• Feature extraction• Dimensionality reduction• Reliable error estimation

Stat

e A

war

enes

sInterrogation

17

InfraredCamera

Physics of FailurePhysics of Failure

Crystal Plasticity Models Scalable Computational Models and Algorithms

Auxiliary Models

Physically BasedLife Prediction Model

Model Development

Model Validation Life Prediction

Crack Initiation andGrowth Prediction

Analytical Predictions• Polycrystal Plasticity• Machining Defects• Inclusions

IDDSInteractive Analysis and

Experiment

Displacement Field Mapping

OIM

Conventional Testing

Probabilistic Material Characterization

• Explicit FEA• Adaptive Meshing• Multi-scale Models

• Subscale Processes• Subscale Properties• Defect Distributions• Microstructural Data• Residual Stresses

300

400

500

600

10 -1 10 0 10 1 10 2 10 3 10 4 10 5 10 6 10 7 10 8

σm

ax (M

Pa)

N (Cycles)

3

4

5

6

7

8

9

10

0 100 200 300 400 500 600 700 800

∆K t

h (M

Pa¦m

)

Temperature (°C)

10 -10

10 -9

10 -8

10 -7

10 -6

10 -5

6 7 8 9 10 20 30

da/d

N (m

/cyc

le)

∆K (MPa¦m)

1

2

3

4

5

0 1 10 6 2 10 6 3 10 6 4 10 6

RT600°C800°C

Nor

mal

ized

CM

OD

Cycles

K5-DuplexR=0.1

300

400

500

600

10 -1 10 0 10 1 10 2 10 3 10 4 10 5 10 6 10 7 10 8

σm

ax (M

Pa)

N (Cycles)

18

Creep HCF LCF

Fret. Embr. Envr.

Physics of Failure

?

? ?

?

The LinkThe LinkSystem Parameters (throttle setting)

Component, e.g., disk

Evolving Material Microstructure

Temperature, stress, time, environment, etc. (incl. distribution)

inputs to state equations

CFD, FEMTemperature, stress, time,

environment, etc. (incl. distribution)

Interrogation Tools for State Awareness

LocalLaser UltrasonicsThermoacousticThermoelectric

GlobalThermalAcousticVibration

Crack Detected

Failure

19

Failure is Neither Random or UnpredictableFailure is Neither Random or Unpredictable

Failure mode DOMAINS well defined (fatigue, creep, corrosion, etc.)

Failure is progressive:

NUCLEATION/INITIATION

PROPAGATION/ESCALATION

COALESCENCE

Reliable failure PREDICTION will be accomplished by combination of;1. Models of physics of failure

Evolution of damageCoupled effects

2. Interrogation tools for state awarenessLocal AND globalSignature manifestations

20

Conceptual Model/State Awareness FusionConceptual Model/State Awareness Fusion

Knowledge of failure domains establishes functional behavior of damage evolution.

Tracking changes not absolute values.

Fidelity/reliability increases with prognosis system usage and maturity.

Short term predictions more reliable than long term “lifing” predictions.

Uncertainty in model predictions modulated by state awareness tools.

Dam

age

Mission/Usage

Dcrit.

Time of prediction

21

How Does it Work?How Does it Work?

0

20

40

60

80

100

120

140

160

180

0 5000 10000 15000 20000

Time, Sec

Stre

ss, k

si

0

100

200

300

400

500

600

Tem

pera

ture

, F

StressTemperature

Model Mission State at time, t

State at time, t+∆t

Signature of “new state” provides confidence that model prediction is

within expected regime.

22

Vision: Materials Damage PrognosisVision: Materials Damage Prognosis

Key science and technology disciplinesCoupled physics-based models of materials damage and behavior

• Interaction of multiple damage/failure mechanisms• Multi-scale, mechanism-based• Microstructurally-based stochastic behavior• Integrated information from state-awareness tools

Interrogation of damage-state• Intelligently exploit existing sensors• Feature extraction from global sensors• Materials-damage-state interrogation techniques and

recorders• Sensor signature analysis

Data management and fusion• Capability matched to mission • Component usage data• Component history and pedigree

23

Concluding RemarksConcluding Remarks

• Urge vigorous discussion

• Constructive dialogue not (just) criticism

• Thinking “out of the box”

• Encourage inter- and multidisciplinary research

• Attack the hard problems

• How do use sensor data to update model predictions?

• How to efficiently link localized (micro) damage to macro behavior?

• How can complexity be reduced?

• Accept that models, sensors and IT techniques will continue to evolve and improve thus attempt to establish methodologies/protocols for communicating, fusing and exploiting the different disciplines now.

24

Questions?

25

Interrogation and State Awareness Interrogation and State Awareness

Conceptual:• Not inspection• Allows the material and structure to communicate

its statePractical:

• Local (embedded/in-situ) or global information• Multi-spectral, -spatial, temporal• May require external perturbation or pre-defined

maneuver(s)• Benchmarked (initially and subsequently?)• MAY demand inspection (last resort)

Computational:• Feature extraction• Dimensionality reduction• Reliable error estimation

Stat

e A

war

enes

sInterrogation

26

Existing Database (History and Past Missions)Existing Database (History and Past Missions)

DO REALLY use past mission history

• Identify salient features of every mission

DO take into account knowledge of the system behavior

• Track trends

DO take into account maintenance history

Exploit expert knowledge

Leverage previous efforts

Exploit IT revolution.

Database:Mission History,

Maintenance, Life Extension and

Design.

27

Damage Evolution Damage Evolution

Use knowledge of applicable physics.

Invoke and exploit coupled and interacting mechanisms.

Use multiple models (if available).

Physics-based and data-driven models will evolve—allow for updates.

Reduced and full models.

Sensor data can modulate model predictions.

Failure physics, damage evolution,predictive models

28

Database:Mission History,

Maintenance, Life Extension and Design.

Prognosis

Failure physics, damage evolution,predictive models

Stat

e A

war

enes

s

Interrogation

Prognosis Reasoning SystemPrognosis Reasoning System

Integrates of all elements, system knowledge and logic

Predicts of capability

Provides multiple decision makers the required information (operator, local commander, theatre director, maintenance, etc.

Provides confidence levels on predictions

Employs sophisticated and evolving reasoners

Conveys pertinent information for easy assimilation

Relies on local and rapid e.g. onboard (reduced) response and more complete e.g., remote control center (full) system models

Benchmarked at convenient times and locations

Based on open and modular architecture

29

Prognosis ContentPrognosis Content

• Science and Technology• Predictive, coupled, multi-scale damage evolution/physics of failure models• (Non-intrusive) state awareness techniques and tools.• Math techniques for feature extraction and characterization of the state of the

system.• Performance projection capability based on current state.• Adaptive mission strategies and (on-board) reasoning/intelligence system• Tools to give multiple users reliable and accurate capability status in “real time”

• Technology• Existing/develop test beds to validate tools and models• Leverage data fusion technologies to implement Prognosis architecture and

reasoning system• Exploit effective data mining techniques (from IT?)

• Demonstrations• Demonstrate impact through analysis and physical demonstrations• Deliver decision tools for pervasive (sub)system manned or unmanned systems.

Back to Main Menu