S. Schuet K. Wheeler D. Timucin M. KowalskiP. Wysocki

Intelligent Systems DivisionNASA Ames Research Center

Moffett Field, California

Aging Aircraft 2009

https://ntrs.nasa.gov/search.jsp?R=20090037095 2020-03-31T08:59:16+00:00Z

Team Members and Contact Information

Name email

Kevin Wheeler kevin.r.wheeler at nasa.govTeam Lead

Dogan Timucin dogan.a.timucin at nasa.gov

Phil Wysocki philip.f.wysocki at nasa.gov

Stefan Schuet stefan.r.schuet at nasa.gov

Agenda0 Introduction & Motivation• Importance of Wiring Faults I• Chafing Faults

© Characterization of Chafing Damage• 3D Commercial EM Simulator f• Experimental Methods• Fault Library

© Model Based Inference• Bayesian Framework• Chafe Model

Recent EWIS IncidentsFebruary 2009 A fire breaks out on-board an A340 Virgin

Atlantic flight en route from Heathrow to Chicago,which could not be extinguished until after theplane landed and depowered. Investigators laterdiscovered problems with the electrical wiring in abar unit of the plane that was specifically adaptedfor the airline.

January 2008 American Airlines B757 Flight 1738 experiencedsmoking in the cockpit caused by arcing within thewindshield heat system resulting in the cockpitwindshield shattering during the emergencylanding.

Chafing is a dominant EWIS failure mechanismo A frequently occurring type of wiring fault

• FAA reports 55% in one study)• Navy reports 37% in another study,

o Precursors to more significant problems:• open and short circuits (cause instrument failure)• arcing (causes smoking, fires, or worse!)

t See K. Wheeler et. a!_, Aging Aircraft Wiring Fault Detection Survey" for an

overview of these studies and more

Detection of Chafing Damage• Efforts in hardware development for chafe detection have

been reported, however, little attention has been focusedon detectability and uncertainty

• Initial investigations on detectability suggest that chafingon unshielded (e.g., power cables) wires is difficult if notimpossible to detect,

• Shielded wire (e.g., high-speed communication cables),however, may generate a detectable signature.

t "The Invisible Fray: A Critical Analysis of the Use of Reflectometry for Fray

Location," Griffiths et al_, IEEE Sensors Journal, vol. 6, no_ 3, June 2006.

9 = y^ 4 C°

Chafing in Shielded Electrical Cablingo Chafe first ablates outer insulation, then shield, leaving

inner conductors intact

MidFigure: Chafe progression: 2k, 4k, and 8k cycles beyond short

o Time domain reflectometry (TDR) between activeconductors and the shield is proposed for chafe detection

9 = y^ 4 C°

3D Commercial Electromagnetic Simulatorv Computer Simulation Technology (CST)'s Microwave

Studio is usedo Wire Types:

• Coaxial Cable• Twisted Shielded Pair

v Fault Types:• Rectangular• Elliptical (pictured)• Multiple Faults

5 MM

3D Commercial EM Simulator: Representative Datav Response of twisted shielded pair (TSP) to TDR

interrogation

p..... ............................... .................... .................... ................... :............

-0.01 .................:.................. ...................: .................. .... .........

-0.02 .................:.................................................................... ......

j -0.03 . ................:............................ ............................. ....................... ........

m 2mmx2mm

2mm x 4mm-0.05 ... ............... :................... ........... .........^.... 2mm x 6mm ........................

o--OM .................. ................... ..................... I.... —2mm x8mm

2mmx10—

-0.07 ............................... ............

sno....

.................:...................................\.................

.............. ........................ . I1Ll:.1V.1 ..... ..................... ........... ..... .^............

0 0.2 OA 0.6 0.8 1Time (ns)

Experimental Methods for Chafe Characterizationo There are two fundamental modes of chafing damage

• Wire movement versus stationary object (e-g-, wire rubbingon a bulkhead)

• Wire movement versus wire movement (e.g., wires in abundle abrating each other)

o Two machines have been developed to mimic thesedamage mechanisms

• Stationary rod abrasion machine• Wire-on-wire abrasion machine

Stationary Rod Abrasion Machine

Specifications Range Optimal SettingStroke Length 1 - 3 cm 1 cm

Stroke Frequency 1-100 Hz 10 Hz

• 200g mass used to pressure wire against diamond coated chafing rod

• Average chafe to inner conductor time for TSP is 12k cycles

19

Wire-on-Wire Abrasion MachinefSpecifications Range Optimal SettingStroke Length 1 cm 1 cm

Stroke Frequency 1-20 Hz 10 Hz

• 500g mass used to pressure wire against diamond coated chafing rod

• Average chafe to inner conductor time for TSP is 25k cycles

t Thanks to AFRL for the design

Rod Abrasion Machine: Representative Data0 Two fault example, one fault fixed and the other growing in size

.................. .................................................................. ....(128]0. 1250)cycles

0.198 .. .....(12670. 5250) ­1 - - -- - -

----- .............. ............... ......... ... ....... ...... ...(126)0.)250) ­1

(12670. 9250) ­1­g

. . . . ................. ......... ....(12—, 1 —) —1- ...............................

0-2 ............... .............. ................................. ................. ..................O

............................................ .............. .. .................... .....

............. .............. ........ ....... ..... .....

............. .................................... ................ .................

rme Ins)

Wire-on-Wire Abrasion Machine: Representative Data

0 Growing braid on wire TDR data set

0A34.......... ........... ...................... ...........

-40aa cyd-

- 6000 cyder

0.132 - — 6000

— ioaaa

- 12aaa qaea

0.13- - 14000 cycle ........... r .......... I .......... I ........... .......... 4 ........- 16000 cyder

0.128- .. ^ .......i-......... ^...........i-... ^^^1, ^. ...... .......

7 ......... . .......... i ...........

0.124 - . ... .......... .......... . ...................... . .......... . ........... . .......... . ....

6 7 B 9 10 11 12 13 14 15 16

Time

Overview of Fault Libraryv Contains TDR response signals from both simulations and

lab experiments• Formatted in ASCII and Matlab binary files (.mat)• Simulated data was collected by growing fault length and

simulating the TDR response• Experimental data was collected by growing faults under

controlled chafing conditions and measuring the TDRresponse

http://ti.arc.nasa.gov/project/wiring/

Why Use Probability Theory for Wire Fault Detection?o Want to infer variables of interest from noisy reflected

electrical signals:• fault location(s)• fault size(s)

o Want to automatically cope with sources of uncertainty:• electrical noise from equipment and environment• unknown or uncertain cable parameters (e.g_, dielectric

permittivity, finite conductivity)• geometric distortions (e-g-, bends, wiggles)• other reflection sources (e.g., splices)• unknown number of faults all mixing together

o Specifically, the Bayesian approach provides a systematicapproach to incorporating these effects and more...

9 = y^ 4 C°

Benefits to the Bayesian Approacho Clearly represents uncertainty inherent within

measurementso Includes uncertainty in known prior information or

expertiseu e.g.,"known"' values, such as permittivity of wire insulation,

are often better represented as random variables knownonly to a certain accuracy

• Quantifies the uncertainty in the inferred parameters• Avoids taking direct inverse in finding optimal model

parameters by seeking the estimate that maximizes theprobability of the observed information

Conducting Research in two areas:

o Forward modeldevelopment:

• LTI Convolution Models• Analytical models• Behavioral models (how

things change)

input F (g output

o Optimization Techniques to retrieve parameters (find mostlikely parameters that explain the observed input andoutput).

• Convex Optimization & Expectation Maximization• Markov Chain Monte-Carlo (MCMC)• Reversible Jump MCMC

1.

Example: LTI System Model

V, (k)TDR _Oo 61

Oz ON_1V^(k)

ZL

V,(k) = OoV, (k)+ e1 V; (k- 1)+... + eN- 1 V; (k— N+1)

o The reflection coefficients O k and input V,(k) are giveno Motivated through physics by assuming the line is lossless

and linear time invariant (LT I)

19

Example LTI System Model

Vi (k)TDR 90 ^^ ^2 6N-1

V (k)

ZLVr(k ) = O * Vi(k)

• O E RN is the variable we want to estimate• F(0) = O x V; = HO represents our model

o H E RN " N , is a convolution matrix• y = F(0) + v, where v E RN is Gaussian noise

• Prior information is that O is sparse, since chafing damageis small and localized

9 = y^ 4 C°

• Likelihood: Prob(ylF, O) x e-2o211F(o)-vllZ• Prior: Prob(OIF) x e-_k o' aklokl

v A heuristic for prior information that (D is sparse• Solve: maximize Prob(OIF,y) x Prob (ff,O) Prob((DIF),

which is equivalent to,

N-1

minimize 2Q2 IIF(^) — y 2 + ^ Akl 0kl (1)k=0

v For fixed Ak , (1) is a convex optimization problem, and thussolvable globally and efficiently, even for large N.

o The Expectation Maximization algorithm can be used toautomatically find the best "tuning" parameters Ak.

Q

Example LTI Convolution Model Estimation Resultv N = 1024, At = 0.04 ns

TDR Data

.d—ren. Mwn r.0

^re amp

__ -ren.ara

o.+4

o.0

0.128 8.5 ] ].5 — 8

5kAt (ns)Fauh Deremon

BrwMt^

20

B BS ] ]S 8

5kAt (ns)

+u

NASA Ames Research Center I Wire Chafe Diagnostics

Mathematical Predictive Chafe Model: Impedance Layeringo Assume that chafe can be thought of as one or a series of

impedance disturbanceso Once impedance of fault is known, it is relatively

straightforward to find the response of the cable infrequency or time domains

z^ z j (^) z.

Computing Capacitance and Inductance0 Capacitance:

7 . 600 = 0Qr = ff V - Dds

f —FVO • (n x z) dl

0 Inductance:

_ µoEL^ C

10 15 20Frequency (GHz)

J5

-40

45N_

50

55

500

-30

Frequency (GHZ)

Frequency Domain Verification : Comparison with CST MWS

o Predicted Return Loss from a 2 x 10 mm chafe (left) and a2 x 15 mm chafe (right) in coaxial cable.

Experimental Verification: Chafe in Coaxial Cable

Q ^

0.1£

f 0.10

0.1£

0.1

Experimental Verification: Comparison to Lab TDR

__ —1 FDTD+2DFDM^^

3.4 3.6 3.8 4 4.2 4.4 4.6 4.8Time (s) x 1 0 e

Conclusions• Machines — Developed two chafing machines used to

mimic effect of chafing on cables.• Datasets — Development of publicly accessible electrical

signature fault datasets.• Algorithms:

• Development of probabilistic Bayesian algorithms forunderstanding and characterizing electrical signatures offaults

• Development of compact and efficient electromagneticsbased forward models for chafe signature

Future Worko Incorporation of forward models within a Bayesian

framework is underway• "Real-life" experimental platforms (NASA wind tunnel and

vertical motion simulator) are being used• Live communication cable interrogation (CAN bus) is being

modeled and contemplated as representative platform.

ReferencesK. Wheeler et al.Aging Aircraft Wiring Fault Detection Surveyhttp://ti.arc.nasa.gov/m/pub/1342h/1342 (Wheeler).pdf

D. SiviaData Analysis, A Bayesian TutorialOxford Science Publications, 2003

S. Boyd, L. VandenbergheConvex OptimizationCambridge University Press, 2003http://www.stanford.edu/—boyd/cvxbook/