model-based diagnostics for wire health … (wheeler...model-based diagnostics for wire health...

Model-based Diagnostics for Wire Health Management

Kevin R .Wheeler, PhD - NASA Ames Research Centerkevin.r.wheeler@nasa.gov

Jules Friederich – MCT Inc. at NASA ARC Stefan Schuet – NASA Ames Research CenterDogan Timucin, PhD – NASA Ames Research CenterXander Twombly, PhD – RIACS at NASA ARCPhil Wysocki – ASRC At NASA ARC

NASA Ames Research CenterMoffett Field, CA

National Aeronautics And Space Administration

Ames Research

Center

Dryden Flight Research

Center

Jet Propulsion Laboratory

MichoudAssembly

PlantStennis Space

Center

Kennedy Space Center

Marshall Space Flight Center

Johnson Space Center

Langley Research Center

Wallops Flight Facility

HQ

Goddard Space Flight

Center

Glenn Research

Center

14 Major Centers

19,000 Civil Service Employees

40,000 Contract Employees

14 Major Centers

19,000 Civil Service Employees

40,000 Contract Employees

White Sands Test Facility

Chafing Wire/Harness 37%

Unspecified Short 18%

Broken Wires 11%

Connector Failure 9%

Other 1%

Unspecified Failure6%

Failure Due to Corrosion

5%

Loose Connection 4%Insulation Failure 3%

Circuit Breaker Failure 2%

Crossmating 2%

Short Due to Corrosion 1% Miswire 1%

MOTIVATION: EWIS Agency Fault Summaries

Navy

Breach in Insulation 55%

Wire cut off 14%

Splices (hidden) 9%

Hot stamp 9%

Did not find 9%

Harness not available 4% FAA Wiring Failures

20%

19%

19%

14%

7%

7%5% 4% 4% 1%

Wire Cut / BrokenConn Damaged / BrokenWire Chaffed Components FaultyConn Pin FailureConn CorrodedWire CorrodedMiswireWire ShortedOther

• According to Captain Jim Shaw, manager of the in-flight fire project for the United States Air Line Pilots Association (ALPA), there are on average three (3) fire and smoke events in jet transport aircraft each day in USA and Canada alone, and the vast majority are electrical [ASW2000].

• According to Air Safety Week, "aircraft are making emergency landings, suffering fire damage to the point of being written off etc, at the rate of more than one a month based on the experience of the past few months" [ASW2001].

• Entire U2 fleet grounded in 2007 due to wire chafing evidence.

Coast Guard

APPROACH: Wire Health Management Development

• Fault library development - Develop publicly accessible wire fault database containing electrical signatures of different faults and wire types.

• Inversion algorithm development - Develop inversion algorithms for retrieval of permittivity along length of each wire in harness.

• Requirements analysis and definition – Develop Questions we need to answer.

Philosophy:Data driven - Recreate wire faults in lab and acquire reflectometry dataSimulation - Simulate faults in commercial simulatorTheory - Theoretical wave equation modeling Bayesian inversion – Retrieving the model parameters over time.

Fault Library

InferenceSoftware

RequirementsAnalysis

Requirements

EmpiricalData

Sims EM Theory

BayesianInference

Requirements Analysis & Definition

• What are typical rates of chafe progression?• How do signals change with worsening faults?• What are the EWIS faults priorities and time-of-life requirements?

• What are the best signals to inject for different types of faults, wire types, and harness configurations?

• How often should data be collected?• When should data be collected?

Should we run additional “Canary” lines (fiber optics)?Should we sense for vibrations (piezo wire)?Should we suggest harness configurations (ribbon)?Junction box best practices for microwaves?

Fault progression analyses

Hardware specifications

Harness specifications for health management

Identify sources of chafing• Bulkheads• Conduits• Control cables• Harness clamps • Hydraulic lines• Installation

Time Domain Reflectometry A Primer

Image from Agilent TDR manual

• A voltage step is applied to the circuit under test• The reflected voltage (ex) is measured at input: Reflections contain

information about the health of the insulation.

Non-uniform transmission line with two discontinuities

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.60

0.2

0.4

0.6

0.8

1

1.2

1.4Coaxial cable

Time in nanoseconds

Vol

ts

InputReflection

Effects of Geometry and FaultsCoaxial Cable

0.6 0.7 0.8 0.9 1 1.1 1.2 1.3

0.08

0.1

0.12

0.14

0.16

0.18

0.2

0.22Coaxial cable

Time in nanoseconds

Volts

The straight section of wire is 60 mm long, and the bend radiusis 10 mm. Each wire has a 0.25mm radius PEC core and 1.5mm

radius dielectric with e=2.08.

Magnitude of reflection due to bend: 26 mVMagnitude of reflection due to fault: 109 mV

Effects of Geometry and FaultsLamp Cord

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.80

0.2

0.4

0.6

0.8

1

1.2

1.4Lamp cord

Time in nanoseconds

Volts

InputReflection

0.5 0.6 0.7 0.8 0.9 1 1.1 1.2

0.586

0.588

0.59

0.592

0.594

0.596

0.598

0.6

0.602

0.604

Lamp cord

Time in nanoseconds

Volts

Magnitude of reflection due to bend: 9.6 mV Magnitude of reflection due to fault: 6.5 mV

Model-based Requirements

• How to answer these questions?– Laboratory experiments– Flight data– Simulations– …

Moving beyond the “What is possible?” stage of investigation requires a model-based approach.

THEORY: Non-uniform Transmission LineContinuous Parameters

Voltage at any point can have a right going wave and a left going wave: ),(),(),( txVtxVtxV −+ +=

Incident waveform: )(),0()()0()( tVtGtVatV tcti ∗+= ++

Reflected waveform: )(),0()),0(2()0()( tVtGltVatV tctr ∗+−= −− τ

attenuation propagation time Compact Green’s function transmitted waveboundary condition

Split representation:)(),()()()),0(,( tVtxGtVxaxtxV tct ∗+=+ +++ τ

)(),()),(2()()),0(,( tVtxGlxtVxaxtxV tct ∗+−=+ −−− ττ

Z(0)

Z(L)Nonuniform line

Lx

Vr(t)

0

Vi(t) Vt(t)

Γ1 T1

Γ2 T2

Γ3T3

V

Lundstedt & Strom 1996

Non-uniform Permittivity Profile

0 20 40 60 80 100 120 140 1600

0.5

1

1.5

2

2.5

3

mm

Per

mitt

ivity

V(incident) V(Transmitted)

Validation of Theoretical and Microwave Simulation Models of a Heterogeneous

Transmission Line

0 0.5 1 1.5 2 2.5 3-0.2

0

0.2

0.4

0.6

0.8

1

1.2

Time (nsec)

Vol

ts

V(transmitted - Greens)V(transmitted- MWS)Residual between V(trans-Greens),V(trans-MWS)V(incident), Greens output and MWS input

Movie of Compact Green’s function step response

x

V(x

,t)

Movie of Compact Green’s function Gauss response

x

V(x

,t)

Non-uniform Transmission Line

Can form mapping between impedance parameters and our representation:

{ } { })(),(),,(),,()(),(),(),( xaxatxGtxGxGxRxcxZ cc −+−+⇔

)()()(

xCxLxZ =where and c(x) is propagation velocity.

Can also form forwards and inverse mappings from observables to model parameters:

{ } { })(),(),()(),(),(),,(),,( tVtVtVxcxaxatxGtxG tricc ⇔−+−+

Forward model

Inverse model (Bayesian inversion)What we really want!

Random uncertainties make inversion stochastic:- wire wiggles, insulation thickness varies - splices introduced- instrumentation noise, wireless packet noise - cables move

0 0.05 0.1 0.15 0.2

1.3

1.35

1.4

x 108

c(x) - reconstructed (blue) - 256 Samples - Solution Method: Iterative

- R = 0.275, G= 1e-022

0 0.05 0.1 0.15 0.2-2

0

2

x 105c - reconstruction error - rms error: 71465.7

0 0.05 0.1 0.15 0.250

52

54

56

Z(x) - reconstructed (blue)

0 0.05 0.1 0.15 0.20

0.005

0.01

Z - reconstruction error - rms error: 0.00216154

0 0.05 0.1 0.15 0.2-500

0

500

Zx(x) - reconstructed (blue)

0 0.05 0.1 0.15 0.2

-0.1

0

0.1

Zx - reconstruction error - rms error: 0.0349996

Reconstruction

An Operational Concept

Op Concept:• Collect data every time plane lands (Vi, Vr, Vt).• Transfer data to ground station.• Algorithm infers state of insulation continuously along critical harnesses

(infer model parameters).• Changes in state over days and weeks are used to detect chafing and

to predict remaining useful life.

Unique Approach:• Change interrogation signal based upon type of faults.• Quantify uncertainties (there are no straight wires).• Produce useful remaining life estimates.

Chafing Measurements

• Goal: Detect chafing while only the shielding is being damaged.

• Experiment Methodology– Measure shielded twisted pair without fault

with TDR– Using abrasive apparatus, chafe wiring

specified number of cycles after shield exposure

– Measure with TDR.

Wire Chafing MethodsGoals:

• repeatable by other laboratories• inexpensive setup• allows for shooting TDR • allows for micrometer access• representative of real-world conditions

- vibration based- conductive chafe

Abrasive rod

Wire mountedunderneath

pivot

Chafed Shielding

BMS13-48, Type 24, Class 2, 2 conductor shielded.A.E. Petsche Co.: BMS13-48T24C02G018

Variability in Chafing to Shield

Since shielding is woven strands (not solid), some variability exists in chafing to shield.

Experiment: Vibrated abrasive chafing mechanism at 10 Hz. counted number of cycles until low voltage d.c. current flowed from abrasive rod to shielding.

Two chafing trials

Progressive Chafing Examples(approximately 1 cm stroke)

4K cycles beyond shield short.Shield still relatively intact.

6K cycles beyond shield short.Shield has holes.

8K cycles beyond shield short.Dielectric of inner conductors is exposed.

10K cycles beyond shield short.Inner dielectric is scraped but intact.

0.02 0.04 0.06 0.08 0.1 0.12 0.140

1

2

3

4

5

6

7

8

9

10Histogram of rate of chafing

mm change in chafe depth

Num

ber o

f occ

urre

nces

Rate of Chafing

0 1 2 3 4 5 6 7 83.7

3.75

3.8

3.85

3.9

3.95

4

4.05Chafing Progression

Thousands of chafing cycles beyond shield exposure

Wid

th o

f wire

in m

m

Consistency in rate of chafing(slope of lines)

Histogram of depth of chafe per2,000 cycles

5 5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8x 10-8

0.14

0.16

0.18

0.2

0.22

0.24

0.26

0.28

0.3

Time in Seconds

Vol

ts

0K cycles 2K cycles 4K cycles 6K cycles 8K cycles 10K cycles

5.2 5.22 5.24 5.26 5.28 5.3 5.32 5.34 5.36 5.38 5.x 10-8

0.134

0.135

0.136

0.137

0.138

0.139

0.14

0.141

0.142

Time in Seconds

Vol

ts

Progressive Chafingon single twisted pair

Chafe location

Started cycle count once shieldingwas first exposed by chafing.

Important to note: Damage is localized to shielding ONLY,No inner conductors were exposed.

Reflection due to

least chafing

Reflection due to

most chafing

TDR Resultsbaseline – shielding exposed

5.2 5.25 5.3 5.35 5.4 5.45

x 10-8

-1

-0.5

0

0.5

1

1.5

x 10-3

Time in Seconds

Am

plitu

de in

mV

Exposed shielding is not detectable from no fault case.

All TDR trials

5.05 5.1 5.15 5.2 5.25 5.3 5.35 5.4 5.45 5.5x 10-8

0.134

0.136

0.138

0.14

0.142

0.144

0.146

0.148

0K cycles 2K cycles 4K cycles 6K cycles 8K cycles 10K cycles

TDR ResultsProgressive chafing – multiple wires

difference from baseline

5.22 5.24 5.26 5.28 5.3 5.32

x 10-8

0

1

2

3

4

5

6x 10-3 Difference from baseline

Time in Seconds

Abs

olut

e di

ffere

nce

in m

V

2K cycles 4K cycles 6K cycles 8K cycles 10K cycles

6K8K

10K

Twisted Pair Geometry

Simulation Results

0.55 0.6 0.65 0.7 0.75 0.8 0.85

0.05

0.055

0.06

0.065

0.07

Time (ns)

Vol

tsSimulated TDR Response for Progressivey Growing

Shield Flaw in Twisted Pair Wiring

Depth 0.6Depth 0.65Depth 0.7Depth 0.75Depth 0.8Depth 0.85Depth 0.9Depth 0.95Depth 1.0Depth 1.1Depth 1.15Depth 1.2

Reality vs. Simulation

5.2 5.22 5.24 5.26 5.28 5.3 5.32 5.34 5.36 5.38 5.x 10-8

0.134

0.135

0.136

0.137

0.138

0.139

0.14

0.141

0.142

Time in Seconds

Vol

ts

0.55 0.6 0.65 0.7 0.75 0.8 0.85

0.05

0.055

0.06

0.065

0.07

Time (ns)

Vol

ts

Simulated TDR Response for Progressivey GrowingShield Flaw in Twisted Pair Wiring

Depth 0.6Depth 0.65Depth 0.7Depth 0.75Depth 0.8Depth 0.85Depth 0.9Depth 0.95Depth 1.0Depth 1.1Depth 1.15Depth 1.2

Kevin.R.Wheeler@nasa.gov