44

Wavelet Analysis of Cockpit Voice Recorder Data

Embed Size (px)

Citation preview

Page 1: Wavelet Analysis of Cockpit Voice Recorder Data
Page 2: Wavelet Analysis of Cockpit Voice Recorder Data
Page 3: Wavelet Analysis of Cockpit Voice Recorder Data
Page 4: Wavelet Analysis of Cockpit Voice Recorder Data
Page 5: Wavelet Analysis of Cockpit Voice Recorder Data
Page 6: Wavelet Analysis of Cockpit Voice Recorder Data

Wavelet Analysis of

Cockpit Voice Recorder Data

Page 7: Wavelet Analysis of Cockpit Voice Recorder Data

Project ResponsibilitiesTeam Leader

Digitize CVR Data, Develop Test Signals

Page 8: Wavelet Analysis of Cockpit Voice Recorder Data

Project ResponsibilitiesCode Developer

Develop Wavelet source code

Page 9: Wavelet Analysis of Cockpit Voice Recorder Data

Project ResponsibilitiesResearch

Research Wavelet Theory

Page 10: Wavelet Analysis of Cockpit Voice Recorder Data

Overview

•Project Background

•CVR Basics & Recording

•Theory

•Wavelet Parameters

•Silent Track Analysis

Page 11: Wavelet Analysis of Cockpit Voice Recorder Data

Project Background

1991 Beech 1900C Twin Turboprop Accident

Page 12: Wavelet Analysis of Cockpit Voice Recorder Data

Accident Location

Project Background

Page 13: Wavelet Analysis of Cockpit Voice Recorder Data

•Right engine mount failed

•Engine separated and struck empennage

•Aircraft broke apart in flight

•Recovered CVR with last 30 min of audio

Project Background

Page 14: Wavelet Analysis of Cockpit Voice Recorder Data

Engine Mount Location

Page 15: Wavelet Analysis of Cockpit Voice Recorder Data

Accident Animation

Page 16: Wavelet Analysis of Cockpit Voice Recorder Data

Project Goals

• Digitize CVR recording

• Isolate tube failure location

• Develop Wavelet analysis program

• Perform Wavelet Analysis on CVR data

Page 17: Wavelet Analysis of Cockpit Voice Recorder Data

Digitize CVR Data

Page 18: Wavelet Analysis of Cockpit Voice Recorder Data

2146:29 INT-2 ah once your established inbound right you're good to what?

2146:34 INT-1 oh #.

2146:34 CAM ((sound of altitude alert chime))

2146:35 INT-2 stop one thing at a time. you're in a bad situation

so correct one thing first.

CVR Transcript and Tape

Catastrophic Event

Page 19: Wavelet Analysis of Cockpit Voice Recorder Data

Goldwave Analysis

Goldwave is a digital audio editing program.

Page 20: Wavelet Analysis of Cockpit Voice Recorder Data

Goldwave Analysis

Windowed Fourier Transform

Page 21: Wavelet Analysis of Cockpit Voice Recorder Data

Goldwave Analysis

Windowed Fourier Results

These frequencies match those found by past research.

EventTurbines

Propeller Blade PassageCVR Noise

Frequency (Hz)5400 & 10300

105400

Page 22: Wavelet Analysis of Cockpit Voice Recorder Data

Goldwave Analysis

Final portion of CVR recording

Page 23: Wavelet Analysis of Cockpit Voice Recorder Data

Integral Transformation Theory

What is it?

Transforms a function from one space to anotherdepending on the transfer function.

Page 24: Wavelet Analysis of Cockpit Voice Recorder Data

The Fourier Transform

Signal processing tool that takes asignal and tells us what frequencies are within that signal.

Page 25: Wavelet Analysis of Cockpit Voice Recorder Data

The Wavelet Transform

Determines the frequency spectrum as a function of time.

Page 26: Wavelet Analysis of Cockpit Voice Recorder Data

Morlet Wavelet Equations

Morlet Wavelet

Scaling/Translation Function

Wavelet Transform

Page 27: Wavelet Analysis of Cockpit Voice Recorder Data

Triboelectric Effect

• Frictional motion creates static electricity

• Aircraft wires vibrate during flight

• Vibrations generate static electricity

• CVR records static electricity

Page 28: Wavelet Analysis of Cockpit Voice Recorder Data

Develop Wavelet Computer Code

• Obtained existing Wavelet code from C. Torrence and G. Compo

• Created interface allowing use of Wavelet code

• Experimented with effect of input parameters

Page 29: Wavelet Analysis of Cockpit Voice Recorder Data

Important Input Parameters

• Wo – Effects Wavelet width

• dj – Effects scale width

• mother – Type of Wavelet

Transform Parameters:

Scale Parameters:

• J1 – # of Scales

• So – Base Scale

Page 30: Wavelet Analysis of Cockpit Voice Recorder Data

An Example

Page 31: Wavelet Analysis of Cockpit Voice Recorder Data

Fourier Analysis

Page 32: Wavelet Analysis of Cockpit Voice Recorder Data

Wavelet Analysis of Test Signal

Morlet Wavelet

Wo = 6dj = .25

Wo = 24

dj = .25

Wo = 24

dj = .025

Page 33: Wavelet Analysis of Cockpit Voice Recorder Data

DOG Wavelet

Wavelet Analysis of Test Signal

m = 2dj = .025

Page 34: Wavelet Analysis of Cockpit Voice Recorder Data

Wavelet Display Test

Wo = 6dj = .125

Morlet Wavelet

Page 35: Wavelet Analysis of Cockpit Voice Recorder Data

• Scale Parameter Equations • Minimum Scale Size

• Maximum Wavenumber

• Beat Phenomenon

• Edge Effects and Tone Width

Wavelet Code Improvements

Page 36: Wavelet Analysis of Cockpit Voice Recorder Data

Scale Parameter Equations

Base Scale Size

Number of Scales

Page 37: Wavelet Analysis of Cockpit Voice Recorder Data

Minimum Scale Size

Page 38: Wavelet Analysis of Cockpit Voice Recorder Data

Maximum Wavenumber5 Hz Test Tone

Page 39: Wavelet Analysis of Cockpit Voice Recorder Data

Beat Phenomenon

50 Hz & 52 Hz Tones

Page 40: Wavelet Analysis of Cockpit Voice Recorder Data

Wavelet Analysis

4p Propeller Mode

Page 41: Wavelet Analysis of Cockpit Voice Recorder Data

Wavelet Analysis

Right Engine Turbine

Page 42: Wavelet Analysis of Cockpit Voice Recorder Data

Future Work

• Continue verifying code consistency

• Obtain controlled triboelectric test data

• Investigate other crash explanations

Page 43: Wavelet Analysis of Cockpit Voice Recorder Data

Acknowledgements

•Dr. Ronald Stearman

•Pete Moss – UT School of Music

•Chris Garland

•Javier Fuentealba

•Marcus Kruger

Page 44: Wavelet Analysis of Cockpit Voice Recorder Data

Questions