Wavelet Homepage                               Table of Contents

Abstract

          Time-series and wavelet analyses were performed on the cockpit voice recorder (CVR) data from a Beech 1900C that crashed off the coast of Block Island, Rhode Island on December 28, 1991.  Previous investigations have shown that a probable cause of the crash was an engine failure.  University of Texas at Austin professor Dr. Ron Stearman has theorized that the CVR is capable of recording mechanical events such as impacts, breaks, or engine failures. A Haar Wavelet analysis program was written in Matlab to find transients in the CVR signal, and the Matlab Wavelet Toolbox was used to perform a Coiflet wavelet analysis on significant events.  With this software, we analyzed all four tracks on the Beech 1900C CVR, a nominal engine tape, and data from tests performed by a spring 1998 design group on a Star-Lite home-built aircraft.  This analysis has allowed us to identify what we believe is a break at the beginning of the silent track (PA system), as well as significant energy in the whirl flutter frequency band just before the final event.  We have determined that this break, as well as “spike” transients evident throughout the tape after the initial event, is produced by the triboelectric effect.  We have determined that other events such as four “middle” events that occur roughly twenty minutes into the thirty-minute tape were caused by electromagnetic sources.  Given this evidence we deduced that the spike transients were caused by the break seen in the initial event and may be the result of the broken member’s movements.  The broken mount did not affect the flight of the plane until the engine was placed into flight idle, at which time the plane entered a whirl flutter situation that caused the plane to crash.  We have outlined a series of tests that we believe will be able to conclusively prove that the source of the initial event is a break.  Another goal of the testing is to establish base reference data to obtain a fundamental understanding of the appearance of mechanical events on the CVR.

 

Wavelet Homepage                               Table of Contents