Wavelet Homepage                               Table of Contents

1 Introduction       

  1.1 Airplane Crash Investigations

            Much of airplane design involves ensuring a high level of safety for every airplane; however, crashes do occur and the resulting loss of life is tragic.  Determining the cause of a crash can improve overall aviation safety by bringing to light any faults in the current design of an aircraft, and it can serve as a guide for improving future designs.  The key problem lies in deciding which of many factors were the primary causes of any single crash.  Often the expert crash investigators disagree.  Generally, the more data that is recorded and available on a flight, the more investigators are able to piece together what happened.  For instance, radar tracking data, weather data, a Cockpit Voice Recorder (CVR), and a Flight Data Recorder (FDR) are all important tools for an aircraft crash investigator. 

Unfortunately, only a limited amount of data is recoverable from most accident sites because of the extent of the damage done to the plane.  It is the job of the investigator to determine the most likely scenario with what information he has.  For many commuter airplanes, the CVR is the investigators’ only insight into the mind of the flight crew.  In fact, even if the crew does not specifically say what the problem is, the cause of the crash can sometimes be determined from a thorough analysis of the CVR. 

In recent years, researchers like Dr. Ronald Stearman at the University of Texas at Austin have begun to use the non-voice signatures present on the recordings to provide insight into the cause of a crash.  Physical events such as the failure of a load-bearing member propagate both through the air in acoustic waves, as well as through materials in the form of pressure changes.  Dr. Stearman has demonstrated that CVRs are sensitive to both the acoustic and non-acoustic spectrums.  The purpose of this report is to analyze the signals present on the CVR from one particular crash and to determine if any signals are the result of events that contributed to the accident, but were unknown to the crew.

  1.2 Accident History

On December 28, 1991, a Beech 1900C twin turboprop commuter airplane (tail number N811BE) crashed in the cold waters of the Atlantic Ocean just off the coast of Block Island, Rhode Island. The crash killed all three people onboard the plane: an instructor pilot (IP) and two student pilots who were trying to make the rank of captain.  The flight was a training mission and involved emergency engine-out situations and other problem scenarios [2].   The National Transportation and Safety Board (NTSB) investigated the crash and determined that “the instructor pilot’s loss of altitude awareness and possible spatial disorientation” caused the crash.  But an independent investigation completed by the Airline Pilot’s Association (ALPA) determined that “the right engine separated from the wing and struck the horizontal stabilizer causing loss of control and eventual breakup of the aircraft while in flight [3].”  According to ALPA’s investigation, a strut (or struts) broke in the right engine mount, causing a whirl-flutter event that tore apart the airplane [3].

  1.3 Previous Work

Determining the true cause of the crash is important because, as of now, the NTSB blames the IP for the deaths of three individuals.  If the IP was not at fault, then his name deserves to be cleared and the responsibility for the crash needs to be properly placed.  Of greater importance, if the Beech 1900C has an inherent safety problem, then the entire fleet needs to be grounded until the problem is fixed.  Clearly, the two proposed causes of the crash are completely different; to resolve this discrepancy, several professional organizations, experts, and ASE 463Q groups have examined several suspected causes of the crash (see Table 1). 

Table 1: List of reports reviewed for this project, including date written and author

 

 

From the previous work, the first conclusion that affects this paper is the evidence in the ALPA report demonstrating that the right wing of the 1900C broke off the plane during the flight.  The fact that the right wing was found undamaged four miles away from the rest of the wreckage suggests the wing came off prior to impact with the water.  Also, a study on the Beech 1900C’s engine mounts found serious problems with struts breaking during flight [4].  Several planes studied had multiple broken struts in the engine mount with only a limited number of flying hours.  In fact, if the lifetime of the engine mount were to be compared to a human’s lifetime, then it would be in the infant mortality category.  As it turns out, if the engine mount did have a fractured strut, then a whirl flutter event is possible.  “Propeller whirl flutter is a dynamic instability that involves the gyroscopic precession of a flexibly mounted engine-propeller system [5].”  This report assumes that a whirl flutter event ripped the right engine and wing from the plane.

Aside from one of the tubes of the engine mount breaking, another potential cause of the whirl flutter event that should be examined is the possibility that one or more of the four bolts supporting the engine mount broke.  If one of these bolts fatigued and broke, then that could have created instability in the propeller motion, which would have initiated a whirl flutter event.

Therefore, two cases need to be analyzed: an engine mount tube breaking and an engine mount bolt breaking.  A previous study showed that mechanical failures, which may or may not produce an audible sound, show up on the CVR tracks [6].  For instance, it has been demonstrated that a signal is recorded by running someone’s fingers along a wire.  The details of how a wire records a signal will be thoroughly discussed later in the paper. 

Finally, the original CVR tapes were digitized in The University of Texas School of Music so they that could be analyzed on a computer [7].  By analyzing the digitized CVR tapes, we hope to confirm whether events on the CVR are mechanical failures or are events that normally occur during a flight, such as raising the landing gear.  All of the above items involved previous work that helped us completes our goals.

  1.4 Project Goals

This project has three primary goals:

1.      Perform a wavelet analysis of the CVR data from the crash to find mechanical events on the CVR and compare those events to test results from both previous semesters, and from this semester, to determine whether a strut or a bolt on the engine mount did break during the flight.

2.      Formulate comprehensive test plans describing the types of tests that are necessary, the materials that are needed, and how to conduct them.  

3.      Conduct the tests with a tape recorder to record the break and the vibrations, digitize the data, and then perform a wavelet analysis of the data to obtain characteristic signatures.

We hope to provide conclusive evidence as to whether or not events on the CVR are either truss members or a bolt breaking in the plane’s right engine mount.

  1.5 Team Member Assignments

We have three group members working on this project.  All of us have been involved in every part of the project.  Although we all shared the work, we did have primary responsibilities: Matt Anderson- Group Lead; Roberto Diaz- Engine Mount Simulation Test Plan; and Joshua Foxworth- Wavelet Analysis.

  1.6 Timeline

 

 

Figure 1: Timeline of Tasks Completed and the Time Required to Accomplish These Tasks