Abstract Aircraft accident forensics include analyzing cockpit voice recorder (CVR) and flight data recorder (FDR) information。 Each recorder is independent of the other and often share no common time frame。 Analysts must align the two time scales using common events on each recorder, such as microphone keying events。 Using test drive development and linear programming, we have created a two-step model to align the CVR and FDR recordings。 The first linear model matches microphone keying events, while the second linear model optimizes alignment of the time scales。 When the constrained optimization problem produces a feasible solution, a standardized computer-generated sentence is created that is useful for forensic reporting  purposes。73745

Keywords:   convolution, linear programming, recorder, time alignment, CVR,  FDR

Introduction

Aircraft accident investigation is an essential process to determine probable cause and propose recommendations to reduce future risk。 Investigators use many types of evidence to determine what caused an accident; one source of information is commonly known as the black boxes (Schueller, 1998)。 Commercial aircraft have two black boxes installed: a flight data recorder (FDR) and a cockpit voice recorder (CVR)。 The FDR records many aircraft parameters such as speed,

About the Authors

William A。 Tuccio is a PhD student in the Aviation program with Embry-Riddle Aeronautical University。 His experience and research interests are in the area of aviation, data-centric solutions, and transportation safety。 Today, William contributes to transportation accident investigations。 This work is the author’s alone and does not reflect the official position of any employer or government   agency。

Harold D。 Townsend is a PhD student in the Aviation program with Embry-Riddle Aeronautical University, where he also acquired a Bachelor degree in Aeronautical Science later followed by an MBA from the University of Alaska Anchorage。 His experience includes that of an airline pilot, aircraft mechanic, flight operations training manager, financial analyst and financial controller。 His academic interest is airline finance and  operations。

http://dx。doi。org/10。5703/1288284314862

altitude, and aircraft configuration。 The CVR records pilot voice transmissions, cockpit noises, and engine sounds。 Through a combination of the recorded information, investigators are able to analyze accident related  events。

Statement of Problem

Forensic accident investigation presents many chal- lenges。 One critical challenge is an accurate and consistent time scale between different pieces of evidence。 This arises when aligning the separate time scales of the aircraft CVR and FDR; Figure 1 shows the nature of the problem。 The technologies used by many in-service CVRs are such that the audio recorded by the CVR lacks an external time reference。1 While FDRs often have an external time reference, for the purpose of this problem statement, it is sufficient to say that the FDR time reference is different from the CVR time reference (Gregor, 2006)。 Investigators successfully use a variety of techniques to align timing of CVR and FDR data, as well as timing of other data sources (i。e。, air traffic control radar, witness statements, etc。)。

Statement of Purpose

The purpose of this paper is to use software engineering test driven development (TDD) to demonstrate how linear programming can be applied to time synchronization between CVR and FDR data。 After delimiting the problem, the paper develops and tests a linear programming model for time synchronization in the LINGO2 programming language。

Nature of Problem

A common technique used by accident investigators to align CVR and FDR data is to use pilot initiated radio transmissions as the basis of alignment (Gregor, 2006)。 That is, an analyst can hear when the cockpit microphone transmit button is depressed (‘‘microphone keying’’) on the CVR, and can further bolster the accuracy of this observation by looking at an audio spectrum of the CVR content。 The accuracy of microphone keying on the CVR can often be identified to within a tenth of a second。 The FDR records many parameters—sometimes thousands of parameters—sampled at various rates varying generally from every four seconds to every 16th of a second, depending upon regulations and installation configuration (Federal Aviation Administration, 2010)。 One of the critical timing parameters recorded is microphone keying。 Whenever a pilot keys the microphone, the FDR records the microphone keying as a binary value (i。e。, 1 for  keyed,