[Federal Register Volume 61, Number 92 (Friday, May 10, 1996)]
[Notices]
[Pages 21522-21526]
From the Federal Register Online via the Government Publishing Office [www.gpo.gov]
[FR Doc No: 96-11725]



-----------------------------------------------------------------------

DEPARTMENT OF TRANSPORTATION
Federal Aviation Administration
[Docket No. 28567]


A Call for the Development of Prototype(s) for a Global Analysis 
and Information Network (GAIN)

AGENCY: Federal Aviation Administration, DOT.

ACTION: Notice; request for comments.

-----------------------------------------------------------------------

SUMMARY: David Hinson, Administrator of the Federal Aviation 
Administration (FAA), stresses that Zero Accidents is the only 
acceptable safety goal for the aviation industry and the FAA. This 
notice offers some ideas for the elements needed to establish an early 
warning capability for existing and emerging safety concerns that would 
move the aviation industry towards Zero Accidents, and challenges the 
aviation industry to participate in developing that capability. Because 
of an emerging combination of improved cooperation between airline 
management, labor, and various governments, advancements in information 
technologies, and the political environment in several countries, the 
international aviation industry has an unprecedented opportunity, by 
sharing and analyzing aviation safety information, to reach Zero 
Accidents.
    This notice (a) solicits comments on the Global Analysis and 
Information Network (GAIN) concept and implementation strategy for 
collecting and analyzing aviation safety data, and (b) invites 
participation in the development of proof-of-concept prototypes. All 
interested parties, whether or not in the aviation community, are 
invited to comment on the ideas presented, offer alternative solutions, 
indicate interest in helping to develop a GAIN prototype or the overall 
system itself, and comment about how government aviation safety 
agencies can best help the industry reach Zero Accidents.

DATES: Comments in response to this call for action must be received by 
June 14, 1996.

ADDRESSES: It is requested that all comments be submitted via the 
Internet by sending an e-mail message with your comments (plain text 
preferred, no graphics please) to: concept __paper@
asyweb01.nasdac.faa.gov.
    Please include your name and organization. Comments must also be 
mailed in hard-copy (two copies) via regular mail to: Federal Aviation 
Administration, 800 Independence Ave., SW., Office of Chief Counsel, 
Attention: Rules Docket (AGC-200), Docket No. 28567, Washington, DC 
20591.
    All comments must be marked: ``Docket No. 28567.'' Commenters 
wishing the FAA to acknowledge receipt of their comments must include a 
pre-addressed, stamped postcard on which the following statement is 
made: ``Comments to Docket No. 28567.'' The postcard will be date 
stamped and mailed to the commenter.
    Comments submitted about this Notice may be examined at the FAA at 
the above address in room 915G on weekdays, except on Federal holidays, 
between 8:30 a.m. and 5:00 p.m. In addition, commenters will be able to 
review all other comments by Internet. Your submission should not 
contain any proprietary or other information that you do not want to be 
made available to the public.

FOR FURTHER INFORMATION CONTACT: Mr. Chuck Fluet, Manager, Safety 
Analysis Division, Office of Aviation Safety, ASY-200, Federal Aviation 
Administration, 400 7th Street, SW., Washington, DC 20590, telephone 
202-267-GAIN (202-267-4246).

SUPPLEMENTARY INFORMATION: The aviation industry has made remarkable 
progress in reducing aviation accident rates. With today's volume of 
flights, the industry would have suffered more than 10,000 fatalities 
last year worldwide if the accident rate had not improved so 
dramatically since 1960. Because of this major decline in the accident 
rate, the industry now suffers an average of less than 800 fatalities 
worldwide per year. However, the rate has remained stubbornly 
consistent for about the last 10 years; and at today's accident rate, 
forecast growth in air transportation demand will lead to more than 
4,500 fatalities worldwide per year by 2025--clearly an unacceptable 
result.
    Sound methods for certifying the safety of new aviation products 
and procedures, as well as surveillance activities that help to ensure 
safe operation and maintenance of these products and procedures, have 
contributed significantly to the current safety levels of the aviation 
industry. Within the framework of these regulatory methods, 
technological advances in engine performance and reliability, airframes 
and materials, air traffic control, cockpit automation, and simulator 
training have contributed to the safety of the aviation system. 
Compliance inspections, accident and incident investigations, special 
studies, and program evaluations are the fundamental methods of 
continuing surveillance in the operating environment, and safety has 
improved significantly over the years in part because of the lessons 
learned by using these methods to understand the mistakes and 
oversights of the past.
    Yet all too often, the industry has not been able to use data about 
accidents, incidents and other system anomalies to become aware of 
existing or emerging safety problems in time to take preventive 
measures. Just as traditional product design and manufacturing methods 
eventually gave way to new, improved principles and methods, a new 
safety information paradigm, with much greater sensitivity to anomalies 
in daily aviation system operations, could help the industry reach Zero 
Accidents. Just as aviation product improvements

[[Page 21523]]

of the past have been fostered by technological advances, improvements 
in aviation safety monitoring and alerting will become possible as a 
result of major advances in information management technology.

An Outline of a New Safety Monitoring Paradigm

    The industry must develop a significantly improved operational 
early warning capability that is sensitive enough to detect and alert 
the aviation community to existing and emerging problems. A major 
aspect of this capability is the sharing of safety information, both 
within categories in the industry, e.g., carriers must share with other 
carriers and manufacturers with other manufacturers; and across 
categories, e.g., labor, management, carriers, manufacturers, air 
traffic controllers, airport operators, and others must share with each 
other. Creating useful information, however, generally requires the 
collection of large amounts of data, and it also involves the careful 
analysis of that data. Rarely would there be any need or desire to 
share any raw data, but the sharing of the analysis of the data--the 
information--could be mutually beneficial. Gathering and analyzing 
large amounts of qualitative and quantitative aviation safety data to 
better understand routine system operations is the foundation of the 
Global Analysis and Information Network (GAIN) concept.
    The GAIN system would be more sensitive to conditions that signal 
increased safety risks because it would contain information about 
normal aviation system operations. The statistical baseline for normal 
aviation operations, constructed with digital flight and ATC radar 
data, among other major and currently untapped sources, would be the 
plumb line from which deviations are measured. The importance of 
obtaining information about a far greater percentage of aircraft 
operations has been illustrated repeatedly by all-too-typical accident 
investigation findings of earlier flights that experienced problems 
similar to the accident aircraft. A truly effective early warning 
capability would involve significant improvements in information 
feedback and analysis for aviation operations. At a minimum, the GAIN 
concept would add the following new elements to the existing monitoring 
systems to improve sensitivity:

--New data sources that would improve risk assessment and provide a 
baseline for normal flight operations, thus improving the chances of 
early anomaly detection.
--New and innovative data management and analytical techniques and 
methodologies that quickly reveal obscure and/or infrequent data 
patterns and associations.
--New methods to disseminate the findings quickly and globally to all 
who could use them to improve aviation safety.

Analytical Strategies and Automated Tools

    The proposed analysis process would be based upon new sources of 
information and new information technology capabilities. First, 
information from voluntary reporting programs (such as the confidential 
Aviation Safety Reporting System (ASRS), or the Air Safety Reports 
(ASR) used by certain airlines) and mandatory incident reporting 
systems (such as the Pilot Deviation or Runway Incursion data bases) 
would be subjected to a range of analysis tools. These include advanced 
data pattern searches--which can be performed autonomously on the data 
by ``intelligent agent'' automation tools to discover patterns or 
associations, finding the ``needle in the haystack.'' ``Intelligent 
agent'' software would aid analysts in discovering thematic 
associations in text data bases, and data visualization tools would 
show the analyst associations in data base elements. Application of 
such data mining analysis tools would provide a more focused 
understanding of operational safety concerns much sooner than current 
analysis techniques. The data management and analysis take place in a 
``data warehouse'' where operational data are extracted from existing 
systems and, through a series of steps that standardize and improve the 
quality of the data, the data are transformed into a data base designed 
for targeted analysis. Within a ``data warehouse'' environment, safety 
analysts can employ various data mining strategies.
    Once existing or emerging safety concerns are identified, 
hypotheses that are developed to explain them can be tested using 
empirical digital flight data, ATC radar data, or other appropriate 
data sources. A focus on remedial measures would at times result from 
an analysis of digital flight data or ATC automated data, both vast 
sources of empirical data.
    As a result of new information technologies, we have the 
capability, for the first time, to monitor and analyze the parameters 
of safe and normal flight. Until very recently, it has been very 
difficult to obtain accurate and reliable information on normal flight 
operations. Now, thanks to new computer technologies, we can use flight 
data recorder and radar information to generate large amounts of very 
accurate and detailed information about flight performance. For 
example, the Boeing 777 records information on 700 flight parameters 8 
times a second. Several countries, mostly in Europe, have programs in 
which a carrier or civil aviation agency routinely monitors and 
analyzes operational data captured on flight data recorders.
    Statistical analysis of digital data or ATC automated data from 
normal flights would yield a baseline of routine operations that can be 
used to detect variations from norms. In addition, baseline statistics 
would help safety analysts quantify operating risks within, as well as 
beyond, the envelope of normal operations. By collecting and analyzing 
information primarily about what went wrong, we are missing the 
opportunity to learn what was done right to avoid an accident or 
incident in earlier situations. The likelihood of detecting problems 
and developing remedies is significantly greater from studying large 
numbers of normal daily operations than from relying primarily upon a 
far smaller number of periodic inspections or accident and incident 
investigations.
    Analysis of digital flight data can provide several types of 
information, including aircraft path analysis, derivation of 
environmental conditions, aircraft configuration time histories, 
aerodynamic coefficients (analysis of coefficients can reveal 
degradation in aerodynamic performance), engine performance, aircraft 
attitude, automated flight control modes and status, warning 
parameters, takeoff and landing distances, and flight loads. Digital 
flight data can be used to detect single anomalies--alerting operators 
when criteria values for selected parameters have been exceeded or when 
particular events occur. Such data also can be used to develop 
descriptive statistics across fleets, to detect deviations from 
statistical norms in the aviation system, or to measure the effects of 
design, procedure, or equipment changes.
    ATC automated data could be used to analyze airplane motion and 
relative position, important factors in analyzing issues such as wake 
vortex and environmental effects. An analysis of air traffic control 
automated data for normal operations could provide insight into methods 
for improving ATC system operations or potential problem areas. Flight 
data anomalies from accidents could be compared to similar anomalies of 
flights that did not crash to learn what was done differently to avoid 
an

[[Page 21524]]

accident. These findings might suggest guidelines on pilot training or 
aircraft design. The same autonomous ``intelligent agent'' analysis 
techniques used to find patterns in data from incident reporting also 
can be applied to digital flight data and ATC radar data--or 
information derived from this data such as paths, flight loads, or 
aerodynamic coefficients--to determine if any otherwise unobserved 
associations exist within the data.

Human Factors Analysis

    This analytical process and the new sources of data under 
consideration could significantly improve our ability to describe what 
is happening in the aviation system, and a comparable human factors 
analysis capability must also be developed. Without a reliable human 
factors analysis tool that addresses the underlying causes or factors 
associated with emerging safety concerns, remedial measures may only be 
temporary ``band aids.'' An effective human performance analysis 
capability developed for use on digital flight data or ATC automated 
data--augmented by feedback from voluntary disclosure systems--is an 
essential part of an early warning system.

A Proposed Architecture for Sharing

    As noted above, for a number of reasons, not the least of which is 
the very large quantity of data, there will probably be little or no 
sharing of raw data, but only of information from the analysis of data. 
Moreover, because of improved networking technologies and capabilities, 
information would not necessarily all be sent to a massive computer at 
one location, but would probably be available to different users to 
different extents by networking--sometimes known as a ``virtual 
database.'' For example, this networking capability makes it possible 
for each carrier, manufacturer, or union to have separate GAIN-type 
systems, or they could do it collectively with one or more others or 
through trade associations, or any combination of them, and the 
information sharing could occur over the network to the extent desired 
or permitted by the owner of each system.
    The information that results from GAIN analyses would ideally be 
available immediately to all recipients who could use it to improve 
aviation safety. The dissemination of vital information can be 
accomplished with existing infrastructure--using the Internet, for 
example, if adequate safeguards can be provided to protect the security 
and confidentiality concerns of the information providers regarding 
identified or identifiable data. The GAIN network would have to 
accommodate different requirements in a user-friendly way, and be able 
to notify automatically all appropriate recipients about potential 
problems without requiring them to know to query the system.

Examples of Proactive Use of Aviation Safety Data

    There are several examples in various countries that demonstrate 
how effectively proactive safety measures can be implemented as a 
result of industry/labor/government partnership sharing of such 
information. When one air carrier's data indicated that pilots were 
frequently disregarding their Ground Proximity Warning System (GPWS), 
the carrier discovered that the frequent disregard was due to a high 
false alarm rate, and further analysis of the data provided the basis 
for developing a software remedy. As a result, that GPWS system was 
improved (to the benefit of all carriers that used it), the false alarm 
rate dropped, and pilots ignored the warning much less.
    Similarly, a carrier that was experiencing frequent altitude 
capture excursions and deviations in one of its aircraft types found 
from the data that the problem was a combination of inadequate pilot 
training and poor altitude capture logic. Analysis of the data provided 
the basis for improving both the training and the logic. Again, the 
logic fix benefited all users of that autopilot around the world, not 
just the carrier that discovered the problem.
    Other examples include improvements to training programs and/or 
operations manuals as a result of high pitch angle takeoffs, more rapid 
that desirable takeoff rotation rates, inadvertent flap/slat retraction 
out of the proper speed range, and unstabilized approaches; design 
fixes for equipment that did not perform as designed or anticipated 
(e.g., an aircraft that was developing cracks from hard landings at 
less than the 2 g cutoff beyond which inspection was mandated); and 
improvements in airport signs and markings to help pilots more 
accurately follow their taxi clearances.
    Also important, of course, is that without the data, it is very 
difficult for carriers, manufacturers, or governments to evaluate 
whether new programs and other fixes are having the desired result.

Concept Implementation Issues

Collection and Analysis of Aviation Safety Data

    In developing an analytical process for an early warning capability 
that would monitor the system and alert the aviation community to 
existing and emerging safety concerns, please consider what data 
requirements, analysis methods, and information dissemination methods 
you would propose. In relation to the analytical process, please 
consider and comment on issues such as the following:

-- What aviation safety data and information are needed to support your 
analysis plan and what, of those needs, is not now being collected?
--Should large quantities of data be collected on a wide range of 
safety issues, or less data on fewer targeted safety issues?
--To what extent is standardization of the data collection or of 
analysis techniques necessary? How should the necessary standardization 
be accomplished?
--How could existing data, such as information from voluntary reporting 
and correction, ASRS, AQP, FOQA, and other such programs, be analyzed 
better to provide meaningful and useful information?
--What could industry and government do to improve existing means for 
data collection?
--Are incentives needed to stimulate the submission of information that 
is not derived from accidents or incidents, as opposed to merely 
removing the disincentives, in order to encourage reporting?
--To what extent can international information sharing occur with a 
``virtual database'' instead of a physically centralized data base?
--What techniques and capabilities are you aware of in the aviation 
industry or in other industries to analyze data effectively and 
generate statistically significant results, with predictive value, from 
large quantities of data describing normal operations?
--What analytical techniques and capabilities are you aware of in the 
aviation industry or in other industries to respond effectively to the 
myriad of human factors issues that arise in operational monitoring 
analysis?
Dissemination of Aviation Safety Information
--To what extent are security measures needed, and what security 
measures are available, to protect information confidentiality while 
still assuring that it reaches all in the industry who could use it to 
improve aviation safety?
--What alerting methods are available to ensure that information is 
automatically distributed to all recipients who could use it to

[[Page 21525]]

improve aviation safety without their having to know to ask?
General
--Will an analysis and dissemination system such as GAIN help the 
aviation industry reach Zero Accidents?
--Are there better ways to help the industry reach Zero Accidents?
--What concerns, if any, do you have about the existence of an analysis 
and dissemination system such as GAIN?
--What should the relationship be between government regulators and 
GAIN for it to be most effective ?
--Although commercial aviation is the initial target for this effort, 
how can other sectors of the aviation industry, including the military, 
help with this initial effort?
--How can the program be expanded to include input from, and the 
development of remedies in relation to information provided by, 
manufacturing personnel, mechanics, flight attendants, dispatchers, 
ramp personnel, and other aviation industry professionals whose input 
could help with the proactive effort?

Considerations for Developing Prototypes

General

    Ultimately GAIN could develop into a comprehensive international 
network of systems for analysis and sharing of aviation safety 
information. However, that development would have to occur 
incrementally, starting with one or more prototypes of various pieces 
of the network. Among the areas that should be evaluated from prototype 
development are: international data standardization, data collection 
protocols, analytical methods, data sharing, alerting mechanisms, and 
the potential value of emerging technologies. By prototyping key 
elements of GAIN, it will be possible to obtain operational proof of 
the most significant new capabilities being incorporated in this early 
warning system. With a minimal initial commitment of resources, risks 
and costs would be reduced, while allowing the overall operational 
feasibility of the concept to be assessed. It would help define 
obstacles and issues associated with the development of GAIN, and 
provide valuable information for future implementation planning. 
Because more types of data, more types of analyses, and more users 
should not generally be added unless experience demonstrates that such 
additions would be useful, prototypes would help to provide the 
experience to determine the desirability of such additions.

Ownership

    For several reasons, the elements of the GAIN network should 
probably not be owned or operated by the FAA or the aviation regulatory 
agency of any other country. Instead, they should probably be owned by 
those members of the international aviation industry that benefit 
economically from its successful performance, analogous to existing 
collectively-owned, non-profit joint ventures in the aviation industry 
that provide services for the owners' mutual benefit. There are several 
reasons for pursuing this type of ownership. First, GAIN would probably 
enjoy better acceptance by the industry if it is not viewed as a 
government effort to gather information for enforcement purposes or to 
protect its own manufacturers and carriers in an international 
marketplace. Second, private ownership, as compared with governmental 
ownership, would facilitate protecting sensitive information from 
public disclosure. Third, the funding of GAIN should not depend upon 
the fiscal situation in any one country.
    Last, but not least, GAIN's existence would be most assured, and it 
would perform most effectively and efficiently, if it were owned 
jointly by those who have a direct economic interest in its success--
namely, the insurers, manufacturers, carriers, pilots, mechanics, 
controllers, and airport operators that make up the industry. Either 
GAIN will improve aviation safety and substantially reduce costs for 
the entire industry--because prevention costs less than accidents--in 
which event industry will want to own and operate it; or it will not 
accomplish these goals, in which event a better way must be developed 
to reach Zero Accidents.
    In determining how GAIN might be owned and structured, we invite 
your comment about:

--What types of prototypes could best demonstrate the concept at the 
lowest cost, given existing data collection and analysis techniques and 
capabilities?
--What entities could help develop prototype projects, how much would 
they cost, and what sources of funding are available?
--What role can you play in the prototyping effort and subsequent 
efforts to develop an operational GAIN?

The Role of the FAA

    The FAA is already engaged in several activities to demonstrate, in 
relatively small scale, the utility of safety data collection and 
analysis, but the GAIN network and its prototypes would probably not be 
FAA systems. The FAA's Office of System Safety could help facilitate 
the creation of GAIN by informing potential participants about the 
concept, and by bringing potential participants together, but the FAA 
will not own or operate GAIN, and will probably not fund its 
development. Instead, the FAA would be one of many users of the 
analytical results and supporting data from GAIN.
    Given the numerous proactive accident prevention activities that 
are already underway in various countries, it is likely that the 
aviation industry would eventually develop an international cooperative 
data sharing system, such as the GAIN network, on its own. The problem 
has been that it is difficult for any one profession, manufacturer, or 
airline to develop a program that systematically facilitates 
international sharing of information to the benefit of the entire 
international aviation community. Thus, in addition to facilitating 
this development by demonstrating its intent to cooperate more with 
industry to reach Zero Accidents, the FAA can play a major role in 
accelerating the progress of private industry by bringing together the 
entities that can help to develop GAIN prototypes--preferably by 
building as much as possible upon the systems that are already in place 
rather than starting anew--and by helping to assure that the prototypes 
are sufficiently standardized and consistent to work together in the 
more comprehensive GAIN network as it ultimately develops.

Conclusion: A Call To Action

    Please let us know of your ideas regarding the development of a 
GAIN network, particularly regarding how you can become involved, 
either in a GAIN prototype or in the more comprehensive permanent 
effort. This is not an invitation for bids or a request for proposals, 
but we are soliciting indications of interest, as well as input 
regarding the viability of this or any other concept to help the 
industry reach Zero Accidents.
    You are encouraged to review the comments (Commenters will be able 
to review all other comments by Internet) and be creative about how 
you, individually or together with other commenters, can begin the 
development of GAIN prototypes. If warranted by the nature and extent 
of the comments, the FAA will host a conference to bring interested 
parties together to discuss refinements of the GAIN concept and the 
development of prototypes.


[[Page 21526]]


    Issued in Washington, D.C., on May 7, 1996.
Christopher A. Hart,
Assistant Administrator for System Safety, Federal Aviation 
Administration.
[FR Doc. 96-11725 Filed 5-9-96; 8:45 am]
BILLING CODE 4910-13-P