NASA Technical Reports Server (NTRS) 19930016782: Anomalous event diagnosis for environmental satellite systems

12 Pages·0.57 MB·English

by  NASA Technical Reports Server (NTRS)

N93-25971
ANOMALOUS EVENT DIAGNOSIS FOR ENVIRONMENTAL SATELLITE SYSTEMS
Bruce H. Ramsay
National Oceanic and Atmospheric Administration
National Environmental Satellite, Data, and Information Service
Camp Springs, Maryland
ABSTRACT computer and communication
operations contractor.
The National Oceanic and
Atmospheric Administration's The Anomaly Reporting Expert
(NOAA) National Environmental Assistant System (AREAS) is an
Satellite, Data, and Infor- interactive, rule-based
mation Service (NESDIS) is demonstration prototype using
responsible for the operation backward chaining goal-directed
of the NOAA geostationary and inference. Upon input of a new
polar orbiting satellites. event's symptom, AREAS queries
NESDIS provides a wide array of a database of prior events with
operational meteorological and associated symptoms and causes,
oceanographic products and and then suggests possible
services and operates various causes to the analyst. AREAS
computer and communication reasons with the archived
systems on a 24-hour, seven events, a rule-based repre-
days per week schedule. sentation of the satellite,
communication, and computer
The Anomaly Reporting System subsystem's physical relation-
contains a database of ships, heuristics acquired from
anomalous events regarding the resident domain experts, and a
operations of the Geostationary mean best fit of prior events
with the new event. Whether the
Operational Environmental
Satellite (GOES) , com- analyst confirms AREAS' sug-
munication, or computer systems gested cause or enters a new
that have degraded or caused one, the event, with its
the loss of GOES imagery. Data related attributes, is entered
is currently entered manually into the database and thus
via an automated query user provides an up-to-date
interface. There are 21 pos- environment in which AREAS can
sible symptoms (e.g., No Data), operate. AREAS includes a help
and 73 possible causes (e.g., system designed to assist new
Sectorizer - World Weather users and it provides technical
Building) of an anomalous information, with graphical
event. The determination of an representation, on the GOES,
event's cause(s) is made by the communication and computer
on-duty computer operator, who subsystems.
enters the event in a paper-
based daily log, and by the Key Words : Knowledge-Based
analyst entering the data into System, Rule-Based, Backward
the reporting system. The Chaining, Goal-Directed
determination of the event's Inference, Environmental
cause (s) impacts both the Satellite Systems, Anomalous
operational status of these Event Diagnosis, Intelligent
systems, and the performance Database
evaluation of the on-site
107
INTRODUCTION operations support for the
GDDS.
The National Environmental
Satellite, Data, and Infor- Why Artificial Intelliqence?
mation Service (NESDIS)
oversees the operation of NESDIS determined to evaluate
civilian satellite systems used the potential of Artificial
for Earth-observation, and the Intelligence (AI) tools and
creation and maintenance of techniques in response to the
global databases in the challenge of sensing,
physical and life sciences. communicating, processing,
NESDIS provides products and analyzing and distributing
services derived from ever-increasing volumes of
environmenta i data that are environmental data and
applied to the protection of products. This increase is due
people and property, national to the larger number of ground-
economic systems, and the based data collection systems,
development and distribution of satellites in orbit with
food, energy and other natural improved sensors, and
resources on a national and additional data shared by other
international level. organizations, both public and
private, iin the United States
and foreign nations.
OBJECTIVES
Four objectives were
established for the development
and demonstration of the
Anomaly Reporting Expert
Assistant System (AREAS)
prototype. They were:
i. Develop a help system for
anomaly reporting,
Figure 1 - GOES East in Orbit
2. Increase personal knowledge
of GDDS.
NESDIS is responsible for the 3. Retain valuable GDDS exper-
operation and maintenance of tise.
the Geostationary Operational 4. Demonstrate the ability of
Environmental Satellite (GOES) , AI to improve administrative
located at 112 ° West as shown and operational tasks.
in Figure !aboye,: and the GOES
Data Distribution System CURRENT ANOMALY REPORTING
(GDDS) . It is staffed with SYSTEM
experienced meteorologists,
oceanographers, compu£er As a result of a computer
specialists, and administrative generated error message or
personnel, as well as employees other indicator, a computer
new to the environmental operator documents the problem
satellite domain. A contractor, in the paper-based Envi-
PRC, Inc., provides com- ronmental Satellite
munications and computer Distribution/Interactive
Processing Center (ESD/IPC)
108
Daily Log. At the conclusion of CAUSE CODES
! a shift, the shift supervisor CODE # ENTRY
synopsizes the Daily Log
01 ...... AIR COND/HEATING
F entries into multiple summary
reports including the
Operational Problem report. A
53 ...... SECTORI ZER-WWB
combined hardcopy daily report,
including the Daily Log and
Operational Problem report,
among others, is then distri- 73 ...... VIRGS
buted to management and staff.
Each morning, contractor and Accompanying the hardcopy
NESDIS personnel meet for a symptom and cause list is a
short discussion of the most GDDS Diagram, part of which is
critical issues encountered in shown in Figure 2. This diagram
the previous 24 hours. On a is not available on-line in the
daily basis a NESDIS staff ARS.
member evaluates the Daily Log
and Operational Problem report The complete diagram (not shown
and enters appropriate here) outlines the flow and
information into the Anomaly processing of data for the
Reporting System (ARS). The maj or communications and
staff shares this task on a computer systems from the GOES
weekly, rotating basis. spacecraft to NOAA's facilities
at Wallops Island, Virginia,
The ARS queries the user for and Suitland and Camp Springs,
the following information: Maryland. With this
information, along with other
Julian Date available documents, assigned
Satellite
staff must evaluate the Daily
Zulu Time
Log and Operational Problem
Symptom(s)
report and determine the cause
Number Products Affected
of the anomalous event. After
Probable Cause(s)
Responsible Division determination of the problem's
Number Expected cause and input of the data, a
Number Actuals
daily report is produced for
Number External Lost
dissemination.
The 21 possible symptoms and 73
ANOMALY REPORTING EXPERT
pOSsible causes are available
ASSISTANT SYSTEM
to the staff in hardcopy or in
an on'line _ text file. _ For
1.0 Problem Identification
example:
A loss of expertise was
SYMPTOM CODES
CODE # ENTRY recently suffered due to the
retirement and promotion of
Ol ...... DATA EARLY several employees. New emp-
loyees needed access to the
lost expertise in order to
14 ...... NO DATA accurately determine the cause
of anomalous events and prevent
future occurrences, if
possible. The current ARS has
21 ...... WRONG DATA
109
WORLD WEATHER BUILDING
(CAMP SPRINGS, MARYLAND) TOMICROWAVE
AT8UITLAND
TELCO 0 & A DATA
VIRGS
TO SOCC
VDUC IFFA
I VIE WMAICVEROI
1
XBAR
SEXRS co Ro,
COMM
COMPUTER
RACK
DCS
I TELCO
DC8 DATA
TO/FROM WALLOPS
I
TELCO DEVICES
Figure 2 - GDDS Diagram: WWB
no help system and several of integrated within a single
the new employees have limited category. The intent is to
knowledge of the GDDS. emphasize the real world
environment in which all three
2.0 Knowledge Acquisition issues are frequently
considered concurrently,
Domain knowledge was acquired especially during initial
through interviews with domain prototyping.
experts, one of whom has since
re£1red. Extensive GDDS docu- 3.1 Knowledge Representation
mentation, including the GDDS
Operations and Maintenance Evaluation of the existing data
Contract, was reviewed. In indicated that an attribute/
addition, the current Anomaly value representation scheme
Reporting System, ESD/IPC Daily combined with rule-based
Logs and Operational Problem processing would be sufficient
reports, Daily and Weekly ARS for initial prototype
reports were also analyzed. development. Since the current
ARS uses a query/response
3.0 Analysis and Design interface it was decided to use
backward chaining, goal
The analysis and selection of directed inference to emulate
knowledge representation and the existing process.
the development tool along with
the system design have been
llO
3.2 Tool Selection 4.0 Prototyping
The criteria for tool Modularity was a key issue
selection, in addition to those during the rapid prototyping of
identified above, were low AREAS since the knowledge
cost, a simple development engineering process was being
environment, and a short used as a learning tool for the
learning curve. The tool GDDS environment. A large
selected was Level5 Expert number of small, easily
System Software (DOS Version modified knowledge bases were
1.3) by Information Builders, initially prototyped to
Inc. This expert system shell establish the relationship
fit the identified small system among various data elements,
prototype requirements : a particularly between symptoms
default query/response and causes, and to model the
interface, symbolic represen- physical subsystems.
tation used in an if/then rule-
base environment, and the need 4.1 Knowledge Sources
to perform simple calculations.
4.1.1 Heuristic Knowledge
3.3 System Design
Heuristic knowledge was
The system design stressed obtained from the doma in
modularity for ease of experts through their
development, explanation, use, explanation of Daily Log and
and maintenance. The shell's Operational Problem report
default user interface was entries and GDDS processing.
employed for the selection of For example, the hardware
menu items and the input of element designation "RTIR"
numeric data. Individual (i.e., RealTime InfraRed) is
knowledge base modules were not specified in the sectorizer
used for each of the primary subsystem node of the Wallops
data items required for version of the GDDS Diagram,
inferencing. The help system's but it was identified by one of
customized graphic and the domain experts as important
narrative explanation screens in distinguishing between
were integrated through sectorizers at the World
Level5's explanation function Weather Building (WWB) and
and separate knowledge base those at Wallops Island. This
modules. The help system knowledge was then incorporated
focused on the three maj or into the rule base of AREAS.
components of the GDDS : the
satellite, data communications, 4.1.2 Documentation
and computer processing
subsystems. A simple mean A number of different documents
statistical analysis was were used as primary knowledge
provided through symptom sources. NESDIS Programs - NOAA
specific knowledge base Satellite Operations identified
modules. This architecture is the organization's mission and
demonstrated in Figure 3., major systems used in carrying
below. out that mission. The GDDS
Operations and Maintenance
Contract was indispensable in
identifying subsystems and
111
STATISTICAL
ANALYSIS
USER KNOWLEDGE MODULES
BASE
INTERFACE _--
MODULES
DATA
BASE
FILES
HELP
SYSTEM
I
I
I COMPUTERSJ
Figure 3 - AREAS Architecture
their constituent components. A 4.2 Process of Discovery
memorandum to all the
organizations responsible for Since one of the objectives of
the GDDS operation explained building AREAS was to gain
the use of ARS as, in part, an additional insights into GDDS,
instructional tool. It included AREAS had to be able to
the GDDS Diagram, of which the represent GDDS physical
WWB segment at Camp Springs, relationships among its
Maryland is shown in Figure 2, subsystems and components. For
and the lists of possible example, the following
symptoms and causes. The "identify symptom" rule
memorandum's express purpose represents the interpretation
was to establish a common of, and the relationship
framework in which to identify between, the shift supervisor's
and respond to anomalous comment in the remarks column
events. of the Operational Problem Log
and the identified symptom.
4.1.3 ARS Database
RULE identify symptom
Evaluation of the ARS data base IF remark IS Short-SZ Halted in RCV
THEN symptom identified
provided input to the data type
AND symptom IS Degraded Data
classifications used in AREAS,
as provided by the expert The remark, "Short - SZ Halted
system shell: numeric, in RCV," means the sectorizer's
attribute/value, and string.
112
processing of the imagery COMM RACK Communications Rack
DCS Data Collection System
product's data set during
IFFA Interactive Flash
transmission was terminated
Flood Analyzer
while in receive mode. The O & A Orbit and Attitude
symptom is thus classified as SEXRS Sectorizers
degraded data (i.e., by SOCC Satellite Operations
Control Center
definition 50 percent or more
TELCO Telephone Data Lines
of the complete image was
VAS VISSR Atmospheric
produced and was of animation Sounder
quality). The next rule, VDUC VAS Data Utilization
"identify responsible Center
VIE VAS Interface
organization," establishes the
Electronics
relationship between the VIRGS VISSR Image
identification of the satellite
Registration and
and a specific sectorizer and Gridding System
the organization responsible VISSR Visible and Infrared
for its operation. Spin - Scan
Radiometer
XBAR Crossbar Switch
RULE identify responsible
4.4 Statistical Analysis
organization
IF symptom identified
AND satellite IS GOES East The initial objective of a
AND hardware element IS Sectorizer statistical analysis of the ARS
6All
data base was to provide the
THEN responsible division
user with background
identified
AND responsible division IS SSD information as to the apparent
relationships between symptoms
and causes. This was accomp-
4.3 Help System lished through a simple mean
analysis of the type and number
of causes attributed to each
The help system provides query
specific information in symptom. This analysis, coupled
narrative and graphical formats with the rule output reviewed
of crucial areas of the GDDS. above, produces a diagnosis as
If the user doesn't understand shown below in Figure 4. The
a particular query, such as user is then at liberty to
"What was the responsible accept or reject the diagnosis.
division?" a help screen is
available with additional 4.5 Introduction of Knowledge-
information explaining the _ Based Systems
physical system relationShipS
and the organizations The focus on simplicity of the
_ponsible for th61r prototype's development,
oversight. Mutually supportive architecture, purpose, and
information from different Operation was important. These
documents is merged as well in issues had to be easily
help screens. For example, a explainable to use AREAS as an
glossary of terms such as the introduction of knowledge
one shown below was merged with engineering concepts. A basic
the GDDS Diagram in Figure 2, approach was taken in knowledge
above. acquisition, representation,
search, and inference for this
purpose.
113
Based on the following information:
Julian Date: 310
Satellite: GOES East
Zulu Time: O20O
Symptom: No Data
Number of Products Affected: 1
Hardware Element: SZ6A11
Do you agree with the diagnosis below?
Probable Cause: Sectorizer-WWB
Responsible Division: SSD
Yes
No
Figure 4 - Diagnosis Screen
5.0 Verification and Validation 6.0 Test and Evaluation
Verification was performed The qualitative test and
through analysis of shell- eva!uati0n of the AREAS
produced knowledge trees demonstratio nprototype fQcused
linking all the goals, rules on its potential use as a help
and attributes in a given system in identifying the
knowledge base in a logical symptoms and causes of
order of precedence starting anomalous events. Feedback from
with the top-level goal. Each user surveys indicated:
logical path through a
knowledge base was also - a Positive reac£ion_'_o _the
manually derived and tested. display of statistical data but
a need to further highlight
Initial validation was only the most prevalent
performed by comparing archived symptom/cause ratios;
results of dome in experts '
analyses to system generated - a desire to have AREAS
conclusions. Subsequent identify individual GDDS
validation was conducted by components and their output of
domain experts through the specific products (i.e.,
evaluation of results from test products are currently assigned
data sets processed by AREAS. identification codes and are
logically linked to specific
hardware elements within GDDS);
114
- approval of the graphics used systems development. Through
but a request for more detail the development of AREAS and
in representing GDDS subsystems preparation of this paper, he
and components; was able to take advantage of
the process of discovery
- the need to allow multiple highlighted by the knowledge
symptom identifications for a engineering process to gain a
single anomalous event (e.g., better understanding of the
the symptoms Data Incomplete GDDS.
and Degraded Data can be
specified for a single event in 3. Retain valuable GDDS exper-
the current ARS), and the tise.
ability to generate multiple
symptom/cause records per The retirement and promotion of
report; and several employees who were very
experienced in the GDDS created
- support for the ability to a potential problem for new
easily review input prior to employees assigned to anomalous
data base update and subsequent event tracking and analysis.
report generation. Part of their expertise,
through the knowledge
CONCLUSION engineering process, was
captured for use through AREAS.
Part, but not ai i, of each
objective was accomplished in 4. Demonstrate the ability of
the development and AI to improve administrative
demonstration of AREAS: and operational tasks.
i. Develop a help system for The only automated tool
anomaly reporting. available within ARS to search
the database requires the user
The current ARS has no help to possess a clear idea of what
feature. One of the expressed is being searched for and
purposes for users and new familiarity with the data types
employees using the ARS is to and structures employed. The
train them in the nomenclature purpose in providing the user
and processes of the GDDS. One with a simple mean analysis of
of the primary objectives of the data represents the initial
AREAS was the linkage of step in providing ready access
relevant narrative and to analytical tools and
graphical information to results. These tools, when
specific user queries. Based on augmented by heuristic rules to
user feedback AREAS has made an constrain the search space,
important step in identifying provide a reasonable method of
user needs in the successful diagnosis to assist the user in
analysis of anomalous events in making decisions. In the future
GDDS. these results may identify
potential trends in specific
2. Increase personal knowledge GDDS subsystems and hardware
of GDDS. elements, as well as processing
methodologies.
The author is a novice with
satellite-based systems but AREAS, with its focus on
experienced in knowledge-based simplicity, provides the
115
opportunity to introduce National Environmental
knowledge-based systems Satellite, Data, and
concepts, development and use Information Service. NESDIS
to employees in a direct, Programs: NOAA Satellite
hands-on way. It highlights the Operations. Washington, DC:
value of knowledge engineering U. S. Government Printing
as a process of discovery. It Office: 1985-472-036, 1985.
also demonstrates the ability
to harness the knowledge of PRC, Inc. GDDS Operations and
disparate sources of Maintenance Contract: Technical
information and provides a Management Proposal. Department
focus for that knowledge on of Commerce Solicitation Number
problem-solving in the domain 52-DGNE-9-00067. McLean,
of anomalous event diagnosis Virginia: PRC, Inc., 1989.
for environmental satellite
systems. Rao, K. & Holmes, H. &
Anderson, R. & Winston, J. &
REFERENCES Lehr, P. (Eds.). Weather
Satellites: Systems, Data, and
Cote, D. "Planning Begins for Environmental Applications.
Major Upgrade of NOAA's Data Boston, Massachusetts: American
Management System." Earth Meteorological Society, 1990.
System Monitor 3(1) (September
1992):1-2.
Harmon, H. & King, D. Expert
Systems. New York: John Wiley &
Sons, Inc., 1985.
Mairs, R. "Anomaly Reporting
System." Memorandum (May _27,
1988) , Satellite Services
Division, National
Environmental Satellite, Data,
and Information Service, Camp
Springs, Maryland.
Miller, D. & Hay, D. "Final
Report of the Artificial
Intelligence Techniques Study
for Environmental Products and
Computer System Operations. "
(September 1992) . U.S.
Department of Commerce Service
Order Number 40AANE201402.
Greenbelt, Maryland: COMSO,
Inc.
Morelli, R. "Using Knowledge
Engineering to Teach Science."
IEEE Expert 5(4) (August 1990):
74-78.
116