Knowledge Requirements and Architecture for an Intelligent Monitoring Aid that Facilitate Incremental Knowledge Base Development

Author(s):  
Ellen J. Bass ◽  
Ronald L. Small ◽  
Samuel T. Ernst-Fortin
1990 ◽  
Vol 80 (6B) ◽  
pp. 1833-1851 ◽  
Author(s):  
Thomas C. Bache ◽  
Steven R. Bratt ◽  
James Wang ◽  
Robert M. Fung ◽  
Cris Kobryn ◽  
...  

Abstract The Intelligent Monitoring System (IMS) is a computer system for processing data from seismic arrays and simpler stations to detect, locate, and identify seismic events. The first operational version processes data from two high-frequency arrays (NORESS and ARCESS) in Norway. The IMS computers and functions are distributed between the NORSAR Data Analysis Center (NDAC) near Oslo and the Center for Seismic Studies (Center) in Arlington, Virginia. The IMS modules at NDAC automatically retrieve data from a disk buffer, detect signals, compute signal attributes (amplitude, slowness, azimuth, polarization, etc.), and store them in a commercial relational database management system (DBMS). IMS makes scheduled (e.g., hourly) transfers of the data to a separate DBMS at the Center. Arrival of new data automatically initiates a “knowledge-based system (KBS)” that interprets these data to locate and identify (earthquake, mine blast, etc.) seismic events. This KBS uses general and area-specific seismological knowledge represented in rules and procedures. For each event, unprocessed data segments (e.g., 7 min for regional events) are retrieved from NDAC for subsequent display and analyst review. The interactive analysis modules include integrated waveform and map display/manipulation tools for efficient analyst validation or correction of the solutions produced by the automated system. Another KBS compares the analyst and automatic solutions to mark overruled elements of the knowledge base. Performance analysis statistics guide subsequent changes to the knowledge base so it improves with experience. The IMS is implemented on networked Sun workstations, with a 56 kbps satellite link bridging the NDAC and Center computer networks. The software architecture is modular and distributed, with processes communicating by messages and sharing data via the DBMS. The IMS processing requirements are easily met with major processes (i.e., signal processing, KBS, and DBMS) on separate Sun 4/2xx workstations. This architecture facilitates expansion in functionality and number of stations. The first version was operated continuously for 8 weeks in late-1989. The Center functions were then transferred to NDAC for subsequent operation. Later versions will be distributed among NDAC, Scripps/IGPP (San Diego), and the Center to process data from many stations and arrays. The IMS design is ambitious in its integration of many new computer technologies, but the operational performance of the first version demonstrates its validity. Thus, IMS provides a new generation of automated seismic event monitoring capability.


1990 ◽  
Vol 80 (6B) ◽  
pp. 1852-1873 ◽  
Author(s):  
Steven R. Bratt ◽  
Henry J. Swanger ◽  
Richard J. Stead ◽  
Floriana Ryall ◽  
Thomas C. Bache

Abstract The Intelligent Monitoring System (IMS) integrates advanced technologies in a knowledge-based distributed system that automates most of the seismic data interpretation process. Results from IMS during its first 8 weeks of operation (1 October through 25 November 1989) are analyzed to evaluate its performance. During this test period, the IMS processed essentially all data recorded by the NORESS and ARCESS high-frequency arrays in Norway. The emphasis was on detection and location of regional events within 2,000 km of these arrays. All events were reviewed and corrected if necessary by a skilled analyst. The final IMS Bulletin for the period includes 1,580 regional events (∼280 events/day). Approximately 55 per cent were smaller than MLg 1, with the largest just over MLg 3. Comparison of IMS locations in southern Finland and northwestern USSR (800 to 900 km from both arrays) with event locations from the University of Helsinki's local network bulletin are used to assess the detection and location capabilities of the system. Two or more phases (minimum needed to locate) were detected for 96 per cent of the events with magnitude greater than 2.5. The median separation between the IMS and Helsinki locations for all common events was 23.5 km. A consistent bias in arrival-time and azimuth residuals was observed for events in small geographic areas, indicating that refined travel-time models and path corrections could further improve location accuracy. The knowledge base in this first version of IMS was based on analysis of NORESS data, and many of the errors in interpretation corrected by the analysts can be attributed to differences encountered when this knowledge is used to interpret ARCESS data. Nevertheless, nearly 60 per cent of the events appearing in the final bulletin are automatic solutions approved without change or moved (by analyst corrections) less than 25 km from the automatic locations. The IMS had the most difficulty interpreting the overlapping signals generated by closely spaced explosions commonly detonated at mines in the Kola Peninsula and northern Sweden. Using the knowledge acquisition facilities included in the system, the deficiencies responsible for these and other errors are isolated, leading to development of new knowledge to be incorporated in the next version of the IMS knowledge base.


1994 ◽  
Vol 33 (01) ◽  
pp. 60-63 ◽  
Author(s):  
E. J. Manders ◽  
D. P. Lindstrom ◽  
B. M. Dawant

Abstract:On-line intelligent monitoring, diagnosis, and control of dynamic systems such as patients in intensive care units necessitates the context-dependent acquisition, processing, analysis, and interpretation of large amounts of possibly noisy and incomplete data. The dynamic nature of the process also requires a continuous evaluation and adaptation of the monitoring strategy to respond to changes both in the monitored patient and in the monitoring equipment. Moreover, real-time constraints may imply data losses, the importance of which has to be minimized. This paper presents a computer architecture designed to accomplish these tasks. Its main components are a model and a data abstraction module. The model provides the system with a monitoring context related to the patient status. The data abstraction module relies on that information to adapt the monitoring strategy and provide the model with the necessary information. This paper focuses on the data abstraction module and its interaction with the model.


2017 ◽  
Vol 20 (1) ◽  
pp. 208-220
Author(s):  
J. F. Coll
Keyword(s):  

Sign in / Sign up

Export Citation Format

Share Document