1999 AAAI Spring Symposium on
AI in Equipment Maintenance Service and Support

Palo Alto, California • March 22-24, 1999

Symposium 
Dates, times, and places 
Format 
List of Accepted Papers 
Paper Submission 
About AAAI 

Equipment Service M&D 
Background 
Examples 

Artificial Intelligence 
Challenge problems 

Software and Data 
Existing software 
Useful datasets 

References 
FAQ 
Web sites 
Bibliography 

Contact 
Comments 
Subscriptions

As a result there is a strong and renewed emphasis on AI technologies that can be used to monitor products and processes, detect incipient failures, identify possible faults (in various stages of development), determine the preventive or corrective action, generate a cost-efficient repair plan and monitor its execution. The service market delivered will range from power generation equipment to aircraft engines, medical imaging systems, and locomotives, just to name a few. 

Monitoring equipment more efficiently can be accomplished in part by employing remotely monitored systems. Big strides have been taken to accomplish these goals for example for in-use monitoring of not only stationary equipment such as manufacturing plants or high end appliances, but also mobile systems such as transportation systems (vehicles, aircraft, locomotives, etc). While advances in hardware development make it possible to perform these tasks, there are new avenues for progress in AI techniques. Some of these approaches have their roots in efforts of years past while others arise from new challenges. 

Characteristics of  typical challenges for AI in Monotoring and Diagnosis (M&D) service can be categorized into input, model, and output. In particular, input questions try to deal with real-time data streams resulting from on-line monitoring equipment. They are required to handle: throughput constraints; noise; non-stationary systems (due to linear drifts or chaotic behavior); erroneous data; data compression and information extraction. Process and product modeling tasks attempt to tackle issues involving non-stationary systems which require constant model update (adaptation, learning). In addition, the signature identification in time-series leading to fault detection and identification needs to be addressed. Moreover, the detection of new (unaccounted for) faults/anomalies/state changes has to be dealt with. Models obtained from first principles and from empirical data render different output which can be deterministic or qualified by confidence level, probabilities, and other type of semantics for uncertainty characterization (randomness, imprecision, vagueness, inaccuracy, inconsistency, incompleteness). These different outputs have to be aggregated to a coherent response. This task can be addressed in information fusion schemes for such heterogeneous models. This workshop aims to address relevant AI technologies which address segmentation, classification, prediction, and decision making in particular in: 

-Adaptation to changing environments, 
-Decision making of autonomous systems (from a service point of view) 
-Information Fusion of various diagnostic modules to resolve conflicts and aggregate information expressing uncertainty in different domains 
-Knowledge extraction from symptom databases 
-Intelligent internet based agents for monitoring tasks 
-Maintenance planning 
-Corrective action planning 
-Trend performance analysis and prognostics, 
-Reliability and margin prediction 
-Machine learning to recognize and classify new system behavior, 
-Autonomous repair 
-Reconfigurability 
 

Organizing Committee

Piero Bonissone  
GE Corporate Research & Development 

Kai Goebel 
GE Corporate Research & Development 

George Vachtsevanos 
School of Electrical and Computer Engineering, Georgia Institute of Technology 
 

Version 1.0
Updated 10/21/98