Ad. Lunardhi et Km. Passino, VERIFICATION OF QUALITATIVE PROPERTIES OF RULE-BASED EXPERT-SYSTEMS, Applied artificial intelligence, 9(6), 1995, pp. 587-621
Citations number
33
Categorie Soggetti
System Science","Computer Science Artificial Intelligence","Engineering, Eletrical & Electronic
Frequently expert systems are developed to operate in dynamic environm
ents where they must reason about time-varying information and generat
e hypotheses, conclusions, and process in puts that can drastically in
fluence the environment within which they operate. For instance, exper
t systems used for fault diagnosis and fault accommodation in nuclear
power plants reason over sensor data and operator inputs, form fault h
ypotheses, make recommendations pertaining to safe process operation,
and in crisis situations could generate command inputs to the process
to help maintain safe operation. Clearly, there is a pressing need to
verify and certify, that such expert systems are dependable in their o
peration and can reliably maintain adequate performance levels. In thi
s article we develop a mathematical approach to verifying qualitative
properties of rule-based expert systems that operate in dynamic and un
certain environments. First, we provide mathematical models for the ex
pert system (including the knowledge-base and inference engine) and fo
r the mechanisms for interfacing to the user inputs and the dynamic pr
ocess. Next, using these mathematical models, we show that while the s
tructure and interconnection of information in the knowledge base infl
uence the expert system's ability to react appropriately in a dynamic
environment, the qualitative properties of the full knowledge-base/inf
erence engine loop must be considered to fully characterize an expert
system's dynamic behavior: To illustrate the verification approach, we
show how to model and analyze the qualitative properties of rule-base
d expert systems that solve a water-jug filling problem and a simple p
rocess control problem. Finally, in our concluding remarks, we highlig
ht some limitations of our approach and provide some future directions
for research.