AI in Equipment Maintenance Service and Support Michael Halasz National Research Council Canada Plenary Session March 23, 1999 Stanford University
What? “renewed” interest in software tools and techniques to improve maintenance process –monitoring –diagnosis and/or prognosis –repair management (parts, location, tools, skills) –handling repercussions (e.g. rescheduling) –capturing and reapplying expertise
Why? cost –can exceed capital outlay (125% - regional jet) –difficult to control (wide swings) quality of service –intangibles (customer perception, side effects) maintenance “outsourcing” –manufacturer: reduced profit margins –operator: not primary mission –maintenance service provider: specialty, economy of scale, risk mitigation
Maintenance: an information rich environment source: equipment, humans, systems data: voluminous, messy, distributed, diverse knowledge: humans, documents, operating practices, software infrastructure: communications & hardware crosses organizational boundaries
applied AI to reason about situations –case-based reasoning –fuzzy logic –neural networks –induction –model based reasoning –rule-based –etc. Leverage through Innovative Information Technologies hybrid systems No Silver Bullet!
Symposium Format 1 tutorial – soft computing, multiple techniques 20 papers/talks –purely technical & application oriented 2 working sessions
Breakdown application areas –automotive, aircraft, HVAC, gas turbines, pumps,… –design for maintainability techniques –networks (belief, bayesian, neural) –MBR, CBR, rules –fuzzy logic –free text interpretation source –45% industrial, 30% academic, 25% blend
Summary Equipment Maintenance Service and Support poses many interesting IT challenges fertile area for AI