An integrated knowledge-based approach to maintenance control systems for automated manufacturing
Abstract
The characterization of the maintenance function in automated manufacturing systems has proven difficult either by analytical methods or by the application of general physical laws. The crux of the problem is attributable to the complexity and variability of maintenance activities which rely on diverse knowledge sources which are very much experience-based. In the event of unexpected failure, the complexity problem is further compounded due to the sophistication and multiplicity of interrelated circuits and mechanisms in automated systems. This research is concerned with the development of a methodological framework that aids the control of timely and efficient maintenance of modern manufacturing systems with major emphasis on the preventive function through malfunction prediction, detection, diagnostics and repair. The proposed five-step planning scheme uses the ICAM definition language commonly referred to as IDEF (only as a means to an end) to permit a systematic description of functional system characteristics of the environment under study. Implementation of this framework centers on the integration of three disparate knowledge sources: (i) model-based stochastic simulations--to generate expectations about the correct system behavior based on inputs and structure and induce knowledge from them, (ii) maintenance database--to maintain dynamic information in an ever-changing environment, and (iii) knowledge-based expert system--to integrate declarative, heuristic and procedural knowledge for maintenance control. The resulting model MCSAM (Maintenance Control Systems for Automated Manufacturing), provided with user interfaces, offers the benefit of taking advantage of both algorithmic and knowledge processing techniques for maintenance control activities. Algorithmic processing is used predominantly for data access, data monitoring, and data verification tasks. On the other hand, knowledge processing is used for those tasks for which algorithmic techniques are inadequate, specifically, for analysis, diagnosis, and recommendation tasks.
Degree
Ph.D.
Advisors
Barash, Purdue University.
Subject Area
Industrial engineering|Artificial intelligence
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.