Design for the Maintainer: Projecting Maintenance Performance from Design Characteristics.
Abstract
We hypothesize that the maintenance activity imposed by an equipment may be effectively projected from one or more computed reference maintenance strategies. These include multi-variable strategies (optimum or 'expert' approaches), single-variable strategies (time-dominant, reliability-dominant, information-dominant, component-dominant), and a stochastic strategy in which tests are selected at random. The optimum performance strategy and the random testing strategy provide bounds on the expected maintenance performance. The single-variable approaches are suspected to be reasonable approximations of human activity under various conditions. The constituents of maintenance performances generated from these strategies, and their associated performance times, are shown to be a direct function of system design. Computed manual times for each of these approaches are presented for one equipment. These preliminary data suggest that maintenance time may be considerably less sensitive to fault diagnosis strategy than expected. Our work leads us to view a troubleshooter as a strategically flexible, data-driven, and opportunistic problem solver. We describe some recent artificial intelligence models of problem solving which support our conception of the troubleshooter. Such models provide a basis from which the computed strategies described elsewhere could arise.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1981
- Accession Number
- ADA102513
Entities
People
- Douglas M. Towne
- Michael R. Fehling
- Nicholas A. Bond
Organizations
- University of Southern California