Dynamic maintenance scheduling using online information about system condition.

Traditional maintenance paradigms, such as scheduled or corrective maintenance, result in production loses caused by unscheduled maintenance due to unpredicted machine downtimes, and unnecessary maintenance performed. The use of online information about system condition will enable the maintenance to be performed only when it is needed and where it is needed in order to achieve maximum combined production benefit. First the prioritization of existing maintenance work orders is studied. A quantitative representation of the value contained inside the system is first presented. Then the most production-beneficial priority setting for the existing maintenance work orders is sought using the Genetic Algorithm. Comparison of the effects of this prioritization method with the effects by other existing maintenance prioritization methods showed that the most production-beneficial priority settings can be generated consistently. Maintenance scheduling for dedicated production system is studied secondly. Maintenance schedules are evaluated, incorporating predicted health information about each individual machine in the system. The score assigned to any possible schedule is calculated as the combined benefit of production and maintenance. It has been shown through simulated results, that Genetic Algorithm based search method can facilitate finding of enhanced maintenance schedules, enabling higher production benefits and lower maintenance cost. The utilization of adjusting machine throughput settings as a maintenance action has also been studied. The result from the analytical approach confirmed that production benefit can be increased when adjusting throughout settings is scheduled with maintenance actions. In order to find the schedule that will maximize the production benefit and minimize the maintenance cost, a novel Genetic Algorithm based optimization using matrix chromosomes and corresponding evolution operators is used. Convergence of the schedule optimization is achieved in the experiments and the corresponding combined production benefit is maximized. In conclusion, a new decision framework is proposed to improve maintenance decision making, by incorporating online information about machine conditions and system conditions. The proposed methods can be used to enable to integration of information, and the benefits of using such information are exemplified in three decision areas.Ph.D.Applied SciencesMechanical engineeringUniversity of Michigan, Horace H. Rackham School of Graduate Studieshttp://deepblue.lib.umich.edu/bitstream/2027.42/124983/2/3163974.pd