Analysis of plant facilities to determine reliability rules regarding repair and maintenance
The purpose of this study was to test a probabilistic failure model against a sampling of process plant equipment. From this failure model a decision rule was to be developed to be a guide in making preventive maintenance decisions. Plant data was procured on the failure characteristics of steam turbines, and the probabilistic failure model was constructed according to the reliability mathematics developed by Bazovsky (2). It was concluded that there was no wear-out or old age characteristic displayed by these turbines with repeated maintenance. The turbines displayed a constant failure rate over the life of each repair. This is to say that upon being serviced the turbine was restored to new condition with its reliability again as good as with a new machine. A decision rule was devised such that the probability of operation for the turbine should be greater than twice that for the system as a whole, and greater than 50% at any time of system maintenance. Otherwise maintenance should be scheduled for the turbine. This 50% probability occurs when the operating time for the turbine is 0.693 of the mean time to failure for the turbines. If the failure rate for the turbines is greater than half of the failure rate for the system, the turbine in the system is the more likely cause of a system failure. Maintenance on the system should then be scheduled according to the reliability of the turbine which is the more critical item. A cost model based on Senju's optimal cost model is proposed. Additional cost data are needed for the application of this model. An extension of Senju's model is recommended to apply to the group of items in a system with consideration of series and parallel units.