PREDICTING THE RELIABILITY OF A REPAIRABLE SYSTEM WITH COMPETING FAILURE MODES

Virtually all of the systems are bound to failure upon usage for certain duration, and
these systems can be categorized as repairable and non-repairable system. Repairable
system is one, which can be restored to get back to work fully by any action such as
parts replacements or changes to adjustable settings other than replacement of the entire
system. While non-repairable system is one, which cannot or not cost effective to be
repaired, such as the microprocessor, light bulb. In this project, research on the
reliability of repairable system was focused, as most of the machines/systems in the
industry field are complex and made from several to millions of parts. The main
objective of this project is to analyze the failure data of centrifugal pumps that were
provided by a petrochemical plant and to develop a reliability model to predict the
failure occurrences for each failure modes. With accurate failure prediction, the plant
will be able to schedule preventive maintenance accurately and thus avoiding the costly
corrective maintenance. Failure data of centrifugal pumps was collected and segregation
was done to narrow the scope of study. Weibull parameters, cumulative number of
failures, MTBF, etc for each failure were analyzed by using Weibull++ 7, while the
reliability of the pumpsystem was modeled by analytical analyses and simulations using
BlockSim 7. The centrifugal pumps from the plant are predicted to have an average
MTBF of 2.81 years. In conclusion, accurate failure prediction on centrifugal pump is
critical so that appropriate actions can be taken to utilize the components folly and most
importantly,minimized the chance ofbreak down ofthe pump.