Reliability and replacement analysis of Great Lakes marine diesels.

Abstract

Reliability and replacement characteristics of Great Lakes marine diesel engines have been studied. A Colt-Pielstick PC2-400 series marine diesel engine has been used as a prototype for our modeling. Failure Modes and Effects, and Fault Tree analyses for our engine have been performed. Censored field data have been processed from six identical engines of the above type. Parameters of Weibull and Exponential PDF's have been evaluated for seven vital system components, namely connecting rod bearing, cylinder head, cylinder jacket, cylinder liner and o-ring, cylinder piston, and fuel cam. Reliability and hazard functions for these items have been derived and discussed using both above PDF's. Exponential PDF's have been found to be inappropriate for representations of components failure times. Reliability based models have been developed and implemented to rationalize current winter layup replacement practices. Two systems have been considered: One for a ship equipped with one engine only and another for a two-engine ship. Incorporating the age dependent nature of system failure characteristics, a semi-Markov competing-process approach has been used in our models, where system failure behavior has been treated as a race among engine components. Howard's one-set competing process model has been implemented and extended to two sets of competing processes. A recursive iteration procedure has been used in the expected cost calculation. An efficient enumeration procedure has been presented to select the replacement policy which produces the minimum expected cost for an operating season. Computer codes have been developed using the above models, and several examples have been examined. Sensitivity analyses have been performed for several parameters for which we had insufficient or no information from the industry to see the influence of their variation on our expected costs and corresponding winter layup policies. We have observed that current replacement practices are conservative. Our results have also indicated that the total expected cost function changes almost linearly with time for our parameters. However, this linearity is very sensitive to the length of operating season. As far as replacement policies are concerned, there is a group of good replacement policies immediately following the best replacement policy; differences between the total expected costs resulting from these policies are not very significant for some cases within our parameter range. Besides, the value of the minimum expected cost function is very sensitive to component ages, and operating costs.