Study on the Wear Failure Rule of Marine Diesel Engine Based on MCMC Method

By analyzing wear data of marine diesel engine cylinder head, a method of probability distribution prediction which combines wear updated data with historical data is proposed based on MCMC method by utilizing WinBUGS through Gibbs sampling. Firstly, great importance is attached to marine diesel engine involving marine safety. Secondly, prior values of gamma process are confirmed on the basis of historical data by MLE. Thirdly, MCMC method is introduced and with Bayesian network, prior parameters can be updated to posterior parameters by WinBUGS. Finally, an example is applied to validate the proposed method, and the posterior distribution matches the practice better. The proposed method satisfies the engineering requirement and provides a foundation for on-condition maintenance decision.