Fault Diagnosis Of Subsea Multiplexed Electro-hydraulic Control System

Recent years the demands for oil and natural gas,which serves as a crucial part of modern industrialization,are rapidly increasing.Offshore oil and gas production has become a hot spot appealed to the industry because of its abundant reservoir.Nowadays the Subsea Production System is the most used production mode controlled and monitored by Subsea Control System,and the multiplexed electro-hydraulic control is widely used.Due to its special conditions and the large delay of maintenance and repair,once an accident occurs,it will cause inestimable consequences.It is necessary to conduct the study on the fault diagnosis.This paper proposes three data driven methods of fault detection,condition evaluation and fault prediction.First,the study uses the Maximum Variance Unfolding which can solve the problem of results varying because of different kernel functions for nonlinear faults that many multivariate statistical methods face.In order to deal with the difficulty of mapping the new data,an incremental improvement method is proposed,and monitoring statistics are constructed to detect the fault.Then,the Continuous Hidden Markov Model is introduced for the system condition monitoring and evaluation,which can generate a likelihood probability for each observation sequence.A performance index and a change index are calculated which measure the deviation from the normal state and quantify the degree of the state changes respectively.Finally,the Auto Regressive Moving Average model is used for fault prediction by forecasting the monitoring statistics.This paper applies Empirical Mode Decomposition because of the nonstationarity and the moving window to estimate the parameters,also a direct multi-step prediction method is presented,which all can improve the accuracy of prediction.