Research On Data-driven Marine Main Engine Operation And Maintenance Decision-making Method

Marine diesel engine is the main power system of the ship.The research on its operation and maintenance decision-making method can effectively avoid major economic losses and disastrous consequences.However,as a complex system,the diesel engine is affected by multiple factors such as environments,personnel operation and complex working conditions,so it is difficult to effectively monitor its operating state.Therefore,in order to improve the monitoring ability of the main engine,this paper will study the condition monitoring and health management methods of the marine diesel engine.In order to monitor the operation of the main engine in real time,the parameter of fuel consumption is selected to represent the overall performance of the diesel engine and predict its values.Considering the complexity of diesel engine working conditions,the prediction model only using a single working condition will have poor results.Therefore,the working condition clustering method is adopted for each characteristic parameter to obtain better clustering characteristics.The improved CDA-RNN model is used to predict the fuel consumption for the input characteristic parameters,and the mapping relationship between each characteristic parameter and fuel consumption is established.The experimental results show that the combination of working condition clustering and the improved CDA-RNN model can effectively predict the fuel consumption.The vibration signals of diesel engine components will change with the change of working conditions.In order to realize the fault diagnosis of main engine components under multiple working conditions,the CNN adopts the Inception structure,and deeply identifies the fault characteristics through the parallel of multiple convolution kernels.The fault identification ability can be further improved by adding ResNet on the basis of the Inception block.In addition,in order to reduce the influence of noise on the vibration signal,singular value decomposition is used to reduce the noise of the signal.Experiments show that compared with the traditional CNN model,the SVDInception-ResNet model can effectively improve the accuracy of fault diagnosis in single load,variable loads and the noisy environments.When making maintenance decisions for diesel engine components,considering the actual situation that the components can not be repaired as new,so the fault rate increasing factor and the service age decreasing factor are introduced.At the same time,the optimal maintenance time of a single component can be solved by taking the lowest maintenance cost per unit time as the goal.However,the best maintenance time for a single component may not be the best maintenance time for the diesel engine system.Each downtime only for a single component will lead to excessive downtime.Therefore,the concept of opportunistic maintenance is introduced,and the opportunistic maintenance threshold of each component is solved by the particle swarm optimization algorithm.It is proved that the opportunistic maintenance can effectively reduce downtime and the total maintenance cost.Combined with the research content of this paper and the needs of shipping companies for the operation and maintenance decision-making methods of the marine main engine,the marine main engine operation and maintenance decision-making prototype system is designed and developed,which provides technical support for shipping companies to carry out the intelligent operation and maintenance of the main engine.