Research On Fault Prediction Of Ship Fuel Pipeline Based On Deep Learning

With the marine industry’s development,the number of ship accidents caused by fuel pipeline failures is increasing.Once the fuel pipeline fails,the losses caused are also very staggering.Studying the problem of failure prediction of ship fuel pipeline can reduce accidents and improve ship fuel equipment stability.Based on the analysis and research of traditional forecasting methods,this article finds that traditional forecasting methods do not consider the connection between data samples and historical data.This problem leads to low accuracy and effectiveness of the prediction results.In order to solve the problem of ship fuel pipeline prediction,this thesis proposes a network model for ship fuel pipeline failure prediction.The better model is applied to the fuel supply unit’s remote operation and maintenance software platform system.The main research contents of the thesis are as follows:(1)Conduct in-depth analysis and research on the failure principle of fuel pipelines,and find that fuel pipelines will gradually degrade over time during application.The problem of fuel pipeline failure prediction is a time series prediction problem.According to the ship’s fault experience,the pressure of the supply pump in the pipeline is selected as the characteristic parameter of the fault prediction.The actual working condition experience verifies the characteristic parameter.(2)Propose a fuel pipeline fault prediction network model.Through the research and analysis of traditional forecasting methods,it is found that it does not consider the influence of historical data on the forecast results.Therefore,this thesis uses long and short-term memory(LSTM)and gated recurrent unit(GRU)network to solve fuel pipeline failure prediction.The network model has a particular cyclic memory unit structure,fully considering the influence of historical data on the prediction results.When exploring better fault prediction models,the theoretical analysis found that it is difficult to determine the optimal solution for network hyperparameters,such as the number of network layers,the number of layer nodes,and the time series size.Through the research and analysis of the network model optimized by the particle swarm optimization algorithm,the gate loop unit network model(PSO-GRU)fused with the particle swarm optimization algorithm is finally proposed to solve ship fuel pipeline failure.(3)Design software platform and analyze experimental results.Analyze the fuel supply unit’s hardware equipment and gradually build the remote operation and maintenance software system of the fuel supply unit through various modules.The network model proposed in the article is compared with experiments,and a better network model is explored through the analysis of experimental results.And embedded a better network model into the remote operation and maintenance software platform system of the fuel supply unit,and achieved good engineering practice results.