Research On Fault Feature Extraction And Diagnosis Method Of Ship Engine Room Equipmen

With the development of automation technology,Internet technology and Internet of Things technology,the degree of automation of modern ships is increasing day by day.The development of automated ships,artificial intelligence algorithms,and "big data"technologies continue to drive the development of smart ships.In the development of intelligent ships,the safety and reliability of the ship’s intelligent engine room system directly affect the safety of ship navigation.In order to improve the safety and reliability of intelligent ships,it is necessary to carry out research on feature extraction and fault diagnosis methods for ship engine room equipment.Although domestic and foreign experts and scholars have achieved certain results in the feature extraction and fault diagnosis of intelligent cabin equipment,there are still some problems to be solved,mainly including the comparison and optimization of vibration signal feature extraction methods and feature selection methods,lack of key correlation Signal reciprocating mechanical fault diagnosis problem,lack of fault data for large diesel engines.In view of the existing problems,this dissertation studies from four aspects:vibration signal feature extraction method,feature selection method,fault diagnosis method for reciprocating mechanical equipment and condition monitoring model of large two-stroke diesel engine.The main research contents include:(1)In order to extract the fault features of the rotating machinery equipment in the engine room,based on the traditional signal decomposition method,an MTAD signal processing method is proposed.EMD,EEMD and VMD methods are used to process the original vibration analog signal,such as noise reduction and feature extraction,and the comparison with the MTAD method shows the effectiveness of the MTAD method.The four methods are compared again using the fan fault simulation test data.The results show that the order of the average errors of the four methods from high to low is:EMD>EEMD>VMD>MTAD,and the recognition accuracy of ship fan faults are:76.79%,83.40%,88.40%and 94.63%respectively.(2)In order to reduce the data dimension after feature extraction and improve the operation speed and accuracy of the classifier,a relative confusion feature extraction method is proposed.This method is combined with SVM in the ship fan fault simulation test,and the highest fault identification accuracy can reach 94.41%.Compared with fuzzy entropy feature selection,Fisher score feature selection,Laplace feature selection and no feature selection,the accuracy rate is higher.They were increased by 0.56%,0.66%,0.51%,and 0.51%,respectively.At the same time,under the premise of the highest accuracy,the calculation time can be shortened by 17.59%.(3)Aiming at the problem of lack of key phase signal in the fault diagnosis of reciprocating machinery,a time-domain alignment processing method is proposed in this dissertation.In order to verify the effectiveness and feasibility of the method,this dissertation takes the ship 4135 high-speed diesel engine as the test object,and carries out the valve clearance fault diagnosis test.Combining continuous wavelet transform and CNN neural network for fault diagnosis,the results show that under the set fault conditions,the model can recognize the fault with 100%accuracy,which can effectively realize the valve clearance fault diagnosis of high-speed diesel engine.(4)Aiming at the problem that it is difficult to carry out fault simulation test and obtain fault data for large two-stroke diesel engines of ships,a state detection model based on adaptive kernel regression is proposed in this dissertation.This method can establish an empirical model based on the monitoring data of the normal state of the diesel engine,then reconstruct each observation vector,and calculate the mean square error value between the reconstructed vector and the observation vector,which can reflect the difference between the current diesel engine state and the normal state,so as to calculate the mean square error value between the reconstructed vector and the observation vector.To achieve the purpose of condition monitoring.Through GT-Power software,based on the test bench data of MAN B&W 6S35ME-B9 diesel engine,the simulation model of diesel engine is established.Five common diesel engine faults,such as insufficient fuel supply,high intercooler temperature,advanced injection angle,delayed injection angle,and exhaust blockage,are set for the simulation model.In the simulated condition monitoring test,the condition monitoring model can timely alarm the abnormal condition of the diesel engine,which proves that the AAKR-based condition monitoring model is suitable for large two-stroke diesel engines lacking fault data.