Research On Condition Monitoring And Fault Diagnosis Of Marine Diesel Engine Based On Performance Parameters

As a thermal system with nonlinear,multi-variable coupling and complex functional structure,diesel engine will inevitably suffer performance degradation or failure during continuous operation.Therefore,monitoring the operating status of diesel engines throughout the life cycle,and performing real-time,quantitative,scientific,and accurate assessment and diagnosis is an important foundation for the safe and reliable operation of diesel engines,and has important research significance for scientifically formulating ship operation and maintenance strategies.This paper takes the main engine of the "Dazhi" smart ship as the research object,and carries out related research around the establishment of diesel engine simulation model,the acquisition of fault data,state monitoring and intelligent fault diagnosis methods.The main work is as follows:Firstly,a simulation model of the Win GD5X52 low-speed diesel engine is established according to the project requirements and the volumetric modeling method.The common failure mechanism is studied and a fault implantation scheme is formulated.The simulation research was carried out;secondly,a fuzzy health degree model was established by combining the fuzzy function and the kernel density estimation method to monitor and quantitatively evaluate the performance degradation of the diesel engine;at the same time,an adaptive dynamic simulation model was established to carry out all-round high-fidelity monitoring and evaluation of the diesel engine.However,the scarcity and imbalance of diesel engine fault data in the actual process seriously restrict the accuracy of fault diagnosis.In this paper,the support vector machine(SVM)with the best classification effect is selected as the base classifier,and it is found that random oversampling and mixed The diagnostic results after sampling are improved by 10.50% and 9.20% respectively compared with the training samples without processing.The problem of unbalanced sample categories can be effectively resolved by resampling the training set.For the unbalanced sample labeling problem,the standard Tri-After the optimization of the training algorithm,the average classification accuracy rate is increased by 19.10%,which maximizes the utilization rate of fault data and the accuracy and generalization of fault diagnosis.In summary,this article has carried out research from the three perspectives of diesel engine simulation modeling,health assessment and fault diagnosis,established a complete simulation model,designed a scientific health assessment plan,formulated a reasonable fault diagnosis strategy,and realized the diesel engine in principle.The effective coupling of performance analysis-health assessment-fault diagnosis methods,functionally integrates intelligent algorithms and simulation models to achieve detailed and accurate fault diagnosis,provides theoretical guidance for formulating scientific and effective diesel engine maintenance strategies,and promotes diesel engine fault diagnosis.The transition from preventive maintenance to predictive maintenance has important theoretical significance and practical value.