Research On Fault Diagnosis Of Marine Diesel Engine Based On Gradient Boosting Decision Tree

The marine diesel engine is an important marine power plant,and its accurate status identification and fault diagnosis are of great significance for the safe navigation of ships.With the development of information technology and intelligent algorithm,the fault diagnosis of marine diesel engine using data-driven and intelligent algorithm is an important part of intelligent ship.The research problem of marine diesel engine fault diagnosis lies in that it is difficult to obtain the fault state operating data,how to select the characteristic data and the low diagnostic accuracy of the fault diagnosis model.To solve these problems,this paper proposes a fault diagnosis scheme combining principal component analysis and particle swarm optimization gradient boosting decision tree.In terms of obtaining marine diesel engine operating data,the 9L34DF four-stroke medium-speed marine diesel engine is used as the research object,and the marine diesel engine is simulated by AVL BOOST.By comparing the simulation calculation results with the diesel engine bench test data,the usability and accuracy of the simulation model of the diesel engine are verified.The simulation scheme of diesel engine fault is designed.Four kinds of faults are selected:uneven oil supply of single cylinder,compression ratio decrease fault,air cooler efficiency decrease fault and supercharger efficiency decrease fault.In order to distinguish the degree of fault,each fault state is divided into light,medium and severe faults.The fault simulation scheme is designed and the experimental calculation is carried out.25 parameters,such as torque,power,average effective pressure,effective fuel consumption rate and cylinder explosion pressure,are selected as the characteristic data of fault diagnosis.The results are compared and analyzed.In terms of feature fusion and fault detection of diesel engine operation data,the principal component model of diesel engine normal operation data is established by using principal component analysis algorithm,and the number of principal components is determined by the method of cumulative variance contribution degree.By comparing the statistics of the new data in the principal component model with that in the residual subspace,we can determine whether the data belongs to the fault data.Using the fault data simulated by AVL BOOST to verify the PCA model,the experimental shows that PCA can effectively carry out feature fusion and fault detection.In terms of diesel engine fault diagnosis model,the gradient boosting decision tree fault diagnosis model is constructed.In order to solve the problem of hyperparameter optimization of gradient boosting decision tree,particle swarm optimization is used to optimize the learning rate and iteration times of gradient boosting decision tree,and the training and testing process of fault diagnosis model of gradient boosting decision tree is designed.In this paper,principal component analysis is used to feature fusion,and particle swarm optimization gradient boosting decision tree is used to build diesel engine fault diagnosis model,and using the fault state operation data simulated by AVL BOOST to verify the model,The results show that the diagnosis performance is significantly improved and the diagnosis accuracy is higher than the decision tree model and the traditional gradient boosting decision tree model,.The fault diagnosis system of gradient boosting decision tree for marine diesel engine is designed and implemented by using Python language.It has certain use and design reference significance.