Marine Main Engine Fault Diagnosis Research Based On Rough Set And Optimized DAG-SVM

With the development of the automation and intelligence of the engine room,the failure mechanism of the machinery equipment in the engine room is complicated and changeable.As the key equipment in the engine room,the ship's main engine plays an important role in the safe navigation of the ship.There is a complex nonlinear relationship between many subsystems included in the ship's mainframe,and the large amount of data collected by many measuring points on the mainframe in a short time will greatly increase the computational overhead of the diagnostic system without processing,and the traditional fault diagnosis method is difficult to complete the task efficiently.Taking the main engine fuel system as the research object,a fault diagnosis method based on rough set theory and optimized directed acyclic graph-support vector machine(DAG-SVM)is proposed.First,the rough set theory in data mining is introduced into the traditional support vector machine(SVM)diagnosis model,and the discretized data is reduced through the difference matrix to establish a support vector machine classifier between each two types of faults,thereby constructing DAG-SVM topology network;then,based on the classification accuracy between classes,optimize the position of the root node and other leaf nodes in the directed acyclic graph,thereby effectively avoiding "error accumulation";finally,based on a very large tank simulator Carry out numerical experiment analysis,under the same conditions,carry out simulation experiments on four typical classification modes,which are 1-vs-1 SVM,1-vs-a SVM,DAG-SVM and the method in this paper.The simulation results show that the fault diagnosis method combining rough set and optimized DAG-SVM can effectively diagnose the ship's main engine fault,and its classification accuracy is improved by 3.4%compared with the traditional DAG-SVM method,and the time consumption is also reduced by 2.42 s,the method of this paper is far superior to 1-vs-1 SVM and 1-vs-a SVM in diagnostic accuracy and time consumption.This diagnostic method has certain reference value for the fault diagnosis research of the ship's main engine,and can also provide data support for the application of SVM in other small sample classifications.