Objectives Complicated non-linear relationships exist among the subsystems of a ship's main engine. For a large amount of data collected by monitoring points in a short time, the traditional fault diagnosis method cannot efficiently complete the task. Taking the fuel system of the ship's main engine 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.
Methods First, the rough set theory in data mining is introduced into the traditional Support Vector Machine (SVM) diagnostic model, and the discretized data is reduced by the difference matrix. A SVM classifier is established between every two kinds of faults to construct a DAG-SVM topology network. Then, based on the classification accuracy of the classes, the positions of the root nodes and other leaf nodes in the DAG are optimized, thereby effectively avoiding the "accumulation of errors". Finally, based on a super-large tanker simulation, numerical and experimental analysis is performed.
Results The experimental results show that the fault diagnosis method based on rough set theory and optimized DAG-SVM can effectively diagnose faults in the main engine of a ship with classification accuracy 3.38% higher than that of traditional DAG-SVM, as well as time consumption reduced by 2.42 seconds.
Conclusions This diagnosis method has certain reference value for research on the fault diagnosis of the main marine engines, and can also provide data support for the application of SVM in the classification of other small samples.
DownLoad: