Fault Classification for Unmanned Surface Vehicles using Supervised Learning Methods

Abstract

The key role of fault detection and diagnosis in the nonlinear system is to maintain safety, reliability, and survivability in case of system failure during a fault. The main idea behind this paper is to explore unmanned surface vehicles (USVs) performance under faulty conditions. K-Nearest Neighbour method has been chosen for faults detection, identification, and classification. A ball balancer system, which is a laboratory setup, has been considered for analysis of faults occur. The translational motion-related control of USVs is represented by ball movement on a plate in the x-y direction. The sway motion of USVs corresponds to x-axis movement of the ball while surge represents the y-axis motion of the ball on a plate. For algorithm implementation, the necessary fault data has been taken from the position of the ball, plate angle, and motor input voltage after inserting the fault in the system. For performance analysis, the results of Simulink and hardware has been validated.