Feature Selection for Unmanned Surface Vehicle Fault Diagnosis Research and Experimental Verification

As interest in unmanned vessel research has increased, emphasis on autonomous navigation capability and the sustainability of normal operation has been highlighted. However, research on fault detection and diagnosis (FDD) for unmanned surface vehicles (USVs) is insufficient compared with research on land machinery or unmanned aerial vehicles. Among prior studies on USV FDD, intensive analysis or diagnostics for individual elements with high failure probability were barely conducted. This study focus on fault detecting features of a USV thruster, one of the most failure-prone elements other than internal causes or software malfunctions. A hypothesis for thruster fault detecting features was made based on general thruster-related equations; the features were tested on a stand-alone thruster test to assess the feature’s acceptability. However, a stand-alone thruster test lacks important effects such as structural vibratory influence of an actual USV, thruster load difference due to inflow velocity and wave-induced vibrations as the USV slides on the water’s surface. Hence, USV capable of acquiring fine quality real-time data for FDD practice was developed, and a surface driving test in a water basin was conducted to confirm the eligibility of the selected features. © ICROS 2022.