A Fault Location Algorithm Based on Convolutional Neural Network for Sensor System of Seafloor Observatory Network

The seafloor observatory network (SFON) covers an extensive area and consists of many network devices functioning in the abyssal environment, which make patrolling inapplicable to fault location in the marine setting. Moreover, finding faults like degradation of precision or zero drift would be rather difficult if such faults are only located by the warning message from a single sensor. To solve this problem and as per the features of SFON, we propose a fault location algorithm based on the convolutional neural network (CNN) for the data transmission system. This algorithm which takes a holistic perspective and considers the features of network device can monitor all the sensors in a unified and centralized way. The algorithm sets the CNN parameters according to the features of the research object, and normalizes the data of sensors to images. It first qualitatively judges a fault, and then recognizes its source and type. The new algorithm has higher precision on fault recognition than the support vector machine.