Application of Adversarial Network Model in Robot Inspection Heterophony Detection

Abstract

With the rapid development of industrial technology, inspection robots need to monitor more and more equipment during inspection, and the demand for fault detection of equipment is increasing. It is particularly important to detect abnormal sounds of equipment such as air compressors in gas collecting stations. The deep convolutional ADversarial network is applied to the abnormal sound detection of air compressor, and the network module of optimized feature learning is introduced before the spectrograph learning of the network, and the structural similarity function is introduced to finally distinguish whether the sound is normal or not. Through simulation experiments, the improved deep convolutional adversarial network algorithm realizes abnormal sound detection of air compressor in gas collecting station, and the discrimination accuracy of abnormal sound is enhanced, which achieves the purpose of abnormal monitoring through equipment sound in actual industry.