基于变分模态分解与对抗多层感知相结合的铁塔局部螺栓松动分析OA

The bolts of transmission towers play a vital role in ensuring the structural stability and operation safety of the towers.Therefore,accurate and efficient bolt looseness detection technology is particularly important for evaluating the condition of these structures.In this paper,a method combining variational modal decomposition(VMD)and against multilayer perceptual networks(AT-MLP)is proposed to identify the looseness of transmission tower bolts.The acoustic sensor collects the voiceprint signal of the iron tower bolts,which is then decomposed using VMD to extract characteristic information,such as the center frequency of the voiceprint signal associated with loose bolts.The time-frequency information of the voiceprint is input into the MLP model as a characteristic parameter to identify the state of the bolts.The results show that the recognition accuracy for bolt looseness of 110 kV and 220 kV transmission towers is 88%and 80%,respectively,while the recognition accuracy for the normal state of the bolts is 100%and 98%,respectively.These results meet the actual operation and maintenance requirements of towers.The method proposed in this paper can accurately distinguish fastening bolts from loose bolts,which provides a basis for the state evaluation of transmission towers.