基于声振信号的多分支动态卷积神经网络风机齿轮箱故障识别  

Multi-branch Dynamic Convolutional Neural Network Fault Identification of Fan Gearbox Based on Acoustic and Vibration Signals

作　　者：赵帅 陈长征[1] ZHAO Shuai;CHEN Changzheng(School of Mechanical Engineering,Shenyang University of Technology,Shenyang Liaoning 110870,China)

机构地区：[1]沈阳工业大学机械工程学院,辽宁沈阳110870

出　　处：《辽宁科技学院学报》2025年第1期10-13,76,共5页Journal of Liaoning Institute of Science and Technology

摘　　要：现有的风机齿轮箱故障诊断方法主要基于单一振动信号,所收集到的信息不够全面,易受外界干扰,导致风电机组齿轮箱故障诊断结果欠佳。针对上述问题,文章提出了一种基于声振信号联合的多分支动态卷积神经网络故障诊断方法。在该方法中,利用多个传感器采集风电机组齿轮箱在不同故障模式下的声振信号;随后,将声振信号输入到针对其特点所设计的卷积神经网络进行学习训练,完成在声振双信号中的自主学习并提取能区分风机齿轮箱故障类型的信号特征;最后,将声振双信号的特征拼接并由Softmax函数完成对应特征的识别。对多信号的深度学习故障诊断的仔细分析发现,由单一振动信号增加到声振双信号的深度学习故障诊断方法具有更高的准确性。实验结果表明,该方法在风机齿轮箱故障诊断方面展现出良好的准确性,具有良好的鲁棒性和自适应性。The existing fault diagnosis methods of wind turbine gearbox are mainly based on a single vibration signal,so the col⁃lected information is not comprehensive enough,and the fault diagnosis results of wind turbine gearbox are vulnerable to external interference.To address this issue,the current paper proposes a multi-branch dynamic convolutional neural network fault diagno⁃sis method based on the combination of acoustic and vibration signals.In the method,a number of sensors are used to collect the sound and vibration signals of wind turbine gearbox under different fault modes.Then,the acoustic and vibration signals are fed into the convolutional neural network designed according to their characteristics for learning and training,in order to complete the independent learning from the acoustic and vibration signals and to extract signal features that can distinguish the fault types of the fan gear box.Finally,the features of both acoustic and acoustic signals are spliced and the corresponding features are identified by Softmax function.Through careful analysis of multi-signal deep learning fault diagnosis,it is found that the deep learning fault di⁃agnosis method which increases from single vibration signal to acoustic and vibration dual signal has higher accuracy.The experi⁃mental results show that the method has satisfactory accuracy,robustness and adaptability in the fault diagnosis of fan gear box.

关 键 词：声振信号 风机齿轮箱 故障诊断 深度学习 

分 类 号：TH165[机械工程—机械制造及自动化]