基于VMD和LSTM模型的航空液压管路卡箍故障诊断

航空发动机液压管路-卡箍系统中卡箍振动信号具有非线性和非平稳性的特点,难以从卡箍故障信号中准确识别出其故障类型。针对该问题,提出了一种基于变分模态分解(Variational Mode Decomposition,VMD)-长短时记忆(Long Short-Term Memory,LSTM)神经网络模型的智能故障诊断方法。首先,利用遗传算法对VMD的模态分量k值和惩罚因子α进行参数优化;然后,将优化后的VMD对卡箍故障振动信号进行分解处理;最后,将分解后的模态分量输入LSTM网络中进行特征学习,从而实现卡箍故障的识别。实验表明:该方法实现了对卡箍螺栓松动状态、根部断裂状态、衬垫磨损等3种典型故障的精准识别,故障总体识别准确率能够达到98.5%以上,有效地提高了航空液压管路卡箍故障识别的准确率。

Fault Diagnosis of Aviation Hydraulic Pipeline Clamp Based on VMD and LSTM Model

In view of the nonlinearity and non-stationarity of clamp vibration signal in aeroengine hydraulic pipeline-clamp system, it is difficult to accurately identify the fault type from clamp fault signal. An intelligent fault diagnosis method based on variational modal decomposition (VMD)-long and short term memory neural (LSTM) network model is proposed. Firstly, genetic algorithm is used to optimize the parameters of the modal component and penalty factor of VMD. Then, the optimized VMD is used to decompose the vibration signal of clamp fault. Finally, the decomposed modal component is input into LSTM network for feature learning, so as to realize the recognition of clamp fault. Experiments show that the clamp fault diagnosis method proposed in this paper can accurately identify three typical faults, such as bolt looseness, root crack and gasket wear, and the overall fault accuracy can reach over 98.5%, which effectively improves the accuracy of clamp fault identification of aviation hydraulic pipeline.