摘要
针对水下无人航行器(Unmanned Underwater Vehicle,UUV)电机故障诊断中现有方法依赖人工特征提取、智能诊断潜力无法充分发挥的问题,提出一种基于双流CNN-LSTM的故障诊断模型。该模型采用卷积神经网络作为特征提取器,无需复杂的预处理步骤,能够自动并行地学习原始信号的低频趋势与高频细节特征,从而实现实时电机状态监测。随后,基于长短期记忆网络的分类器利用提取的特征深入挖掘时序依赖关系,以识别电机故障。试验基于自主搭建的UUV电机故障模拟平台,设置了多种转速与负载工况以验证模型性能。结果表明,该方法能够高效诊断UUV电机中的六种典型状态,平均诊断准确率达到97.22%。试验证明,该模型在UUV电机故障诊断领域具有良好的有效性和鲁棒性。
To address the limitations of existing methods for underwater unmanned vehicle(UUV)motor fault diagnosis,which rely on manual feature extraction and do not fully leverage the potential of intelligent diagnosis,a two-stream CNN-LSTM fault diagnosis model is proposed.The model employs convolutional neural networks as feature extractor,which can learn the low frequency trend and high frequency detail features of the original signal without complex pre-processing steps,making real-time motor status monitoring possible.Afterwards,the classifier based on the long short-term memory network uses these features to explore temporal dependencies and identify motor faults.Experiments are conducted on a self-constructed UUV motor fault simulation platform,and the performance of the model is validated by setting multiple speeds and load conditions.The results show that this method can efficiently diagnose six typical states in UUV motors and achieve an average diagnostic accuracy of 97.22%.These findings demonstrate the model's effectiveness and robustness in UUV motor fault diagnosis.
作者
陈雪倩
沈钧戈
白俊强
谭浩声
黄浩然
CHEN Xueqian;SHEN Junge;BAI Junqiang;TAN Haosheng;HUANG Haoran(Unmanned System Research Institute,Northwestern Polytechnical University,Xi′an,710072;National Key Laboratory of Unmanned Aerial Vehicle Technology,Northwestern Polytechnical University,Xi′an,710072)
关键词
水下无人航行器
电机
人工智能
故障诊断
卷积神经网络
长短期记忆网络
underwater unmanned vehicle
motor
artificial intelligence
fault diagnosis
convolutional neural networks
long short-term memory
