基于改进时间信息融合模型的液压管路故障诊断研究

Fault diagnosis of hydraulic pipeline based on improved time information fusion model

航空发动机液压管路故障信号中存在强大的噪声干扰,导致其诊断模型的故障识别率较低和诊断模型的泛化性不强。针对这一问题,提出了一种基于改进的时间信息融合模型的航空液压管路故障诊断方法。首先,基于循环神经网络原理,设计了正向和反向的时间信息融合的变形结构,构建出了航空液压管路时间信息融合模型,并通过引入LeakyReLU函数,对模型进行了改进;然后,将实测的一维航空管路时序数据集输入到改进的时间信息融合模型双向循环神经网络(Bi-RNN)中,进行了权重参数的更新;最后,基于同一实测的数据集,分别将其输入到改进的时间信息融合模型、长短期记忆神经网络(LSTM)、循环神经网络(RNN)、支持向量机(SVM)和反向传播神经网络(BPNN)5种故障诊断方法中,进行了训练和对比分析,对相关方法的优越性进行了验证。研究结果表明:利用改进的时间信息融合模型可以对液压管路健康状态和裂纹、凹坑等故障状态进行精准识别,并且准确率可以达到99.2%,总体的准确率和综合指标F 1-sore均可以提高5.1%;在综合性能、准确精度等指标上,改进时间信息融合模型明显优于其他故障诊断模型,可为航空发动机液压管路故障诊断提供一条新的思路,具有一定的工程应用价值。

Aiming at the issues of strong noise interference,low fault recognition rate,and weak model generalization in the fault signals of aviation engine hydraulic pipelines,an improved temporal information fusion model for fault diagnosis of the aero-engine hydraulic pipeline was proposed.Firstly,based on the principle of recurrent neural network,the deformation structure of forward and reverse time information fusion was designed,and the time information fusion model of aviation hydraulic pipeline was constructed,and the LeakyReLU function was introduced to improve the model.Then,the measured one-dimensional aviation pipeline time series data set was input into the improved time information fusion model bidirectional recurrent neural network(Bi-RNN)to update the weight parameters.Finally,based on the same measured data set,the proposed improved temporal information fusion model,long short-term memory neural network(LSTM),recurrent neural network(RNN),support vector machine(SVM)and back propagation neural network(BPNN)were respectively input into five fault diagnosis methods for training.The superiority of the proposed method was verified by comparative analysis.The research results indicate that the improved time information fusion method proposed in this article for hydraulic pipeline fault diagnosis has achieved accurate identification of hydraulic pipeline health status and fault status such as cracks and pits,with an accuracy rate of 99.2%.The overall accuracy and comprehensive index F 1-sore have been improved by 5.1%.In terms of comprehensive performance,accuracy,and other indicators,it is significantly superior to other fault diagnosis models,providing a new approach for the diagnosis of hydraulic pipeline faults in aviation engines,and has certain engineering application value.