基于模型的发动机旋转机械信号处理及故障分类方法

Model-based Signal Processing and Fault Classification Method for Engine Rotating Machinery

随着发动机旋转机械数据的指数级爆炸,传统的故障分类方法不仅成本高且效率低下,本文提出了一种全新的应用于发动机旋转机械的故障分类方法,结合旋转机械故障数据的特征与成熟的深度学习算法模型,设计出一种适合处理旋转机械振动信号的改进的门控循环单元模型——C-GRU模型。该模型通过将1D-CNN的优势与GRU的优势进行结合,利用一维卷积层对原始数据进行深层特征提取,凭借门控循环单元使得网络获得了“记忆”的功能,即使在噪声干扰下依旧能取得不俗的效果。

With the exponential explosion of engine rotating machinery data,the traditional fault classification method is not only cost of idle time but also low efficiency.This paper proposes a new fault classification method applied to engine rotating machinery.Combining the characteristics of rotating machinery fault data and mature deep learning algorithm mode.This paper designs a new improved gated recurrent unit(GRU)model,C-GRU model,which is suitable for processing vibration signals of rotating machinery.By combining the advantages of 1D-CNN with the advantages of GRU,the model uses a one-dimensional convolutional layer to extract deep features of the original data,and relies on the gated recurrent unit to make the network obtain the function of"memory",which can still achieve good results even under noise interference.