摘要
鉴于GFM RF-40型卧式数控径向精锻机生产过程中锻造主机的动力电机出现故障会导致严重的生产事故,提出了一种基于注意力增强一维卷积神经网络的精锻机电机故障诊断方法。该方法通过提取原始振动信号,然后对原始震动信号进行滑动窗口分割、转换Numpy数组等预处理,再利用一维卷积神经网络对预处理后的振动信号进行特征提取,将提取到的特征输入注意力机制层计算权重,根据注意力权重生成上下文向量,最后使用分类器输出每个故障类型的概率。1D-CNN-Attention使用数据训练通过反向传播算法不断调整参数,使模型能够准确的识别电机的故障类型。
Considering that the power motor of the forging main unit may fail during the production process of GFM RF-40 horizontal CNC radial precision forging machine may lead to serious production accidents,this paper proposes a motor fault diagnosis method for precision forging machines based on an attention-enhanced one-dimensional convolutional neural network.This method extracts the original vibration signal,and pre-processes it by applying the sliding window segmentation and converting Numpy arrays.Then,the pre-processed vibration signals are used in an one-dimensional convolutional neural network to carry out feature extraction,and the extracted features are input into an attention mechanism layer to calculate the weights.Based on the attention weights,a context vector is generated,and a classifier is finally used to output the probability for each type of fault.The 1D-CNN-Attention continuously adjusts parameters through data training by backpropagation algorithm,which can make the model accurately identify fault type of the motor.
作者
刘忠兴
曹立文
Liu Zhongxing;Cao Liwen(School of Mechanical and Electrical Engineering,Heilongjiang University,Harbin 150080,China)
出处
《防爆电机》
2025年第4期107-112,共6页
Explosion-proof Electric Machine
基金
黑龙江省自然科学基金项目(SS2021C005)
黑龙江省重点研发计划项目(GZ20220121)。
关键词
卷积神经网络
注意力机制
电机故障诊断
Convolutional neural networks
attention mechanisms
motor fault diagnosis
