残差胶囊网络在旋转机械故障诊断中的应用研究

打开文本图片集

摘要：针对旋转机械中的故障诊断需求，在传统的胶囊网络中引入残差块和模糊C均值聚类算法，构建残差胶囊网络故障诊断模型。在残差胶囊网络的基础上，引入注意力机制和G-K动态路由算法，构建注意力胶囊网络故障诊断模型。仿真分析表明：两种模型都能对故障进行精准测试，具有较强的表达能力和泛化能力。

关键词：旋转机械；故障诊断；胶囊网络；残差块；注意力机制

中图分类号：TP277文献标志码：B文章编号：1671-5276（2024）06-0244-03

Abstract：To meet the fault diagnosis requirements in rotating machinery， a residual capsule network fault diagnosis model is constructed by introducing residual blocks and fuzzy C-means clustering algorithm in traditional capsule networks. On the basis of residual capsule network， attention mechanism and G-K Dynamic routing algorithm are introduced to build a fault diagnosis model of attention capsule network. Simulation analysis shows that both models can accurately test faults and have strong expressive and generalization abilities.

Keywords：rotating machinery; fault diagnosis; capsule network; residual block; attention mechanism

0引言

在工业4.0时代的背景下，机械故障诊断的核心目标是利用先进的技术手段提高机械设备使用寿命，减少由于机械故障带来的经济损失［1］。（剩余3762字）