面向螺丝锁附序列的多分辨率融合卷积神经网络

刘天宇; 周稻祥; 李明; 李心宇

doi:10.7652/xjtuxb202003020

您当前的位置：

首页 >

文章列表页 >

面向螺丝锁附序列的多分辨率融合卷积神经网络

更新时间：2025-07-09

- 面向螺丝锁附序列的多分辨率融合卷积神经网络
- Multi-Resolution Fusion Convolutional Neural Network for Screw Locking Series
- 西安交通大学学报 2020年54卷第3期页码：161-168+178
- 作者机构：
  
  太原理工大学大数据学院,太原,030024
- 作者简介：
- 基金信息：
  
  : 2019-09-12。作者简介: 刘天宇(1995—),男,硕士生;周稻祥(通信作者),男,讲师。基金项目: 国家自然科学基金资助项目(11771321);国家自然科学基金青年基金资助项目(11901423);山西省社会发展科技攻关计划资助项目(2017 03D321032)
- DOI：10.7652/xjtuxb202003020
  中图分类号： TP183
- 网络首发：2020-03-10，
  
  纸质出版：2020
- 稿件说明：
移动端阅览
刘天宇, 周稻祥, 李明, 等. 面向螺丝锁附序列的多分辨率融合卷积神经网络[J]. 西安交通大学学报, 2020,54(3):161-168+178.

Multi-Resolution Fusion Convolutional Neural Network for Screw Locking Series[J]. 2020, 54(3): 161-168+178.
刘天宇, 周稻祥, 李明, 等. 面向螺丝锁附序列的多分辨率融合卷积神经网络[J]. 西安交通大学学报, 2020,54(3):161-168+178. DOI： 10.7652/xjtuxb202003020.

Multi-Resolution Fusion Convolutional Neural Network for Screw Locking Series[J]. 2020, 54(3): 161-168+178. DOI： 10.7652/xjtuxb202003020.

摘要

为了准确识别螺丝锁附是否发生故障和具体故障类型

提出了一种多分辨率融合卷积神经网络。使用原始序列数据作为输入以提高识别速度和精度; 为了提取多尺度特征

分别在分辨率(数据长度)为4 000、2 000和1 000的特征向量上进行一维卷积运算; 在Fusion层通过上采样、下采样和1×1卷积等策略

将各分辨率特征向量融合得到3组新特征向量

使得该网络能够获得锁附序列的整体和局部特征信息; 在输出层使用类别加权交叉熵(CWCE)损失

通过为损失函数设置惩罚系数来加大对样本较少类别的惩罚力度

缓解了各类别数据不平衡的问题。收集了3 149条螺丝锁附序列

并在该数据集上进行了实验

结果表明:在6分类实验中

所提方法的准确率为96.00%

宏F1为93.93%

均高于其他方法; 在2分类实验中

所提方法的准确率为99.36%

CWCE损失的有效性得到了验证; 所提方法能够有效地判别锁附故障

并具有较好的实时性。

Abstract

For identifying whether a screw locking fault occurs and its type accurately

a multi-resolution fusion convolutional neural network is proposed. To improve recognition rate and accuracy

the network uses original series data as the input. To extract multi-scale features

1-D convolutional operation is performed on the feature maps with resolutions(data length)of 4 000

2 000 and 1 000

respectively. In the Fusion layer

the resolution feature maps are fused by upsample

downsample and 1-1 convolution to obtain three sets of new feature maps

so that the network can obtain the global and local information of locking series. In output layer

class weighted cross entropy(CWCE)loss is used to alleviate the issue of unbalanced data in different classes by setting a penalty coefficient for loss function to strengthen the punishment for small sample categories. 3 149 screw locking series datasets are collected

and experiments are performed on this dataset. In the experiments of six categor

关键词

Keywords

references

DIETTERICH T G. Machine learning research: four current directions [J]. AI Magazine, 1997, 18(4): 97 -136.

DIFILIPPO N M, JOUANEH M K. A system combining force and vision sensing for automated screw removal on laptops [J]. IEEE Transactions on Automation Science and Engineering, 2018, 15(2): 887-895.

罗川, 黄迟, 曾文绶, 等. 基于随机森林的智能锁螺丝机锁附判别研究 [J]. 济南大学学报(自然科学版), 2019(5): 410-416.

LUO Chuan, HUANG Chi, ZENG Wenshou, et al. Discrimination of intelligent screw machine based on random forest [J]. Journal of University of Jinan(Science and Technology), 2019(5): 410-416.

MYERS C, RABINER L, ROSENBERG A. Performance tradeoffs in dynamic time warping algorithms for isolated word recognition [J]. IEEE Transactions on Acoustics, Speech, and Signal Processing, 1980, 28(6): 623-635.

CORTES C, VAPNIK V. Support-vector networks [J]. Machine Learning, 1995, 20(3): 273-297.

BREIMAN L. Random forests [J]. Machine Learning, 2001, 45(1): 5-32.

KRIZHEVSKY A, SUTSKEVER I, HINTON G E. ImageNet classification with deep convolutional neural networks [J]. Communications of the ACM, 2017, 60(6): 84-90.

ZHU X X, HOU D N, ZHOU P, et al. Rotor fault diagnosis using a convolutional neural network with symmetrized dot pattern images [J]. Measurement, 2019, 138: 526-535.

LI D, LI X M, ZHAO J M, et al. Automatic staging model of heart failure based on deep learning [J]. Biomedical Signal Processing and Control, 2019, 52: 77-83.

SWAPNA G, KP S, VINAYAKUMAR R. Automated detection of diabetes using CNN and CNN-LSTM network and heart rate signals [J]. Procedia Computer Science, 2018, 132: 1253-1262.

BALOGLU U B, TALO M, YILDIRIM O, et al. Classification of myocardial infarction with multi-lead ECG signals and deep CNN [J]. Pattern Recognition Letters, 2019, 122: 23-30.

庞彦伟, 李潇, 梁金升, 等. 基于卷积神经网络的第一导联心电图心拍分类 [J]. 天津大学学报(自然科学与工程技术版), 2018, 51(10): 997-1004.

PANG Yanwei, LI Xiao, LIANG Jinsheng, et al. Classification of first lead electrocardiogram heartbeats based on convolutional neural networks [J]. Journal of Tianjin University(Science and Technology), 2018, 51(10): 997-1004.

原永朋, 游大涛, 渠慎明, 等. 卷积神经网络在心拍类识别中的应用 [J]. 计算机应用, 2018, 38(12): 3638-3642, 3648.

YUAN Yongpeng, YOU Datao, QU Shenming, et al. Application of convolution neural network in heart beat recognition [J]. Journal of Computer Applications, 2018, 38(12): 3638-3642, 3648.

RUSSAKOVSKY O, DENG J, SU H, et al. ImageNet large scale visual recognition challenge [J]. International Journal of Computer Vision, 2015, 115(3): 211-252.

SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognition [C]∥Proceedings of the 3rd International Conference on Learning Representations(ICLR). London, UK: ICLR, 2015: 1-14.

GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation [C]∥Proceedings of the 2014 IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Piscataway, NJ, USA: IEEE, 2014: 580-587.

SUN K, XIAO B, LIU D, et al. Deep high-resolution representation learning for human pose estimation [C]∥Proceedings of the 2019 IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Piscataway, NJ, USA: IEEE, 2019: 5693-5703.

REISS A, INDLEKOFER I, SCHMIDT P, et al. Deep PPG: large-scale heart rate estimation with convolutional neural networks [J]. Sensors, 2019, 19(14): 3079.

IOFFE S, SZEGEDY C. Batch normalization: accelerating deep network training by reducing internal covariate shift [C]∥Proceedings of the 32nd International Conference on Machine Learning(ICML). Princeton, NJ, USA: International Machine Learning Society(IMLS), 2015: 448-456.

LIN M, CHEN Q, YAN S. Network in network [C]∥Proceedings of the 2nd International Conference on Learning Representations(ICLR). London, UK: ICLR, 2014: 1-10.

KINGMA D, BA J. Adam: a method for stochastic optimization [C]∥Proceedings of the 3rd International Conference on Learning Representations(ICLR). London, UK: ICLR, 2015: 1-15.

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

蜉蝣优化双通道网络在齿轮箱故障诊断中的应用

一种适用于混合三端直流输电线路的故障定位方法

采用格拉姆角场-卷积神经网络-时序卷积网络混合模型的锂离子电池健康状态估计

一种基于FPGA的卷积神经网络协处理器设计

一种高分辨率遥感图像视感知目标检测算法