Page 69 - 无损检测2024年第六期
P. 69

刘  建,等:

              基于深度学习的列车车轴缺陷超声检测

              4  结语                                                  and transfer learning[J].Sustainable Cities and Society,
                                                                     2021,70:102898.
                  基于深度卷积神经网络并通过多种数据样本                             [9]  LIN T Y,GOYAL P,GIRSHICK R,et al.Focal loss
              增强方法,引入EMblock机制,提出一种基于EM-                             for  dense  object  detection[J].IEEE  Transactions  on
              YOLO v5s的超声检测数据识别方法, 有效提高了列                            Pattern Analysis and Machine Intelligence,2020,42(2):
              车车轴缺陷的超声检测性能,最后验证了不同扩充                                 318-327.
              数据对检测能力的影响。结果表明,所提出网络架                              [10]  LIU Z,GU X Y,WU W X,et al.GPR-based detection
              构能适应相控阵探头的图像和扩充的不同数据的缺                                 of  internal  cracks  in  asphalt  pavement:a  combination
                                                                     method  of  Deepaugment  data  and  object  detection[J].
              陷检测任务,与传统的YOLO v5s以及同类方法相
                                                                     Measurement,2022,197:111281.
              比,该方法在性能上有明显提升,具有应用于实际车
                                                                  [11]  YU  H  M,LI  Q  Y,TAN  Y  Q,et  al.A  coarse-to-
              轴缺陷检测的水平和潜力。
                                                                     fine  model  for  rail  surface  defect  detection[J].IEEE
                                                                     Transactions  on  Instrumentation  and  Measurement,
              参考文献:
                                                                     2019,68(3):656-666.
                [1]  WANG X Y,LOU Z F,WANG X D,et al.Prediction     [12]  LEE H,EUM S,KWON H.ME R-CNN:multi-expert
                   of  stress  distribution  in  press-fit  process  of  interference   R-CNN  for  object  detection[J].IEEE  Transactions  on
                   fit  with  a  new  theoretical  model[J].Proceedings  of  the   Image Processing,2020,29:1030-1044.
                   Institution  of  Mechanical  Engineers,Part  C:Journal     [13]  QIAN  Y,LI  X  L,ZHANG  Q,et  al.SPP-CPI:
                   of  Mechanical  Engineering  Science,2019,233(8):   predicting  compound-protein  interactions  based  on
                   2834-2846.                                        neural  networks[J].IEEE/ACM  Transactions  on
                [2]  SHU Y L,YANG G X,LIU Z M.Simulation research    Computational Biology and Bioinformatics,2022,19(1):
                   on  fretting  wear  of  train  axles  with  interference  fit   40-47.
                   based  on  press-fitted  specimen[J].Wear,2023,523:     [14]  RANI S,GHAI D,KUMAR S.Object detection and
                   204777.                                           recognition  using  contour  based  edge  detection  and  fast
                [3]  彭朝勇,高晓蓉,王艾.车轴压装部相控阵超声波探                         R-CNN[J].Multimedia  Tools  and  Applications,2022,
                   伤的各向异性扩散去噪改进算法[J].中国铁道科学,                         81(29):42183-42207.
                   2017,38(3):77-82.                              [15]  KZLOLUK  S,SERT  E.Hurricane-Faster  R-CNN-
                [4]  周素霞,卢俊霖,吴毅,等.基于直流电位降的高铁                         JS:hurricane detection with faster R-CNN using artificial
                   车轴裂纹检测研究[J].机械工程学报,2022,58(14):                   Jellyfish Search (JS) optimizer[J].Multimedia Tools and
                   288-295.                                          Applications,2022,81(26):37981-37999.
                [5]  曹贞全.动车组空心车轴非接触式超声波检测设备的                      [16]  ZHANG  R  Y,SONG  Y.Non-intrusive  load
                   研究[J].铁道机车与动车,2017(12):46-47,8.                   identification  method  based  on  color  encoding  and
                [6]  MARSHALL M B,LEWIS R,DWYER-JOYCE R S,           improve  R-FCN[J].Sustainable  Energy  Technologies
                   et  al.Ultrasonic  measurement  of  railway  wheel  hub‒  and Assessments,2022,53:102714.
                   axle  press-fit  contact  pressures[J].Proceedings  of  the     [17]  HWANG  Y  J,LEE  J  G,MOON  U  C,et  al.SSD-
                   Institution  of  Mechanical  Engineers,Part  F:Journal  of   TSEFFM:new  SSD  using  trident  feature  and  squeeze
                   Rail and Rapid Transit,2011,225(3):287-298.       and extraction feature fusion[J].Sensors,2020,20(13):
                [7]  HE  S  Y,HU  D  Y,YU  G,et  al.Trackside  acoustic   3630.
                   detection  of  axle  bearing  fault  using  wavelet  domain     [18]  XU  Z  J,SU  J  J,HUANG  K.A-RetinaNet:a
                   moving  beamforming  method[J].Applied  Acoustics,  novel  RetinaNet  with  an  asymmetric  attention  fusion
                   2022,195:108851.                                  mechanism for dim and small drone detection in infrared
                [8]  ZHENG  Z,QI  H  Y,ZHUANG  L,et  al.Automated    images[J].Mathematical  Biosciences  and  Engineering:
                   rail surface crack analytics using deep data-driven models   MBE,2023,20(4):6630-6651.










                                                                                                          35
                                                                                         2024 年 第 46 卷 第 6 期
                                                                                                  无损检测
   64   65   66   67   68   69   70   71   72   73   74