结合图像编码器与2D-CNN的钢桁梁桥损伤识别

打开文本图片集

中图分类号：U446.2；TU317 文献标志码：A 文章编号：1007-2683（2025）05-0027-14

Combining Image Encoder and 2D-CNN for Steel Truss Bridge Damage Recognition

CHEN Mengyuan， RAN Zhihong， PAN Jie （College of Architecture and Planning，Yunnan University，Kunming 65O5OO，China）

Abstract：Aiming attheproblemthattraditionalmachinelearingisdificulttodeal withthecomplicateddataunderlong-term monitoring，acombinationofimageencoderand2D-C（two-dimensionalconvolutionalneuralnetwork）inteligntdamageition methodisproosdA2.5stltrusststbdgenditsoespodingfiteementoelareedfotesudyndiidsf imageencodersareusedtoconvertthtime-rangedisplacementdataintoequal-sizeimage matrices，andthematricesareusedasthe two-dimensional rainingsamplesofCNN（convoutionalneuralnetwork）tobtainthedplearning modelfordamagerecognition，so astovalidateteroustessoftediferentimageencodingmetodsindamageinformationextractioandpattenrecognitioResults show that the method improves the accuracy by 23.4% compared with the traditional machine learning algorithm and 1D-CNN（onedimensionalconvolutioalneuralnetwork），andafterfurtherexperimentalstudies，thestabilityandapplicabilityofGAF（gramian angular field）are found to be optimal，and its damage recognition accuracy reaches 100% ：

Keywords：damage identification； image encoder; 2D-CNN；bridge health monitoring；machine learning

0 引言

短、节省材料、拓扑特征明显等优点，被广泛用于桥梁结构。（剩余17130字）