基于域自适应的塔式起重机结构损伤诊断方法研究  

Research on domain adaptive structural damage diagnosis method of tower crane

作　　者：宋世军[1] 朱昆贤 安增辉 宋连玉 杨蕊 Song Shijun;Zhu Kunxian;An Zenghui;Song Lianyu;Yang Rui

机构地区：[1]山东建筑大学机电工程学院,济南250101 [2]山东富友科技有限公司,济南250101

出　　处：《起重运输机械》2023年第20期46-52,共7页Hoisting and Conveying Machinery

基　　金：国家级青年科学基金项目资助(52005300)。

摘　　要：塔式起重机结构损伤诊断存在诊断过程复杂且低效、收集数据量大且有标签数据少、不同塔式起重机间收集到的数据因有差异而造成的网络模型诊断精确度不稳定等问题。为解决这些问题引入迁移学习方法,文中采用迁移学习中的域自适应方法,利用过往的标签数据与现有工况无标签数据建立工况间的联系,从而学习到不随工况改变的结构损伤特征,扩大了模型的应用领域。并且,以深层神经网络为基本框架,运用深层多核最大均值差异对不同塔机收集到的数据特征的分布进行评价,实现了少量多种塔式起重机数据的智能结构损伤诊断。There are some problems in the structural damage diagnosis of tower cranes,such as the complex and lowefficient diagnosis process,the large amount of data collected and the small amount of labeled data,and the unstable diagnosis accuracy of network model due to the difference of data collected from different tower cranes.In order to solve these problems,the transfer learning method is introduced.By using the domain adaptive method in transfer learning,the relationship between the past labeled data and the unlabeled data of the existing working conditions was established,so that the structural damage characteristics that do not change with the working conditions were obtained,and the application field of the model was expanded.In addition,based on the deep neural network,the distribution of data characteristics collected by different tower cranes was evaluated by using the deep multi-core maximum mean difference,and the intelligent structural damage diagnosis of a small number of tower crane data is realized.

关 键 词：塔式起重机 域自适应 迁移学习 结构损伤 最大均值差异 

分 类 号：TH213.3[机械工程—机械制造及自动化]