矿用刮板的磨损曲面重建方法研究  

Research on Worn Surface Reconstruction Method for Mining Scrapers

作　　者：王艺卓 李建儒 龚堰珏 赵罘[1] Wang Yizhuo;Li Jianru;Gong Yanjue

机构地区：[1]北京工商大学,北京市100048

出　　处：《工程机械》2025年第4期89-96,I0004,I0005,共10页Construction Machinery and Equipment

基　　金：国家自然科学基金项目(51975006)。

摘　　要：为实现矿用刮板的自动化高质量修复,对失效刮板磨损部位的形态进行机器人激光熔覆路径的设计。使用线结构光扫描系统获得待熔覆部位的点云,针对点云精度不足的问题,研究提出一种结合点云插补算法和卷积神经网络(Convolutional Neural Networks,CNN)的方法进行磨损曲面的点云重建。点云插补算法通过Nurbs曲线拟合得到待加工曲线并利用构建基准线确定误差阈值的方法提高点云数据的纵向精度,利用基于网格搜索超参数优化的CNN-LS TM对点云数据进行横向预测,所提CNN-LS TM模型在评估过程中表现出高拟合优度和良好的泛化能力,能够有效捕捉数据特征,并在未知数据上做出可靠预测,刮板的磨损曲面可以得到真实还原。为复杂曲面工件的自动化修复提供了新的技术途径。To achieve automated high-quality repair of mining scrapers,a robotic laser cladding path is designed for the morphology of the worn parts of the failed scraper.The point cloud of the parts to be cladded is obtained by using a linear structured light scan system,and in view of the problem of insufficient accuracy of the point cloud,a method combining the point cloud interpolation algorithm with Convolutional Neural Networks(CNN)is proposed for point cloud reconstruction of the worn surfaces.The point cloud interpolation algorithm improves the longitudinal accuracy of the point cloud data by fitting the Nurbs curve to obtain the curve to be processed and using the method of constructing the reference line to determine the error threshold.The CNN-LSTM based on grid search hyperparameter optimization is used for lateral prediction of the point cloud data.The proposed CNN-LSTM model shows a high goodness-of-fit and a good generalization ability in the evaluation process,which can effectively capture data features and make reliable predictions on unknow data,and the worn surface of the scraper can be truly restored.It provides a new technological approach for automated repair of workpieces with complex curved surfaces.

关 键 词：矿用刮板 NURBS曲线 插补算法 超参数优化 卷积神经网络 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]