输电线路的关键部位包括塔身、导线、绝缘子、避雷线以及引流线,无人机精细化导航的首要任务是构造输电线路的点云地图并从中分割出上述部位。为解决现有算法在输电线路的绝缘子、引流线等精细结构分割时精度低的问题,通过改进PointNet+...输电线路的关键部位包括塔身、导线、绝缘子、避雷线以及引流线,无人机精细化导航的首要任务是构造输电线路的点云地图并从中分割出上述部位。为解决现有算法在输电线路的绝缘子、引流线等精细结构分割时精度低的问题,通过改进PointNet++算法,提出了一种面向输电线路精细结构的点云分割方法。首先,基于无人机机载激光雷达在现场采集的点云数据,构造了输电线路点云分割数据集;其次,通过对比实验,筛选出在本输电线路场景下合理的数据增强方法,并对数据集进行了数据增强;最后,将自注意力机制以及倒置残差结构和PointNet++相结合,设计了输电线路关键部位点云语义分割算法。实验结果表明:该改进PointNet++算法在全场景输电线路现场点云数据作为输入的前提下,首次实现了对引流线、绝缘子等输电线路中精细结构和导线、杆塔塔身以及输电线路无关背景点的同时分割,平均交并比(mean intersection over union,mIoU)达80.79%,所有类别分割的平均F_(1)值(F1 score)达88.99%。展开更多
Taking lower speed gas tungsten arc welding of thin stainless steel plate as study case, two dimensional and symmetric Gaussian distribution mode of heat flux from the arc is applied to calculate the temperature field...Taking lower speed gas tungsten arc welding of thin stainless steel plate as study case, two dimensional and symmetric Gaussian distribution mode of heat flux from the arc is applied to calculate the temperature field and thermal cycles. APDL of three or two dimensional array is effective to express the planar load of heat flux. The dynamic variation of welding temperature field and thermal cycles at different points are numerically simulated. The effect of the distribution parameter σq of Gaussian heat flux on the temperature profiles is investigated. The increasing of σ q causes the dropping of the maximum temperature, and this decreasing is clearer at the range of distance 0 -4 mm away from the weld centerline. Beyond the range with a distance 4 - 10 mm along the transverse direction, minor vibration of temperature occurs, but the temperature difference is limited. As the value of σq rises, the calculated weld pool shape gets contracted. The predicted weld widths at both top and bottom surfaces match well with the experimentally measured ones.展开更多
文摘输电线路的关键部位包括塔身、导线、绝缘子、避雷线以及引流线,无人机精细化导航的首要任务是构造输电线路的点云地图并从中分割出上述部位。为解决现有算法在输电线路的绝缘子、引流线等精细结构分割时精度低的问题,通过改进PointNet++算法,提出了一种面向输电线路精细结构的点云分割方法。首先,基于无人机机载激光雷达在现场采集的点云数据,构造了输电线路点云分割数据集;其次,通过对比实验,筛选出在本输电线路场景下合理的数据增强方法,并对数据集进行了数据增强;最后,将自注意力机制以及倒置残差结构和PointNet++相结合,设计了输电线路关键部位点云语义分割算法。实验结果表明:该改进PointNet++算法在全场景输电线路现场点云数据作为输入的前提下,首次实现了对引流线、绝缘子等输电线路中精细结构和导线、杆塔塔身以及输电线路无关背景点的同时分割,平均交并比(mean intersection over union,mIoU)达80.79%,所有类别分割的平均F_(1)值(F1 score)达88.99%。
基金Acknowledgements The authors are grateful to the financial supports from Shandong Province Natural Science Foundation (ZR2010EM073) and Qingdao Science and Technology Fundamental Research Foundation (09-1-3- 37-jch ).
文摘Taking lower speed gas tungsten arc welding of thin stainless steel plate as study case, two dimensional and symmetric Gaussian distribution mode of heat flux from the arc is applied to calculate the temperature field and thermal cycles. APDL of three or two dimensional array is effective to express the planar load of heat flux. The dynamic variation of welding temperature field and thermal cycles at different points are numerically simulated. The effect of the distribution parameter σq of Gaussian heat flux on the temperature profiles is investigated. The increasing of σ q causes the dropping of the maximum temperature, and this decreasing is clearer at the range of distance 0 -4 mm away from the weld centerline. Beyond the range with a distance 4 - 10 mm along the transverse direction, minor vibration of temperature occurs, but the temperature difference is limited. As the value of σq rises, the calculated weld pool shape gets contracted. The predicted weld widths at both top and bottom surfaces match well with the experimentally measured ones.