基于异构计算的电弧增材制造熔池尺寸在线检测算法  

An online detection algorithm of wire arc additive manufacturing molten pool dimension based on heterogeneous computing

作　　者：周天生 董博伦 林三宝[1] 蔡笑宇 范成磊[1] Zhou Tiansheng;Dong Bolun;Lin Sanbao;Cai Xiaoyu;Fan Chenglei(Harbin Institute of Technology,Harbin 150001,Heilongjiang,China)

机构地区：[1]哈尔滨工业大学,黑龙江哈尔滨150001

出　　处：《机械制造文摘（焊接分册）》2022年第4期6-9,共4页Welding Digest of Machinery Manufacturing

摘　　要：电弧增材制造过程中的熔池形态影响沉积态金属的成形,为了得到良好的增材产品,首先需要实现对增材过程的在线监测,以获取熔池形态尺寸等数据,为进一步反馈控制提供前提条件。在线监测的重点之一在于计算机的图像处理速度,而图形处理器GPU作为强大的图形处理设备,比CPU更适合做大量图像相关的计算。为此,文中采用CPU/GPU异构协同并行计算方法,试图通过调用GPU设备加速图像处理过程,以实时获取电弧增材制造过程中的熔池形态尺寸信息。试验结果表明,无论GPU调用与否,同一处理程序对同一视频处理所得到的熔池宽度结果基本相同,而当调用GPU后,每帧图像的处理时间大幅降低,处理效率得到很大程度的提高。因此,异构计算为电弧增材制造熔池尺寸在线监测提供了可能。Morphology of molten pool affects formation of deposited metal during wire arc additive manufacturing.In order to obtain good additive products,it is necessary to realize online detection of additive process first,so as to obtain shape and size of molten pool and other data,and provide prerequisite for further feedback control.One of the key points of online detection is image processing speed of computer.As a powerful graphics processing equipment,graphics processing unit(GPU)is more suitable for doing a large number of image-related calculations than CPU.Therefore,CPU/GPU heterogeneous collaborative parallel computing method was adopted in this paper to accelerate image processing process by calling GPU devices,so as to obtain information of shape and size of molten pool in the process of arc additive manufacturing in real time.The results showed that whether GPU was called or not,results of weld pool width obtained by the same video processing program were basically same.When GPU was called,handling time for a single frame was reduced significantly,and efficiency of algorism was highly improved.Therefore,heterogeneous computing provided a possibility for online detection of size of molten pool by arc additive manufacturing.

关 键 词：异构计算 在线检测 增材制造 电弧 

分 类 号：TG444[金属学及工艺—焊接]