节点文献

基于计算机视觉的熔池检测与分析

Detection and Analysis of Molten Pool Based on Computer Vision

【作者】 杨帆;

【导师】 宋立军;

【作者基本信息】 湖南大学 , 机械工程, 2020, 硕士

【摘要】 激光增材制造技术是使用高能激光把基体表面和粉末材料一起熔化,然后凝固,使基体表面与粉末聚积在一起的特殊制造技术。与传统制造技术相比,增材制造技术实现了零件“自由制造”,具有性能好、不需要模具、加工周期短、不受零件材料和结构限制等一些优点,已经广泛应用于航天航空、汽车、军工、医疗等许多领域。激光增材制造技术主要用于涂层加工、堆积成形、零件制造和修复等方面,近年来相关技术的发展使其应用到了更加广泛的地方,但是仍然存在不足之处,特别是激光增材制造智能化、自动化研究这一方面。本文以提高激光增材制造智能化、自动化为主要目标,使用基于计算机视觉的方法对激光增材制造过程中的熔池进行检测和边缘提取,并研究在工艺参数变化时,熔池几何特征的变化规律,为后续建立激光增材制造智能控制系统奠定良好的基础。为此,本文主要完成以下内容:(1)说明本文的研究背景和意义。对增材制造的现状和现有的熔池采集技术,检测技术及其检测结果进行分析,了解到现有研究中的不足和可提升的地方。确定下来研究内容是基于计算机视觉的熔池检测与边缘提取并分析熔池几何特征随工艺参数变化的规律。(2)分析熔池图像的特点,通过调研和实验,对熔池采集所需的相机、辅助光源、镜头、滤波片和衰减片进行挑选,搭建合适的熔池采集装置,采集到清晰度高、质量好的熔池图像。(3)提出基于改进目标检测和语义分割的熔池边缘检测方法,通过使用改进卷积结构、改进损失函数和引入多尺度特征融合网络结构等方法,大大提高熔池检测和边缘提取的准确率和速度。(4)通过使用改进后的算法获取到了熔池的面积、长度和高度,并研究了激光功率、送粉率和扫描速度等工艺参数对熔池几何特征的影响规律。

【Abstract】 Laser additive manufacturing technology is a special manufacturing technology that uses a high-energy laser to melt the surface of the substrate and the powder material together and then solidify,so that the surface of the substrate and the powder accumulate together.Compared with the traditional manufacturing technology,it has realized the "free manufacturing" of parts.It has some advantages such as good performance,no mold,short processing cycle,and is not limited by the material and structure of the parts.It has been widely used in aerospace,automotive,military industry,medical and many other fields.The main applications of laser additive manufacturing technology are coating processing,stacking forming,parts repair,etc.In recent years,the development of related technologies has made it applied to more extensive areas,but there are still shortcomings,especially in the aspect of laser additive manufacturing intellectualization and automation.In order to improve the intellectualization and automation of laser additive manufacturing,this paper uses computer vision-based methods to detect and extract the molten pool during the laser additive manufacturing process,and studies the change rule of the geometric characteristics of the molten pool when the technological parameters of laser additive manufacturing change,which lays a good foundation for the subsequent establishment of the intelligent control system of laser additive manufacturing.Therefore,this paper mainly completes the following contents:(1)Explain the research background and significance of this paper.This paper analyzes the current situation of laser additive manufacturing and the existing molten pool collection technology,detection technology and detection results,and finds out the shortcomings and potential improvements in the existing research.It is determined that the research content is molten pool detection and edge extraction based on the computer vision,and analyzes the rule of molten pool geometry characteristics changing with laser additive manufacturing process parameters.(2)Analyze the characteristics of the molten pool image,through research and experiment,select the camera,auxiliary light source,lens,filter and attenuation filters needed for molten pool collection,build appropriate molten pool collection device,and collect the molten pool image with high definition and good quality.(3)Propose the edge detection method based on improved object detection and semantic segmentation.By the use of improving convolution structure,improving loss function and adding multi-scale feature fusion network structure,the accuracy and speed of molten pool detection and edge extraction are greatly improved.(4)Obtain the area,length and height of the molten pool by using the improved algorithm,and study the influence of technological parameters such as laser power,powder feed rate and scan speed on the geometric characteristics of the molten pool.

  • 【网络出版投稿人】 湖南大学
  • 【网络出版年期】2022年 03期
  • 【分类号】TP391.41;TP391.73
  • 【下载频次】150
节点文献中: 

本文链接的文献网络图示:

本文的引文网络