【作者】 张永强；

【导师】 陈武柱；

【作者基本信息】 清华大学 ， 材料科学与工程， 2006， 博士

【摘要】 激光切割是目前工业上最先进和最重要的材料裁切方法之一,对激光切割质量的实时检测与控制是该领域的前沿课题。本文通过同轴视觉传感系统同时获得切割前沿和切割火花簇射视觉图像,成功地提取了产生切割缺陷(过烧、挂渣)和切割面粗糙度变化时图像特征信号的变化规律,首次实现了以无切割缺陷并且在一定工艺条件下获得最高切割质量(最低切割面粗糙度)为目标的切割速度自寻优控制。在工艺实验的基础上,建立了切割面分层理论模型,实验研究和模型分析表明:切割面近下缘处是切割质量的最薄弱环节,此处粗糙度最高,切割前沿在下缘处的温度是决定下缘质量的主要因素。提出以近下缘粗糙度作为切割质量评价和检测的主要指标较合理。通过切割前沿视觉图像,无法获得切割面粗糙度的信息,但可以作为识别切割缺陷的判据。实验证明,过烧缺陷的特征是图像中切割前沿上缘和下缘波动发生向上跳变,而挂渣缺陷的图像特征是仅有下缘波动发生向上跳变。从侧面研究了切割过程中的火花簇射行为,发现了火花簇射视觉图像特征与切割速度以及切割面近下缘粗糙度之间有良好的对应关系:在一定工艺条件下,随着切割速度的变化,火花簇射喷射长度以具有极大值的倒U形曲线形式变化,最大的火花簇射喷射长度(即火花簇射视觉图像不同亮度带像素数的极大值)对应于最低的切割面近下缘粗糙度,此时的切割速度为该工艺条件下的最佳切割速度。这一发现为切割面质量的在线检测与寻优控制奠定了基础。为适应切割方向和切割位置的变化,论证了切割过程火花簇射主要视觉特征同轴检测的可行性,证明火花簇射的同轴视觉图像中最大火花簇射长度(火花簇射图像不同亮度带最大像素数)与最低的切割面粗糙度的对应关系仍然存在。在此基础上,建立了具有现场实用性的激光切割同轴视觉检测和控制系统及相应的图像处理方法。同时利用切割前沿和火花簇射同轴视觉图像上述特征信号的变化规律,进行了激光切割质量的实时检测与自寻优控制。结果表明,系统可以迅速寻找到最佳切割速度,达到无切割缺陷而且该工艺条件下切割面近下缘粗糙度最小。更多还原

【Abstract】 Laser cutting is one of the most advanced and important technology for metallic materials processing. Process monitoring and control is significant for high-quality laser cutting. By coaxially detecting the images of the cutting front and the sparks jet, the relationship between the detected signals and cutting quality, including cutting defects (burning or dross) and surface roughness, has been obtained. Self-optimizing control aiming to the best roughness on some certain cutting conditions has been realized.A laminated model of cutting face for evaluating cutting quality has been built up on the basis of cutting experiments. The experimental results and the theoretical analysis of the model show that, the lower part of the cutting face, adjacent to the bottom edge, has the worst cutting quality due to its worst roughness. The quality of the bottom face is mainly depended on the temperature of the lowest part of cutting front. Therefore, the method with this roughness as the main criterion for quality evaluation and detection is reasonable.The vision image of cutting front could not indicate the roughness variation of the bottom edge, but could be used as the criterion identifying cutting defects. The characteristic signal responding to burning occurrence is the jumping standard deviation of both DCF (Distance from Beam Center to Front Edge) and DCB (Distance from Beam Center to Back Edge) of the detected image, while dross occurrence only inducing the jumping standard deviation of DCB.The behavior of sparks has also been investigated by side imaging the sparks jet. It is found that the sparks shapes are related to both the cutting speed and the roughness of the bottom face. On the given condition, the sparks jet length, which corresponds to the pixel quantity of different gray scales, shows an inverted U-curve relationship with the cutting speed. The maximum sparks length occurs at the status of the best roughness of the bottom face, which relates to the optimal cutting speed.Coaxial detecting the sparks jet has also been tried for application to variouscutting directions and attitudes. It is proved that the maximal sparks length of the coaxial image, which can be determined by the highest pixels of different brightness ranges of the sparks image, still corresponds to the best roughness of cutting face. A system for coaxial visual detection and control in laser cutting, as well as the corresponding image processing method has been built up.Experiments for coaxial detecting and optimum controlling of laser cutting quality were performed by simultaneously utilizing the images of both cutting front and sparks jet. The results show that the optimal cutting speed for free defect and the best roughness of the bottom edge could be found automatically and quickly.更多还原