【作者】 高俊平；

【导师】 王克鸿；

【作者基本信息】 南京理工大学 ， 材料加工工程， 2010， 硕士

【摘要】 本文针对焊接过程中存在较大干扰、对任意曲率的焊缝跟踪时存在较大误差以及系统无法满足高速焊接要求的难题,开展了图像处理和智能控制的深入研究,突破图像处理在焊缝跟踪系统中精确性与实时性的矛盾,使焊缝跟踪系统能够对一定曲率范围内的曲线实时跟踪,在此基础上实现高速化焊接。采用激光视觉传感方法,对视觉传感装置进行改进,选用980nm波长的半导体激光二极管、配套的980nm窄带滤光片以及1%的减光片等组成的激光视觉传感装置,获得了清晰的铝合金接缝坡口图像,很好的解决了铝合金对激光的二次反射与焊接过程中的飞溅与弧光强干扰等问题,提高了系统的适用性,同时为后续坡口图像的处理打下了良好的基础。对铝、钢在多种熔化极气保焊接时采集的图像进行研究并分类整理,按照噪声信号的分布特征把图像分为两类,在此基础上对两类坡口图像进行分析并选择合适的图像处理流程。制定焊缝跟踪初始化界面,并对界面传递的关键信息进行整合,从而制定出最优的系统处理与控制流程。为了达到兼具实时性与精确性的图像处理效果,本文对处理算法如中值滤波、均值滤波、二维otsu阈值分割等算法进行优化设计,并采用多线程的处理思想,把原来的一个线程分为一个主线程与两个子线程,使用关键代码段把图像采集处理与偏差纠正并发运行,最终整个流程的耗时控制在40ms以内,完全满足实时控制的要求。文中建立了步进电机的传递函数模型,利用Simulink分别对Fuzzy-PID分段控制与P-模糊监督PID (P-FPID)分段控制进行仿真分析,仿真结果表明后者的控制精度与响应速度均优于前者,因此把Fuzzy控制做为PID参数的控制器,较Fuzzy控制直接控制系统输出效果好。最后为了验证本文所提出的图像处理系统与控制系统的性能，进行铝合金斜线接缝与曲线接缝的实时跟踪试验。研究表明,对于焊接速度在0.9m／mim～1.2m／min范围内,斜率在10°～20°之间的斜线接缝,焊接实时跟踪的最大误差在1mm范围内。在0.6m／min焊接速度下对曲线焊缝跟踪时,焊接实时跟踪的最大误差在0.7mm范围内。更多还原

【Abstract】 With regard to problems that the heavy interferences in welding process, and deviation in tracking of curvatures,system can’ meet the requirements of high-speed welding, image processing and intelligent control were studied. The conflicts between the accuracy and real-time of image processing was worked out,and the system could achieve real-time within all curves of a certain curvature, in which foundation the high speed welding will be realized.To improve the visual sensing devices which based on structured light,980nm wavelength semiconductor laser diode,980nm narrow band passing-filter and 1% cutting-filter were choosed to make up the laser vision sensor, then clear images of the joint were obtained and the reflection of light on the Aluminum alloys, spatters and arc interferences in the process of welding were eliminated. The new system had good flexibility and opened the way to the subsequent image processing.The Images of common material in different welding process were studied and classified, and the images with two kind of noise which matched different image process were analyzed. Then the initialization interface of seam tracking was figured out, from which the system integrated the critical information and derived the optimal system processing and control flow,In order to make the image processing both real-time and accuracy, several important processing algorithms were optimized, for example, median filtering, mean filter, two-dimensional otsu threshold segmentation. Then based on multi-threaded processing, the traditional process was seperated into one main thread and two subordinate thread,which improved the using of CPU and at the same time made the entire process faster,At last the entire process costed less than 40ms which met the requirements of real-time control.The transfer function of the system was established, and simulation based on Fuzzy-PID、P and fuzzy control to supervise parameters of PID (P-FPID)sub-section control were conducted with Simulink software.The results showed that the accuracy and response speed of the latter control were better, so fuzzy control to supervise parameters of PID control was more effective than the fuzzy control alone.Finally, to verify the performance of image processing and control system had been improved, real-time tracking test of aluminum alloys of diagonal and curved joints was carried out. As a result, the maximum deviation existed in 1mm when the welding speed was in 0.9m/min～1.2m/min and the slope of joints were in the 10°～20°,and the deviation existed in 0.7 mm when the welding speed was 0.6m/min in tracking the curve seam.更多还原