【作者】 李建；

【导师】 赵军； 马丽霞；

【作者基本信息】 燕山大学 ， 材料加工工程， 2009， 博士

【摘要】 自河北青县巨龙钢管有限公司引进了我国第一条大口径直缝埋弧焊管JCO生产线后,我国开始对管坯JCO成形工艺及相关技术进行了消化和吸收,由于关键技术和设备从国外引进,对管坯成形质量控制及工艺参数调整只停留在经验层面上,存在着工艺参数调整误差大、人员工作强度大及产品质量波动大等现实情况。为此,本文提出了大口径直缝埋弧焊管JCO成形智能化控制技术,以实现管坯成形质量的实时监测和工艺参数的实时预测。本文在板材弯曲智能化控制技术的研究成果基础上,分析了管坯JCO成形智能化控制需要解决的关键技术,并对管坯JCO成形过程理论解析、成形过程中的信息监测、材料性能参数识别及工艺参数预测等方面的相关问题展开了系统研究。从板料成形状态和板料内质点变形状态出发,基于塑性弯曲工程理论,建立了管坯JCO弯曲成形过程的力学模型,推导出了板料任一成形状态板料内任一质点卸载前后的转角、弯曲半径数学表达式,分别给出了描述弯曲力和弯曲行程、弯曲角和弯曲行程之间关系的更为精确的解析表达式。根据卸载理论,建立了精确的回弹计算模型,并利用有限元和实验研究等方法对理论计算结果进行了数值验证,为管坯JCO成形智能化控制技术中识别和预测模型的建立奠定了理论基础。利用机器视觉和传感器技术,开发了管坯JCO成形过程实时监测系统,实现了弯曲力、弯曲行程和弯曲角等物理参数的非接触、高精度的实时测量。详细研究了管坯端面图像处理和直线检测算法,在试验和结果比较基础上,提出了适合管坯端面的图像处理流程。系统地研究了摄像机标定方法,为克服传统标定方法的标定过程繁琐、需高精度标定参照物的不足,本文提出了适合现场使用的正三角形标定方法,并用实验验证了该方法的正确性和有效性。通过对管坯JCO成形及弹复过程中的特征变量进行分析和研究,完成了神经网络识别模型的输入、输出层设计,利用MATLAB编程语言,开发了管坯JCO成形过程中的参数识别和预测程序。选择前馈BP网络作为识别和预测模型的神经网络结构、LM算法作为网络优化算法,提高了网络收敛和泛化精度。在上述研究基础上,利用Visual C++编程语言开发了管坯JCO成形智能化控制系统软件,主要内容包括:管坯弯曲成形过程理论解析计算程序开发,神经网络训练样本自动创建程序开发,弯曲力、弯曲行程数据采集程序开发,管坯端面图像处理及直线检测程序开发,摄像机标定程序开发以及Visual C++与MATLAB接口程序开发。最后利用PCI总线数据采集卡、面阵CCD及相关硬件,完成了管坯JCO成形智能化控制硬件系统开发。利用管坯JCO成形智能化控制系统进行了物理模拟实验,实验结果表明系统运行稳定,行程工艺参数预测精度可靠,制品椭圆度具有较高的一致性。作为研究工作的最终成果,成功地将自主开发的智能化控制系统移植到了大口径直缝埋弧焊管JCO成形生产线上。更多还原

【Abstract】 Since the first longitudinal-seam submerged arc welded (LSAW) pipes production line with JCO forming process was established in Julong Steel Pipe Co., Ltd, the unfinished pipe forming process and the correlated technology have been investigated by the researchers. Due to the key technology and the forming press from abroad, the systematical basic research on the forming process has not been conducted yet. And the pipe quality and processing parameters adjustment are conducted according to accumulated experience with the characteristics of large error, high personnel working strength and unstable production quality. As one of the effective method to improve the pipe quality, the intelligent control technology is introduced to real-time monitor the pipe quality and to adjust the processing parameters automatically.Based on the research achievements of the intelligent control technology for the sheet metal bending, the key technologies of intelligent bending for unfinished pipe forming with JCO process are analyzed, and the relevant issues to the theoretical analysis, the process monitoring, the identified model of material properties and the predicted model of optimal processing parameters have been studied in this paper.In this investigation, based on the elementary theory of plastic bending, the mechanical model which can describe the bending process and springback behavior adequately has been established. The equations for the rotation angle and radius of each point in neutral surface for any bending process are derived. And the calculated model of springback after removing the punch load is derived according to the uploading theory. The finite element method and experiment were introduced to verify the validation of the mechanical model established before and analyze the dominant factors influenced on the bending and springback process. And the investigation is carried out to provide the theoretic basis for the intelligent control for unfinished pipe forming with JCO process.Based on machine vision and sensor technology, the system for real-time monitoring the sheet metal bending process is developed, which can measure the bending force, punch displacement and the unbend angle with high precision. The processing algorithms of image pre-processing and line detection, which are simple, efficient and suitable for the pipe ending surface image, are studied in detail. According to the comparison of the image processing results, a series of algorithms suitable for the images are identified. Focused on the live condition, a new camera calibration method has been put forward to transform the angle in image into the actual angle. The experimental data indicates that the calibration method proposed has advantages of simple operation, fast completion and the high precision, which is very suitable for live system calibration.The identification model of material properties and the prediction model of optimal processing parameters have been constructed employing the neural network technology. In the identification model, the bending force, the punch displacement and the unbend angle are denoted as the input variables, and the material properties are the output variables. The Levenberg-Marquarat is chosen as the optimal algorithm of neural network whose topology structure is feedforward network, and Matlab language is to program.Employing the research results above, an intelligent control code for unfinished pipe forming with JCO process is developed with Visual C++ language, including the development of calculated code theoretic analysis for sheet metal air bending process, the development of building code of the neural network training sample data, the development of gathering code of bending force and the bending displacement, the development of calibration code, the development of unbend angle identifying code, and the development of the interface code between the signal control and the identification model. At last, employing the portable DAQ card and camera CCD and other hardware, the intelligent control system for the unfinished pipe forming with JCO process is established.The physical simulated experiments have been conducted with the intelligent control system, which showed that the system operated stably, the punch displacement predicted is reliable, and can obtain high quality pipes. As the final achievement of research work, the intelligent control system has been transplanted to the LSAW pipe production line with JCO process successfully, improving the production efficiency and the product quality and realizing the intelligent control for the unfinished pipe bending process.更多还原