【作者】 温宏愿；

【导师】 陈延如；

【作者基本信息】 南京理工大学 ， 光学工程， 2009， 博士

【摘要】 转炉炼钢在线终点的准确控制是全世界冶炼行业一直亟待解决的难题。转炉炼钢是世界上最主要和效率最高的炼钢方法。转炉终点的准确控制在提高炼钢质量、节能减排、降低生产成本和提高吹炼自动化水平等方面均具有重要意义。然而在转炉吹炼过程中,由于加入原材料的不稳定性、吹炼过程中复杂的化学反应和吹炼钢种范围的严格性等各方面的原因,对终点的碳、温度进行准确控制至今仍然难以实现。为了准确地判定转炉终点,本文展开了一系列研究,具体内容为：构建了炉口辐射多频道信息获取系统,该系统主要包括光纤光学系统和火焰视频捕获系统两大部分,其中光纤光学系统细分为炉口辐射获取分系统,光纤谱分复用分系统和多光谱复合探测分系统。光纤光学系统在避免高温和污染的环境下,一方面可以把可见光波段炉口辐射的光谱信息较完整的采集下来,利用光纤谱分复用技术,使得到达各通道探测器前的光谱能量均匀一致,并把转炉炉口光辐射信息进行光电转换处理后,通过串口传送给计算机用于后续判定；另一方面,同时采集炉口火焰的视频图像信息,利用颜色模型空间转换技术,实现了图像特征信息的提取。综合考虑分析光谱光强和图像的特征曲线,得到了能够反映转炉吹炼过程的潜在光辐射吹炼规律,该规律曲线可以分为前中后三个阶段,不同阶段的曲线变化情况都较好地符合该阶段的炉内碳氧反应情况。为了准确地判定炼钢终点,本文选用上述规律中的一些特征值作为模型的输入输出变量,利用各自训练样本数据建立了用于判定转炉终点的回归预测模型和神经网络预测模型,并把建立好的两种预测模型对非样本数据空间的其他炉次情况进行了实际预测,其中回归预测模型在4秒内的预测精度为90.4%,神经网络预测模型在5s内的预测精度为76.9%,系统响应时间为1.688s,满足预测精度和快速判定的要求。研究所用的方法和技术为今后转炉终点控制技术的发展提供了一定的参考价值,它将对提升传统产业科技水平,推进工艺装备现代化有较好的效果。更多还原

【Abstract】 The Basic Oxygen Furnace (BOF) steelmaking end-point’s accurate control online is a problem to be solved urgently in the world smelting industry. The BOF steelmaking is the most important and the most efficient steelmaking method in the world. The accurate control of the steelmaking end-point has important significance for the improvement of the steel quality, energy-saving and emission-reduction, the reduction of production cost and the improvement of the blowing automation level. However, it is difficult to control the end-point carbon and temperature accurately because of many reasons in the blowing process, such as the instable raw materials, the complex chemical reactions, and the strict steel grade scope and so on.To judge the steelmaking end-point accurately, a series of research have been done as below:A BOF mouth radiation multi-frequency information acquisition system was designed, which mainly included the optical fiber optical system and the mouth flame image capturing system. The optical fiber spectrum was divided into the mouth radiation acquisition subsystem, fiber spectrum division multiplexing subsystem and the multi spectrum light intensity detection subsystem. On the one hand, the visible band radiation information was collected completely and divided into the same spectrum energy using the fiber spectrum division multiplexing technology when they arrived at various channel detectors. The spectrum radiation information was processed by the electro-optical transformation and then transmits to the computer for subsequent judgment through the serial port. On the other hand, the flame image information was captured, and then the image characteristic information is extracted by the color space conversion method. Comprehensive consideration on the spectrum light and image characteristic curve, the latent optical radiation blowing law was obtained, which could reflect the steelmaking blowing process. The law could divide into three stages, and the curve change of different stage coincided with the carbon-oxygen reaction in the same stage. The end-point regress prediction model and the end-point neural network prediction model were established based on the respective training sample data and the parameters selected from the optical radiation blowing law. The non-type data space was forecasted by the models and the anticipated effect is achieved. The regress model prediction accuracy is 90.4% when the measurement erro is less than 4s. The neural network model prediction accuracy is 76.9% when the measurement erro is less than 5s, and the model response time is 1.688s. The results show that the anticipated prediction accuracy and the requirements of online end-point judgment are achieved.The method and technology in this study have important reference value for the development of the BOF steelmaking end-point control technology. It will have good effect on promoting the traditional industry technology level and the advancement craft equipment modernization.更多还原