论文部分内容阅读
针对国内外转炉炼钢终点控制的现状,建立了一种用于终点预测的神经网络模型。以炉口辐射信息获取系统为实验平台,运用光纤谱分复用和颜色空间模型转换技术,分析发现了光谱与图像信息特征量在吹炼过程中呈现出中前期类似、末期相反的规律。从得到的特征规律曲线中选用一些关键特征量,在改进的修正系数算法基础上,进行了模型的训练和预测分析。实验结果表明:响应时间在2s以内,满足快速判定的时间要求;改进算法的模型预测精度高于常规算法,该系统可以正常工作在转炉炼钢的恶劣环境下,达到了预期效果。
Aiming at the status quo of BOF endpoint control at home and abroad, a neural network model for endpoint prediction is established. Taking the radiation information acquisition system of the mouth as an experimental platform, the spectral spectral multiplexing and color space model conversion techniques were used to analyze the characteristics of spectral and image information in the blowing process. Based on the improved correction coefficient algorithm, some key features are selected from the characteristic curves obtained. The model training and prediction analysis are carried out. The experimental results show that the response time is within 2s, which meets the time requirement of fast decision. The improved prediction accuracy of the improved algorithm is higher than that of the conventional algorithm. The system can work normally in the harsh environment of converter steelmaking and achieve the expected results.