基于大数据挖掘的低碳钢连铸工艺参数优化

摘要：针对热轧卷表面缺陷，基于大数据挖掘技术中的神经网络预测模型，提出了一种优化连铸工艺参数的新方法(prediction model method，简称PMM)。PMM方法可以得到各连铸参数对表面缺陷发生可能性的多样本连续变化图，并以此得到对应影响规律、关键工艺参数及临界值。结果表明，吹氩参数中，保护氩气流量对低碳钢热轧卷表面缺陷影响最为明显且呈负相关关系，塞棒与水口位置的最佳吹氩流量分别为3.0和1.8L/min。结晶器热流参数中，内弧侧水流量影响最明显，各面水温差最佳范围为7~9℃，最佳进水温度在35℃附近。同时，表面缺陷发生可能性随拉速提高、板坯宽度、浇铸长度增加而增加明显，但随中间包钢水质量增加而逐渐降低。此外，对比发现浇铸速度、板坯宽度、保护氩气流量与结晶器冷却水流量等参数是影响热轧卷表面缺陷形成的关键连铸工艺参数，且缺陷发生可能性对结晶器冷却水总流量的波动最为灵敏，其临界下限值为8700L/min。

Parameter optimization for continuous casting of low carbon steel based on big data mining

Abstract：As to the surface defects in hot rolled coils, based on the neural network prediction model in big data mining, a new method (prediction model method, PMM) for optimizing continuous casting process parameters was proposed. The multi sample continuous variation chart of the possibility of surface defects in each continuous casting parameter can be obtained by the PMM method. And based on it, corresponding influence law, key process parameters and critical values can also be obtained. The results show that among the parameters of argon blowing, the protective argon flow has the most obvious effect on the surface defects of low carbon steel hot rolled coils and has a negative correlation. The optimal argon blowing flow for the stopper rod and the nozzle position are 3.0L/min and 1.8L/min, respectively. Among the heat flow parameters of the mold, the influence of the water flow on the inner arc side is the most obvious, and the best range of water temperature difference on each surface is 7-9℃, the best water inlet temperature is about 35℃. At the same time, the possibility of surface defects increases significantly with the increase of the casting speed, the width of the slab, and the increase of the casting length, but it gradually decreases with the increase of the weight of the molten steel in the tundish. The casting speed, slab width, protective argon flow and mold cooling water flow are the key process parameters that affect the formation of surface defects in hot rolled coils. And the possibility of defect occurrence is the most sensitive to the fluctuation of the total cooling water flow of the crystallizer, and its critical lower limit is 8700L/min.