基于无约束小波权重TSVR的转炉炼钢终点静态预测模型
|
摘要
转炉炼钢是一个极其复杂的物理化学反应过程,采用智能方法建立转炉炼钢的数学模型是近些年来的一个热点问题。针对熔池碳含量和温度的终点命中率问题,提出了一种新的静态预测模型的建模方法。在传统的孪生支持向量回归机的基础上,将小波权重矩阵引入到目标函数中,然后将目标函数转换成无约束优化问题求解,提高了算法的性能和运算效率;最后基于某炼钢厂260 t转炉的实际生产数据,建立了转炉炼钢终点静态预测模型。试验结果表明,预测模型的终点碳质量分数(误差±0.005%)和温度(误差±10℃)的单命中率分别为94%和96%,双命中率达到90%。通过与现有的方法比较,所提出的预测模型取得了最优的结果,不仅能够指导实际生产,也可用于冶金行业的其他应用背景的数学建模。
Converter steelmaking is an extremely complex physical and chemical reaction process,and the research on prediction model with intelligent method is a hot spot in recent years. Aiming at the end-point hit rate of carbon content and temperature,a new modeling method of static prediction was proposed. Based on the traditional twin support vector regression,the wavelet weight matrix was introduced into the objective function. Then,the objective function was transformed to the unconstrained optimization problem,which improved the performance and computational efficiency of the algorithm. Finally,the datasets of 260 t BOF were collected from some steel plant to establish the static prediction model. The experimental results showed that the end point carbon mass fraction(error±0.005 %) and temperature(error±10 ℃) of the prediction model achieved a hit rate of 94 % and 96 %,respectively,and the double hit rate was 90 %. Compared with the existing methods,the proposed prediction model achieved the optimal results,which could not only guide the actual production,but also was used for the mathematical modeling of other application background in metallurgical industry.
Converter steelmaking is an extremely complex physical and chemical reaction process,and the research on prediction model with intelligent method is a hot spot in recent years. Aiming at the end-point hit rate of carbon content and temperature,a new modeling method of static prediction was proposed. Based on the traditional twin support vector regression,the wavelet weight matrix was introduced into the objective function. Then,the objective function was transformed to the unconstrained optimization problem,which improved the performance and computational efficiency of the algorithm. Finally,the datasets of 260 t BOF were collected from some steel plant to establish the static prediction model. The experimental results showed that the end point carbon mass fraction(error±0.005 %) and temperature(error±10 ℃) of the prediction model achieved a hit rate of 94 % and 96 %,respectively,and the double hit rate was 90 %. Compared with the existing methods,the proposed prediction model achieved the optimal results,which could not only guide the actual production,but also was used for the mathematical modeling of other application background in metallurgical industry.
引文
[1] 孙永涛,吴永刚,秦波.基于IPSO优化BP的转炉炼钢终点预测研究[J].内蒙古科技与经济,2017(19):71-73.
[2] 祁子怡,高坤,赵宝芳,等.基于RBF神经网络在转炉炼钢终点预报中的应用研究[J].无线互联科技,2017(4):106-107,129.
[3] 朱亚萍,王文龙,徐生林.基于量子微粒群的BPNN在转炉炼钢静态模型中的应用[J].机电工程,2011,28(5):598-600.
[4] 李长荣,赵浩文,谢祥,等.基于L-M算法BP神经网络的转炉炼钢终点磷含量预报[J].钢铁,2011,46(4):23-25,30.
[5] PENG X. TSVR: an efficient twin support vector machine for regression[J]. Neural Networks the Official Journal of the International Neural Network Society,2010,23(3):365-372.
[6] RASTOGI R,ANAND P,CHANDRA S. A v-twin support vector machine based regression with automatic accuracy control[J]. Applied Intelligence,2016,46:1-14.
[7] XU Y,LI X,PAN X,et al. Asymmetric v-twin support vector regression[J]. Neural Computing and Applications,2017(2):1-16.
[2] 祁子怡,高坤,赵宝芳,等.基于RBF神经网络在转炉炼钢终点预报中的应用研究[J].无线互联科技,2017(4):106-107,129.
[3] 朱亚萍,王文龙,徐生林.基于量子微粒群的BPNN在转炉炼钢静态模型中的应用[J].机电工程,2011,28(5):598-600.
[4] 李长荣,赵浩文,谢祥,等.基于L-M算法BP神经网络的转炉炼钢终点磷含量预报[J].钢铁,2011,46(4):23-25,30.
[5] PENG X. TSVR: an efficient twin support vector machine for regression[J]. Neural Networks the Official Journal of the International Neural Network Society,2010,23(3):365-372.
[6] RASTOGI R,ANAND P,CHANDRA S. A v-twin support vector machine based regression with automatic accuracy control[J]. Applied Intelligence,2016,46:1-14.
[7] XU Y,LI X,PAN X,et al. Asymmetric v-twin support vector regression[J]. Neural Computing and Applications,2017(2):1-16.