合理的轧制规程是使轧制过程达到最佳状态的重要保证。规程设定中采用具有自学习功能的BP神经网络取代传统轧制力数学模型，选用LevenbergMarquardt算法对轧制力进行预报。采用等相对负荷目标函数，考虑到现场和设备所受限制，确定约束条件。利用罚函数法将有约束的最优问题转换成无约束的最优问题，对某厂冷连轧现场规程进行了优化设计，并对优化前后的轧制规程进行了分析和比较，优化效果令人满意，满足实际生产要求。
The reasonable rolling schedule is an important guarantee to make the rolling process achieve the optimum condition. BP neural network with selflearning function is adopted to replace traditional rolling force models and then LevenbergMarquardt algorithm is adopted to predict the rolling force. The equal relative load object function is used and the constrained condition considering the limitation of the field and facility is confirmed. Sequential unconstrained minimization technique (SUMT) is used to change the optimal problem from constrained condition to unconstrained condition when the mill’s use of relative load as the optimization goal in a factory field is set. Analysis and comparison with existing schedules are offered, and the performance of the optimal rolling schedules is satisfactory.
参考文献
- 下载量()
- 访问量()
- 您的评分：
-
10%
-
20%
-
30%
-
40%
-
50%