论文部分内容阅读
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Levenberg-Marquardt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。
Reasonable rolling schedule is to ensure that the rolling process to achieve the best condition of the important guarantee. In the procedure setting, a BP neural network with self-learning function was used to replace the traditional mathematical model of rolling force. The Levenberg-Marquardt algorithm was used to predict the rolling force. Equal relative load objective function is adopted, taking into account the restrictions on site and equipment, to determine the constraints. The penalty function method was used to convert the constrained optimal problem into the unconstrained optimal problem. The site procedure of a tandem cold mill was optimally designed. The rolling procedure before and after the optimization was analyzed and compared. People are satisfied with the actual production requirements.