薄板连铸GA-LM-BP漏钢预报模型研究  

Breakout Prediction in Thin Slab Continuous Casting Process Based on GA-LM-BP Neural Network

作　　者：张本国[1,2] 展邦华[3] 刘军[1,2] 夏建生[1,2] 

机构地区：[1]盐城工学院机械工程学院,江苏盐城224051 [2]江苏省模具智能制造工程技术研究中心,江苏盐城224051 [3]郑州机械研究所,河南郑州450000

出　　处：《铸造技术》2017年第8期1936-1939,共4页Foundry Technology

基　　金：江苏省基础研究计划资助项目(BK20150429);国家自然科学基金资助项目(51505408)

摘　　要：针对传统BP神经网络在训练过程中存在收敛速度慢、易陷入局部最优解的缺陷,将遗传算法全局寻优能力与LM算法的局部寻优能力引入到BP神经网络训练过程,建立了GA-LM-BP神经网络漏钢预报模型;结合某钢厂连铸现场历史数据,对该预报模型进行了训练和测试。结果表明,GA-LM-BP漏钢预报的收敛速度较传统BP神经网络明显加快,其泛化能力和对漏钢温度特征的识别精度也有了较大提高。Slow convergence and local optimal solution in the training process are two terrible drawbacks of the traditional BP neural network. The global optimization ability of genetic algorithm and the local optimization ability of LM algorithm were introduced into the training process of the BP neural network to improve its converge property, and then a GA-LM- BP neural network was established. The GA-LM-BP neural network model was trained and tested with the historical data collected from a steel plant. The testing results show that the convergence rate of the GA-LM-BP neural network model is faster than the traditional BP neural network significantly. The generalization capability and the recognition accuracy for the temperature characteristics of the breakout prediction system are greatly improved after using GA-LM-BP neural network.

关 键 词：薄板连铸 漏钢预报 遗传算法 LM算法 BP神经网络 

分 类 号：TF777.1[冶金工程—钢铁冶金]