基于粒子群算法优化支持向量机的铝热连轧机轧制力预报  

Rolling Force Prediction Based on Support Vectors Machine with Particle Swam Optimization

作　　者：杨景明[1,2] 陈伟明[1,2] 车海军[1,2] 吕金[3] 贾林[1,2] 

机构地区：[1]燕山大学电气工程学院,河北秦皇岛066004 [2]燕山大学工业计算机控制工程河北省重点实验室,河北秦皇岛066004 [3]天津电气传动设计研究所,天津300180

出　　处：《计量学报》2016年第1期71-74,共4页Acta Metrologica Sinica

基　　金：国家自然科学基金钢铁联合基金资助(U1260203);河北省科学技术研究与发展计划基金(10212157);河北省高等学校创新团队领军人才培育计划(LJRC013)

摘　　要：为了提高热轧带材的轧制力预报精度，提出了粒子群算法和支持向量机结合的方法来预报轧制力。根据轧制原理用支持向量机建立轧制力预报的模型，通过粒子群算法优化支持向量机参数来提高预报精度。为了进一步提高轧制力预报精度，还提出了支持向量机网络与数学模型相结合的方法，对某“1＋4”铝热连轧厂现场采集的5052铝合金轧制数据进行离线仿真，仿真结果可以看出支持向量机网络与数学模型结合的方法预报轧制力，提高了轧制力预报速度并使其轧制力预报精度控制在7％以内。In order to improve the prediction accuracy of rolling force in hot rolled strip, combined with particle swarm optimization algorithm and support vector machine method to predict the rolling force, According to the principle of rolling, rolling force prediction model is established by using SVM, the parameters of SVM was optimized by particle swarm algorithm to improve the prediction accuracy. In order to further improve the precision of rolling force prediction, network based on SVM combined with mathematic model method is proposed. The acquisition of a large number of rolling data offline simulation was based on a ＂1 ＋ 4＂ aluminum strip rolling factory. The simulation results can be seen that rolling force predicted the method of network based on SVM combined with mathematic mode, improves the speed of rolling force prediction and makes the forecast accuracy of rolling force control within 7%.

关 键 词：计量学 轧制力预报 支持向量机 粒子群 热轧 

分 类 号：TB93[一般工业技术—计量学]