【作者】 张蕾；

【导师】 张利；

【作者基本信息】 辽宁大学 ， 软件工程（专业学位）， 2022， 硕士

【摘要】 随着汽车工业、机械制造业、日用五金等领域蓬勃发展,带钢需求量日益增加,且对带钢质量的要求也越发严格。带钢厚度精度是衡量带钢质量的重要指标,因此控制带钢出口厚度精度是轧制工业领域出口优质带钢成品亟需解决的首要问题。然而,在实际轧制过程中,影响带钢厚度的众多因素具有时变性、耦合性、严重非线性等特点,传统带钢厚度预测算法难以准确建模非线性关系并且往往忽略掉一些重要因素。与此同时,现存人工智能算法对非线性带钢时序序列特征挖掘还不够深入,预测精度不高。因此寻找一种新颖且适合处理带钢厚度预测问题的人工智能算法是本文研究的重点。近年来,深度学习算法给非线性时序预测问题带来新的解决方法。本文提出改进边牧算法优化长短期记忆网络(IBCO-LSTM)带钢厚度预测系统,由于LSTM预测精度受到隐藏层神经元数目和学习率等关键超参数影响,所以引用优秀的边牧优化算法对LSTM进行参数寻优。同时针对算法存在初始种群分布不均和种群部分个体寻优状态不准确的问题进行改进。首先,提出利用帐篷映射(Tent map)初始化种群分布,提高初始种群分布质量,进而改善算法全局搜索能力;其次,将动态加权策略引入种群部分个体速度更新公式,提高算法的局部寻优精度。最后利用IBCO搜索LSTM最佳参数,建立最优IBCO-LSTM预测模型,实现带钢厚度预测。本系统设计开发五个功能模块,依次为注册登录、用户管理、数据处理、预测模型构建和厚度预测模块。数据处理模块引用互信息对数据进行特征选择,然后采用最值归一化方法对数据进行归一化处理,最后将数据集划分为输入到模型中的训练集与测试集;预测模型构建模块利用IBCO搜索LSTM最佳参数,训练最优模型;厚度预测模块利用训练好的模型实现带钢厚度预测。本文将国内某钢铁集团实测带钢数据作为实验数据。通过对比实验验证IBCO-LSTM模型相较传统模型具有更好的预测效果。经后续反复测试,证实本文提出的IBCO-LSTM带钢厚度预测系统能够在带钢轧制领域中有效应用。更多还原

【Abstract】 With the vigorous development of automobile industry,machinery manufacturing industry,daily hardware and other fields,the demand for strip steel is increasing,and the quality requirements of strip steel are more stringent.The thickness accuracy of strip is an important indicator to measure the quality of strip,so controlling the thickness accuracy of strip export is the primary problem that needs to be solved urgently for the export of high-quality strip products in the rolling industry.However,in the actual rolling process,many factors affecting strip thickness have the characteristics of time-varying,coupling and serious nonlinearity.The traditional strip thickness prediction algorithm is difficult to accurately model the nonlinear relationship and often ignores some important factors.At the same time,the existing artificial intelligence algorithm for nonlinear strip sequence feature mining is not deep enough,and the prediction accuracy is not high.Therefore,finding a novel and suitable artificial intelligence algorithm for strip thickness prediction is the focus of this paper.In recent years,the deep learning algorithm has brought new solutions to the nonlinear time series prediction problem.In this paper,an improved Border Collie Optimization is proposed to optimize the strip thickness prediction system of longshort-term memory network model(IBCO-LSTM).Because the prediction accuracy of LSTM is affected by key super-parameters such as the number of hidden layer neurons and learning rate,the excellent Border Collie Optimization is used to optimize the parameters of LSTM.At the same time,the algorithm has the problems of uneven distribution of initial population and inaccurate optimization state of some individuals in the population.Firstly,tent map is proposed to initialize the population distribution,improve the quality of the initial population distribution,and then improve the global search ability of the algorithm.Secondly,the dynamic weighting strategy is introduced into the velocity updating formula of some individuals in the population to improve the local optimization accuracy of the algorithm in the search space.Finally,IBCO is used to search the optimal parameters of LSTM,and the optimal IBCO-LSTM model is established to predict strip thickness.The system design and development of five modules,registration login,user management,data processing,prediction model building and thickness prediction module.The data preprocessing module uses mutual information to select the feature of the data,and then normalizes the data using the maximum normalization method.Finally,the data set is divided into training set and test set as the input model data set.The prediction model construction module uses the improved edge-grazing optimization algorithm to search the optimal parameters of LSTM and train the optimal model;the thickness prediction module uses the trained model to realize the thickness prediction of strip steel.This paper selects the measured strip data of a domestic steel group as the experimental data.The comparison experiment verifies that IBCO-LSTM model has better prediction effect than the traditional model.After repeated tests,it is confirmed that the IBCO-LSTM strip thickness prediction system proposed in this paper can be effectively applied in the rolling industry.更多还原