During the continuous casting process, the instantaneous abnormal mold level fluctuation has great detrimental effects on slab quality, thus the mold level fluctuation is a key parameter for continuous casting process of high-quality steel. In the present study, the slab continuous casting process data for low carbon steel, medium carbon steel, hypo-peritectic steel and peritectic steel are collected to investigate the effect of process parameter on the instantaneous abnormal mold level fluctuation. The frequency domain coherence results illustrate that the stopper-rod position is the key influence factor on the instantaneous abnormal mold level fluctuation during the continuous casting process of different steel grades. The time–frequency characteristics of instantaneous abnormal mold level fluctuation and stopper-rod position were analyzed by continuous wavelet transform (CWT), and the results show that before the instantaneous abnormal mold level fluctuation occurrence, the high-frequency region amplitude of stopper-rod position shows a linear increase trend. For the continuous casting process with constant casting speed, the increase rate is within the range of 0.1/s to 0.2/s. For the instantaneous abnormal mold level fluctuation induced by steep casting speed change, the increase rate can reach 0.67/s. For the instantaneous abnormal mold level fluctuation induced by ladle change, the increase rate can reach 0.85/s. Therefore, the instantaneous abnormal mold level fluctuation can be predicted by CWT analysis for the high-frequency region amplitude of stopper-rod position.
为了解决国内工业加热炉普遍存在的燃烧效率低、自动控制投用率不高等问题,针对目前加热炉的燃烧控制方案进行了改进性的研究.在传统自动燃烧控制方案的基础上,创新性地研究设计了三参数串级交叉自动燃烧过程控制系统,即在传统加热炉自动燃烧控制回路仅利用空煤配比系数实现燃烧控制的基础上,增加了煤气热值和残氧值两个重要参数参与回路调节,实现了三参数共同作用的串级交叉自动燃烧过程控制方案,并通过仿真模型验证最大动态偏差率平均为62％,平均调节时间为20 s,平均衰减比为8∶1,改进后控制方案的稳定性、准确性和快速性等性能指标均较传统燃烧控制方案优越,具有较好的应用前景.
通过与钢铁企业实际生产相结合,面向订单坯料设计时多种可选坯型的特殊场景,充分考虑了炉次计划编制过程中所涉及的客户特殊要求、生产工艺约束、设备约束等条件,以降低钢水冶炼成本、减少无委托钢坯库存、提升客户服务水平为目标,基于约束理论建立了数据模型.通过挖掘模型特点,提出了一种基于贪婪算法和回溯法、约束传播方法混合的约束满足问题的求解方法.该算法与传统人工编制计划相比,计划编制效率平均提升2 897倍,钢包利用率平均提高0.56％,人工调整率平均为2.38％.试验结果表明,本文所提供的炉次批量计划优化方法有效解决了人工编制计划模式下的效率低下、人工经验依赖程度高、重复性工作量大等问题.
在连铸过程中,结晶器液面波动是限制连铸速度和铸坯质量的关键参数之一,因此,液面波动行为的准确预测一直是冶金学者的研究重点.基于此,本文利用Python对结晶器液面波动的振幅值进行预测.首先,选取中间包的质量、塞棒的位置、拉力和结晶器振动作为模型的输入,对数据快速傅里叶变换和归一化处理.然后,构建4×3×1的反向传播(back propagation,BP)神经网络模型,并利用鲸鱼算法(whale optimization algo-rithm,WOA)对初始权值和阈值优化.通过训练预测,相比BP神经网络,WOA-BP神经网络能较好地对结晶器液面波动进行预测,且预测值与结晶器液面波动振幅吻合较好,拟合决定系数(R2)为0.841 4;当振幅偏差为±0.02时,命中率可达到91％.
In order to solve the problem of low set and control precision of traditional roll bending force setting model for tandem cold rolling in production practice, a hybrid roll bending force prediction model based on fruit fly optimization algorithm（FOA） and generalized regression neural network（GRNN） was proposed.The smoothing factor parameters of GRNN network is optimized by FOA algorithm to ensure the best performance of the model.The hybrid FOA-GRNN roll bending force prediction model for tandem cold rolling in this paper was compared with the corresponding back propagation neural network（BPNN） prediction model.The comprehensive performance of the two models were evaluated by error indexes.It is proved that the hybrid FOA-GRNN model can better predict the roll bending force compared with BPNN model in tandem cold rolling.
针对钢铁企业订单呈现多品种规格、小批量趋势下的生产计划组织难点问题,充分考虑订单结构、设备、原料、工艺等因素,以钢铁企业生产计划为主线构建了钢铁企业智能钢轧一体化管控平台.平台以数据中心为支撑,融合工厂模型、业务模型,并集成算法求解引擎,以支撑订单计划、批量计划、生产调度等业务需求.在管控平台基础上能够实现钢铁企业各个业务系统之间的数据共享,达到综合决策、协同优化的目标.该系统在宁波钢铁有限公司投入使用,在生产计划组织方面起到了较好的效果,为智能化转型奠定了基础.
At present, the intelligent rating and impurity deduction schemes for scrap in the industry are based on the target recognition model, but compared with the semantic segmentation model, the target recognition model cannot accurately depict the boundary of scrap, resulting in inaccurate area estimation and feature collection of scrap.However, there are many semantic segmentation models at present.How to select a model suitable for scrap grading scenarios is a problem to be solved.To solve this problem, a 1∶3 physical model was established in the laboratory to simulate different types of scrap entering the factory for quality inspection. Then 2K resolution camera was used to collect image data. Finally,20 mainstream semantic segmentation models were compared and analyzed. The experiment shows that on 139 scrap datasets,full convolutional network（FCN） model and high resolution net（HRNet） are used to segment scrap semantically. The improved efficient semantic segmentation model SegFormer-B5 based on Transformer should be used on the 1 529 scrap steel data sets with image enhancement to predict and classify scrap. From the mean intersection over union（ mIoU） index, the HRNet backbone network used by FCN and object-contextual representations network（OCRNet） is 6. 6 percentage points higher than the residual network（ResNet）.
Oxygen is an important gas energy in the process of converter steelmaking.Accurate prediction of oxygen consumption in converter steelmaking is conducive to improve the stable control of steelmaking process.In addition, it is also a powerful guarantee for the orderly operation of steelmaking plant.According to the mechanism characteristics of converter steelmaking process, IC-XGBoost data model considering interval-like constraints was established to improve the prediction precision and accuracy.The improved XGBoost model was comparatively tested by using the actual data of converter steelmaking process with noise disturbance.Compared with traditional data models such as neural network, SVM and XGBoost, the established model can efficiently and effectively obtain satisfactory prediction precision and accuracy, which has good anti-disturbance ability.
Iron and steel industry is not only the pillar of the national economy, but also the major energy consumption section considering the whole society.With the raise of goals such as "emission peak hit" "carbon neutrality reach",greenization is currently imperative for iron and steel industry.Owing to the fact that the production process and key equipment are comparatively complete and fixed, energy management optimization becomes an effective way for energy saving, emission reduction and low-carbon running.As a result, related methods and techniques consequently become hotspots for research and application in this field.Aiming at energy management and optimization for iron and steel industry, the research progress was summarized in four aspects, including state awareness and trend forecasting, real time optimization and equipment control, scheduling decision-making and system optimization, platform development and engineering practice, covering multiple theoretical and technical systems such as mechanism modeling, data driven, industrial Internet, artificial intelligence,etc. Finally,taking current status and development trend into consideration,some challenges and future topics were summarized.
为确定焦比、煤比、燃料比等经济技术指标调控与高炉利用系数之间的关系,结合贝叶斯优化后的极端梯度提升(extreme gradient boosting,XGBoost)回归高炉利用系数预报模型在多元线性和回归等方面的优势,借助灰狼优化算法(gray wolf optimization,GWO),构建了高炉利用系数提升的鼓风制度预报模型.在经济技术指标可调控区间上,智能推荐技术指标不超过20％的区间内,智能推荐模型利用系数提升达9.1％以上,且系统运行稳健、仿真效果有效,具有在钢铁企业可观的应用前景和推广价值.
在环保与去产能化的双重影响下,中国钢铁开始向高质、智能、绿色的生产模式转变,传统的高能耗、高污染的高炉冶炼理念已不再适用于"十四五"规划的发展方向.随着大数据与人工智能技术的兴起,新一代钢铁工业在智能制造的推动下向着绿色制造迈进,通过分析钢铁企业多年积累的数据而建立各种预测模型已成为一种大趋势.本文首先以高炉智能化转型作为研究背景,通过由简入繁的方式介绍了当前高炉冶炼指标预测模型及冶炼过程监测系统.然后,分析了数据处理与专家决策优化策略的重要性,并简要阐述了当前各企业高炉大数据云平台的搭建情况.最后,对高炉智能化转型作出了相应的结论与展望.
Blast furnace gas（BFG） is an important secondary energy generated in the iron-making process of iron and steel enterprises, and optimal scheduling of its consumption can help reduce the carbon emission and improve the economic efficiency.In view of the lack of typical scheduling scenarios for BFG system, a scheduling scenario generation method for BFG system based on improved generative adversarial network（GAN） was proposed. A scheduling scenario with generation and consumption differences, gas tank level and adjustable unit consumption flow as elements was established,and by decomposing them,multiple generators were used to learn to fit the distribution characteristics of different scenario elements to reduce the data fluctuations caused by the interactions among the elements. In addition,Wasserstein distance was used as the loss function,and a gradient penalty（GP） strategy was introduced to improve the stability and convergence speed of the model training process. Experiments were conducted with the actual operation data of BFG system in a large domestic steel enterprise,and the results show that the generated scheduling scenario set conforms to the statistical characteristics and temporal correlation of the actual operation process data of BFG system,which verifies the effectiveness of the proposed method,and the scheduling scheme based on the generated scenario set can ensure that the gas tank can operate stably within the safety interval.
连铸坯低倍质量检测是评价连铸坯质量的重要手段,正广泛应用于连铸生产过程.然而,钢厂大多采用人工方法对连铸坯低倍质量进行评价.这种人工方法依赖于评级人的经验,缺乏检验的一致性、客观性及准确性.为了准确地对连铸坯凝固组织及中心偏析评级,以U-Net网络为基础,集成了残差模块、金字塔池化模块(pyramid pooling module,PPM)以及注意力模块,提出了 APR-UNet模型.模型中,残差模块可避免深层网络出现退化问题;PPM可聚合不同区域的上下文信息,加强模型感受野,以提高网络获取全局信息的能力;注意力机制可抑制无用信息,提高模型对铸坯凝固组织及缺陷的分割精度及模型鲁棒性.使用相同数据集分别训练U-Net模型和APR-UNet模型.试验表明,对连铸坯等轴晶区分割,APR-UNet模型的交并比(inter-section over union,IOU)达93.54％,较U-Net模型提高了 1.06％;对中心偏析的评级,APR-UNet模型的评级成功率达91.2％,较U-Net模型提高了3.6％.APR-UNet模型有效改善了原模型分割结果中出现的过分割现象,在连铸坯凝固组织及缺陷的评级方面具有很大潜力.
轧机是广泛用于钢铁及有色金属轧制生产的核心关键设备,其驱动电机及电气传动系统具有装机容量大、系统可靠性高、过载能力强、动态响应快等显著特点.在回顾总结轧机电气传动系统发展历程的基础上,比较了几种用于轧机主电机控制相关的大功率电力电子器件及其系统方案,重点介绍了采用新型全控型电力电子器件的三电平脉冲宽度调制(pulse width modulation,PWM)中压变频传动系统的结构和技术特点,同时对国内外目前大功率轧机电气传动系统的研究现状及发展趋势进行了展望.
随着5G、移动边缘计算(mobile edge computing,MEC)、云平台技术的发展,超700 Mbps大上行5G专网已成功落户制造行业,这也促使工业非结构化实时数据参与云化产品质量分析成为可能.结合钢铁行业特点,阐述了基于终端层、网络层、云化层和模型层的5G+云表检技术框架.同时,针对5G工业专网下私有云表检系统给出了解决方案,重点介绍了系统多种实现方式、优化网络承载、分布式软件设计、多元化模型融合、流程化算法编排等方面的内容.最后介绍了鞍钢集团在5G工业专网下云化带钢表面质量检测系统的应用及发展情况.
In order to solve the contradiction between batch production modes of hot galvanized sheet and personalized demands of customer orders, reduce the variety and size switch in the production process, so as to improve the production efficiency and reduce the production cost, an optimization model of hot galvanized production scheduling was established.Considering the key scheduling factors such as material priority, roll change programming, post-processing method, specification switching programming and production economy in hot dip galvanizing production, a material priority setting method was proposed based on planner′s know-how, and then a hot dip galvanizing optimization scheduling method was proposed by combining material priority with search algorithm.In this scheduling method, the batch type is first determined according to the material, and then the batch materials are initially sorted to generate the scheduling plan set.Then, the remaining materials are inserted into a suitable scheduling set to obtain a minimum scheduling plan of hot dip galvanizing unit. Finally,the search strategy is used to further optimize the scheduling plan and obtain the final scheduling plan.Based on the above methods,an optimization scheduling system of hot dip galvanizing unit was developed. The application results show that the above methods can significantly improve the planner′s work efficiency and scheduling quality.
为了确保有效利用轧机设备能力,以某钢厂棒材生产线的轧机电机负荷数据为研究对象,利用PyTorch搭建基于长短期记忆(long short term memory,LSTM)神经网络的预测模型,定义模型初始网格结构参数,选定单元结构激活函数,并针对模型超参数的选择问题,采用自适应学习算法(adaptive moment estimation,Adam)进行参数优化,迭代降低损失值,提高模型的预测精度.通过试验设计,采用生产两种规格棒材的轧机负荷数据进行验证,结果表明,与未优化的负荷预测模型对比,均方误差SME分别降低了 3.28、1.76,证明了所建立模型的预测效果更好,具有较高的稳定性.
As for the variable load problem of coal and by-product gas mixed fuel boilers in iron and steel industrial parks, an intelligent anti-disturbance controller was designed for achieving multi-mode predictive control of boilers of different types/specifications in complex environments based on the analysis of boiler nonlinearity, large time delay, and multi-mode characteristics.A dynamic model of data driven boiler under all operating conditions was established based on a linear variable parameter model.On this basis, a Smith predictive compensator was used for the time-delay compensation to improve the stability and rapidity of the control system.Considering the impact of different types of disturbances in the complex industrial environments, disturbance observers were deeply integrated with model predictive control（MPC） to suppress input and output disturbances.Different types of boiler scenarios were selected for validation analysis.Compared to existing algorithms, the results show that the proposed algorithm has significant advantages in suppressing disturbances with the tracking average percentage error increasing of 0.35% and the root mean square error decreasing of 0.164.At the same time, it can also be used for solving the variable load control problem of different types/specifications of boilers in iron and steel industrial parks with good scalability.
There are many factors that affect the FeO content of sintered ore, and there is a nonlinear relationship between the FeO content and each factor, making it difficult to predict.Aiming at the problem that the content of FeO in sinter is difficult to predict directly, a prediction model of FeO content in sinter was proposed, which combined Dropout algorithm, Adam algorithm and four-layer BP neural network.In order to improve the prediction accuracy of FeO content in sintered ore, combined with the sintering process, the temperature characteristics of the thermal imaging key frame of the tail section of sinter with strong correlation with the FeO content of sintered ore were specially selected as the parameter input of the model.The Dropout algorithm was used to improve the structure of the four-layer BP neural network, and the Adam algorithm optimized the training process of the four-layer BP neural network,so as to improve the prediction accuracy and generalization ability of the model.Experiment shows that when the error values of FeO content in sinter predicted by the improved model are ± 0. 5, ± 0. 8 and ± 1. 0,respectively,the hit rate reaches 77. 42%,88. 71% and 96. 77%.Compared with the three-layer BP neural network prediction model and the support vector regression（ SVR） model,the error of this model is smaller,and the prediction accuracy is also significantly improved.
The blast furnace operation profile is closely related to blast furnace long life, blast furnace operation and technical and economic indexes, etc.A reasonable operating furnace type is helpful to ensure high quality, low consumption, high production and long life of blast furnace production.The clustering analysis of the cooling stave temperature data of blast furnace can effectively and reasonably characterize the changes of blast furnace operation profile, which has important guiding significance for the blast furnace production.K-means and Gaussian mixture model（GMM） were used to cluster the data set based on cooling stave temperature data of Shasteel 5 800 m~3 blast furnace, and the clustering results were evaluated based on the principles of the two clustering algorithms,combined with Davies-Bouldin indicator（DBI） and Silhouette coefficient（ SC）,and analyzed the blast furnace smelting situation corresponding to the production status of the obtained cluster categories.The clustering results are better when the K-mean clustering algorithm is used and the furnace profile is clustered as 3,based on the sample data selected in this paper. And the average coke ratio,coal ratio,fuel ratio,gas utilization rate,hot metal temperature and output of the 3rd furnace profile corresponding to 357. 62 kg/t,163. 18 kg/t,512. 34 kg/t,47. 51%,1 502. 045 ℃ and 12 472. 59 t/d,respectively,so it is more suitable for the daily production of this blast furnace. This study can provide a strong reference for the selection of clustering algorithm and the evaluation of clustering results in the analysis of blast furnace ironmaking big data.
