【作者】 丁敬国；

【导师】 刘相华； 胡贤磊；

【作者基本信息】 东北大学 ， 材料加工工程， 2009， 博士

【摘要】 中厚板轧制过程对现场检测仪表的精度要求很高,考虑到中厚板轧制过程工作环境非常恶劣,各种干扰直接影响仪表检测精度,一旦检测仪表出现故障或精度下降,都将直接影响正常生产。本文结合国内某3500mm中厚板现场实际生产,开发了中厚板轧制过程软测量系统平台,根据中厚板生产工艺结合现场已有各种检测仪表,采用合理的软测量模型对在线控制参数进行测量,并在此基础上开发人工免疫网络对在线仪表进行故障诊断,从而提高过程控制精度以及产品的质量。主要的研究进展如下:(1)建立数据采集和处理平台。研究了采用OPC和Socket流式套接字技术实现基础自动化和软测量系统之间实时数据的数据通讯,开发了以Oracle 10g为数据库的数据采集系统,搭建了数据处理平台,对随机误差数据、错误数据、噪音数据以及粗大数据进行处理,为处理软测量系统面对的海量数据提供了可靠的平台。(2)提出了中厚板轧制过程产品尺寸参数的软测量方法。分析了前滑和钢板打滑对钢板速度的影响以及热膨胀及氧化铁皮对钢板体积变化的影响,研究了轧件在展宽轧制阶段横向流动规律,并考虑了发生打滑时电流的变化趋势以及钢板头尾不均匀变形头部宽度变化规律,推导出能够实用的在线长度、厚度和宽度软测量模型。提出了打滑程度因子概念,通过实测电流变化定量评价打滑现象,采用族划分和族相关软测量在线校正技术,对厚度、宽度、温度和钢种进行族划分,应用灰色关联度模型计算族与族之间关联度获取软测量模型的最优校正因子。在线实际应用表明,各种规格产品的厚度软测量值与人工卡量各点平均值的偏差均在±0.08mm之内,宽度软测量偏差均在0.4%之内。(3)提出了中厚板轧制过程变形抗力软测量方法。分析了化学成分、轧制温度、应变速率以及变形程度等因素对变形抗力的影响,对于控制轧制工艺,残余应变对变形抗力的影响比较大,分析合金元素Nb以及奥氏体晶粒尺寸变化对残余应变的影响,构建了残余应变模型以及基于径向基神经元网络的变形抗力软测量方法,通过对不同钢种、应变速率条件下的软测量值和实验值进行对比,二者吻合较好。(4)研究基于软测量技术的温度推断控制方法。分析了热辐射和对流、高压水除鳞、与轧辊接触产生的热传导、轧制塑性功等主要因素对中厚板温度变化的影响,研究了黑度系数与板厚之间的关系,构建了氧化铁皮随时间变化的数学模型,根据氧化铁皮厚度变化动态修正热辐射的换热系数。对测温仪实测温度进行置信区间计算,采用部分最卸朔ㄉ杓仆评砉兰破?以可信度最高的测温仪为基准,对在线其它测温仪测量值进行推理估计并确定其测量偏差。(5)开发基于软测量技术的仪表故障诊断系统。介绍了在中厚板生产在线检测仪表系统结构与功能概况,分析了检测仪表检测数据之间的相互关系。研究了人工免疫网络诊断技术,针对仪表实时检测数据,采用亲和度评价、个体浓度评价、激励度计算、免疫选择、克隆、变异、克隆抑制和种群刷新等算法构建在线检测仪表人工免疫诊断网络,通过实际应用表明,该系统对在线检测仪表故障诊断有较好的效果,为轧制过程控制数据分析和仪表故障诊断分析提供合理的指导。上述研究成果已经在我国某中厚板厂得到应用,使生产中遇到的一些难题得到解决。中厚板生产软测量技术是一个广阔的研究领域,其理论和方法需要进一步丰富、发展和完善,有许多问题有待进一步研究解决。更多还原

【Abstract】 The dependence of measuring instruments is extraordinary high for plate rolling process, but producing condition of which is so formidable that all kinds of interference will affect the accuracy of measuring instruments, if there is an error or accuracy decreases in measuring instruments, the normal manufacture of plate will be affected. This paper combines with actual manufacture of national certain 3500mm plate mill, exploits soft sensor system platform of plate processing, combines with all kinds of field measuring instruments and plate production technology, measures on-line controlling parameters by rational soft sensor models and exploits artificial immune networks for fault diagnosis, so it will improve the process control accuracy and product quality. The specific research contents were shown as follows:(1) Data acquisition and handling platform was established. Study the data communication of real-time data between basic automation system and soft sensor system by OPC technique and stream Socket technique, exploit data acquisition system which used oracle 10g as database, built up data handling platform, which deal with random error data, misdate, noise data and rough data, and it supplied a reliable mass data platform for soft sensor system.(2) Product size parameters soft sensor method of plate rolling process was built. The impact of steel plate velocity by forward slip and steel dragging as well as the influence for steel plate volume changing by thermal expansion and iron scale were analysed, lateral flow regular pattern of work piece at spreading stage was studied, the current change tendency when steel dragging happens and changing regular pattern of head width when steel plate tongue shape deformation happens were considered, and then high accuracy on-line soft sensor models of length, thickness and width were deduced. A forward slipping extent conception was presented which evaluated the slipping phenomenon quantitative based on the changing of current. At the same time it put forward soft sensor on-line correction technique as race division and race correlation, which devised the race by thickness, width, temperature and steel grade, and then optimal division factor of soft sensor model is obtain by using association degree model which calculates the association degree of races. On-line actual application shown that the deviation between thickness soft sensor value and average of manual measuring could be controlled within 0.2mm, deviation of width soft sensor could be controlled within 8mm.(3) Deformation resistance soft sensor method of plate rolling process was built. The influences for deformation resistance by factors such as chemical composition, rolling temperature, strain rate, deformation and so on were analysed, for the residual strain affectes resistance to deformation greatly as to control rolling process, the influences to residual strain by Nb and grain dimension of austenite were also it analysied at the same time, and the residual strain model and deformation resistance soft sensor method by RBF neural networks were built up, by comparison of soft sensor value and laboratory value at condition of different steel grade and strain rate, the accuracy between measuring value and experiment value of this method was almost anastomotic.(4) Studies on rolling temperature inference controlling based on soft sensor technique were carried out. The influences to temperature changing of plate by major factor such as heat radiation, convection, high-pressure water descale, thermal conductivity of roll contact, rolling plastic work done and so on were analysed, and the relationship between blackening coefficient and plate thickness was analysed, then mathematical models in which thickness of iron scale changes as time going on were built up, which dynamic corrected the heat transfer coefficient of heat radiation as the changing of iron scale thickness. The confidence interval of pyrometer measured temperature and designs inference estimation implement by using partial least square solution were calculated, which considered pyrometer with highest confidence as conference, inference and estimates measured values of other on-line pyrometers and determines their measuring deviation.(5) Measuring instruments fault diagnose system based on soft sensor technique was built. The system structure and function survey of on-line measuring instruments were introduced, and the relationship among measuring instruments datas were analysed. Fault diagnosis technique of artificial immune networks was used based on real-time data of instruments, which built up by amicable numbers evaluation, individual density evaluation, calculation of motivation degree, immune selection, clone, dissociation, clone rejection and swarm update and so on. From actual application it shows that there was a better effect to fault diagnoses of on-line measuring instruments of this system and it supplied a reasonable guiding action to rolling process controlling and instruments fault diagnose.Research findings above have been applied to some national plate mill, which solved several puzzles in production. Plate production soft sensor technique was also an extensive research field, whose theory and method should be abundant, developed and sounded further, and many problems should be solved further.更多还原