以工业炉管离心铸造生产过程监测数据为基础的产品质量早期识别

发布时间：2018-03-22 13:09

本文选题：工业炉管　切入点：离心铸造　出处：《北京化工大学》2013年硕士论文　论文类型：学位论文

【摘要】：工业炉管是乙烯裂解装置等石化生产装置中反应炉的核心组成部分,可以分为辐射段炉管和对流段炉管。其中裂解反应主要在辐射段炉管内进行,辐射段炉管需要长期在高温低压的条件下工作,容易产生渗碳、腐蚀、断裂等问题。因此,其质量的优劣对石化产品质量和石化生产安全起着重要的作用。辐射段炉管的生产大多数采用离心铸造的方式,因为采用这种方式所得的产品具有组织致密、晶型完整、金属夹渣物少和机械性能良好等优点,能够很好地适用于各种石化生产。在一些铸造企业中,工人大多根据生产经验来进行生产,并通过对单一参数的控制来保证炉管的质量。但炉管生产过程复杂,影响因素众多,且各影响因素之间存在一定的相关性,各因素协同影响炉管最终的质量。因此,依据经验或对单一参数控制的方法无法实现对产品质量严格控制和废品产生原因的有效查找。本文以某企业辐射段炉管铸造生产过程为对象,按照批次对每根炉管的离心铸造生产过程信息进行全流程的采集,具体包括物料信息,离心铸造过程信息和质检信息等。以这些过程采集的数据信息为基础,采用多元统计模型对炉管生产过程进行监测。操作工人可对模型预测有质量问题的炉管进行全面的检测,避免不合格产品混入合格产品中。本文以主元分析方法为基础,通过分析各参数对故障贡献的大小,从而找出引起产品质量不合格的主要参数,为实际生产过程提供参考,在下一根炉管的生产中及时调整生产条件,以保证生产的正常进行,达到提高炉管产品质量,降低生产成本的目的。因此炉管质量监测模型的建立具有重要的工业实际应用价值和研究意义。本文的主要研究内容包括： (1)对原料熔炼和离心铸造这两个影响炉管质量的关键环节进行了详细的分析。讨论了主要操作变量对炉管质量的影响。结合文献资料、生产经验和变量收集的可行性等因素确定了建立炉管质量监测模型所需的变量种类、变量采集方式和采集频率等重要内容。 (2)对建立炉管质量监测模型所需的训练数据进行预处理,以消除建模数据的异常点,保证模型的准确性。用交叉验证法确定模型最优主元个数,采用该模型对实际生产数据进行监测。 (3)为提高炉管质量监测模型的效果,讨论了建模变量的数量和种类以及建模数据的数量等因素的影响,并通过贡献图的方法对质量不合格炉管的生产数据进行分析,确定引起炉管质量问题的主要参数。将模型对炉管质量预测的结果同工厂实际检测结果相对比,可得模型正确预报质量不合格炉管的比率为86.67%,错误预报质量不合格炉管的比率为2.86%,对实际生产过程具有一定的指导意义。
[Abstract]:Industrial furnace tube is the core part of reaction furnace in petrochemical production plant such as ethylene pyrolysis unit, which can be divided into radiation section furnace tube and convection section furnace tube, in which the cracking reaction mainly takes place in the radiation section furnace tube. The radiation section furnace tube needs to work under the condition of high temperature and low pressure for a long time. It is easy to cause carburization, corrosion, fracture and so on. Its quality plays an important role in the quality of petrochemical products and the safety of petrochemical production. Most of the tubes in the radiation section are produced by centrifugal casting, because the products obtained by this way are compact in structure and complete in crystal form. The advantages of less metal slag and good mechanical properties can be well applied to various petrochemical production. In some foundry enterprises, most of the workers produce according to their production experience, and ensure the quality of the furnace tube by controlling the single parameter, but the production process of the furnace tube is complicated and the influence factors are many. There is a certain correlation among the influencing factors, and all factors influence the final quality of furnace tube. Therefore, the strict control of product quality and the cause of waste can not be effectively found according to experience or the method of single parameter control. In this paper, the whole process information of centrifugal casting of each furnace tube is collected according to the batch, which includes the material information, taking the casting process of the furnace tube in a certain enterprise as the object. Based on the data collected from these processes, the multivariate statistical model is used to monitor the tube production process. Based on principal component analysis (PCA), this paper analyzes the contribution of each parameter to the failure, and finds out the main parameters that cause the unqualified quality of the product, and provides a reference for the actual production process. Adjust the production conditions in time in the production of the next furnace tube in order to ensure the normal production and improve the product quality of the furnace tube. Therefore, the establishment of furnace tube quality monitoring model has important practical application value and research significance in industry. The main research contents of this paper are as follows:. 1) two key links which affect the quality of furnace tube are analyzed in detail, that is, raw material melting and centrifugal casting. The influence of main operating variables on tube quality is discussed. Factors such as production experience and the feasibility of variable collection determine the types of variables needed to establish the quality monitoring model of furnace tubes, variable collection methods and acquisition frequency and other important contents. In order to eliminate the abnormal points of the modeling data and ensure the accuracy of the model, the training data needed to establish the furnace tube quality monitoring model are preprocessed, and the optimal number of principal components of the model is determined by cross-validation method. The model is used to monitor the actual production data. In order to improve the effect of the quality monitoring model of furnace tube, the influence of the quantity and type of modeling variables and the quantity of modeling data are discussed, and the production data of unqualified furnace tube are analyzed by the method of contribution diagram. The main parameters causing tube quality problems are determined. The prediction results of furnace tube quality by the model are compared with the actual test results in the factory. The ratio of the unqualified furnace tube predicted correctly by the available model is 86.67, and the ratio of the unqualified furnace tube predicted by the error prediction is 2.86, which is of certain guiding significance to the actual production process.
【学位授予单位】：北京化工大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TQ086.3

【参考文献】