基于改进ICA算法的高炉故障诊断方法

发布时间：2018-01-28 02:41

本文关键词： 高炉故障 ICA MEWMA SVM　出处：《浙江大学》2017年硕士论文　论文类型：学位论文

【摘要】：钢铁作为重要的基础原材料,在我们的日常生活与生产中随处可见,其地位极其重要。有资料表明,中国目前已成为全球最大的钢铁生产国,同时也是最大的钢铁消费国。我国钢铁冶炼产业发展很快,已成为国民经济中的重要支柱产业。然而,必须注意到,我国的钢铁冶炼产业背后存在的问题不容小觑,如能耗过高;产业过于分散;更多的是生产较为低级别的钢材,在高品质钢材的生产方面能力较差;供应链管理水平较差等。为此,必须努力提高我国钢铁产业的产品品质,降低能源消耗。高炉是钢铁冶炼生产中的重要装备,能耗巨大,对后续冶炼品质有重要影响。因此,为了降低钢铁产业能耗,提高钢铁产品品质,本文从高炉安全顺行的角度研究其故障监测与诊断问题。高炉本身是一个非常复杂而又庞大的系统,其生产过程的机理模型亦非常复杂,涉及多个不同的生产模态,要实现高炉冶炼安全顺利,就必须对过程中出现的故障进行准确的检测与诊断,从而做出相应的处理,保证炉况平稳进行。考虑到实际高炉生产过程中,现场的数据采集装置会收集大量数据,且这些数据大多服从非高斯分布,故考虑在独立元分析算法(IndependentComponentAnalysis,ICA)的基础上,采用数据驱动的方法,对高炉故障展开研究。然而,ICA算法本身存在一定缺陷,也不完全适用于高炉故障,故首先对ICA算法进行改进,从独立元的数量以及重要性排序方面入手,提出了一种改进ICA算法,在一定程度上提高了故障检测的准确性。其次,针对高炉故障缓慢变化的特点,将ICA算法与多元指数加权平均(Multivariate Exponent Weighted Moving Average,ME WMA)方法结合,通过加权的方式,将历史数据的影响更好地考虑在内,改善了对于高炉故障的检测效果。最后,考虑到高炉冶炼过程中,操作从安全性的角度出发,往往会在可能出现故障之前就采取了相应措施,因此故障样本有限。故本文将针对小样本的支持向量机(Support VectorMachine,SVM)算法与改进的ICA算法结合使用,一旦采用改进的ICA算法检测到有故障发生,即采用SVM算法对故障的类型进行诊断。
[Abstract]:As an important basic raw material, iron and steel can be seen everywhere in our daily life and production, and its status is extremely important. According to some data, China has become the largest steel producer in the world. China's iron and steel smelting industry has developed rapidly and has become an important pillar industry in the national economy. However, attention must be paid to it. The problems behind the iron and steel smelting industry in China should not be underestimated, such as high energy consumption; The industry is too scattered; More is the production of lower grades of steel, in the production of high-quality steel capacity is poor; The level of supply chain management is poor. Therefore, it is necessary to improve the product quality and reduce energy consumption in China's iron and steel industry. Blast furnace is an important equipment in the production of iron and steel, and the energy consumption is huge. Therefore, in order to reduce the energy consumption of iron and steel industry, improve the quality of iron and steel products. In this paper, the problem of fault monitoring and diagnosis of blast furnace is studied from the point of view of safe and direct operation. The blast furnace itself is a very complex and huge system, and the mechanism model of its production process is also very complex. It involves many different modes of production. In order to realize the safe and smooth smelting of blast furnace, it is necessary to accurately detect and diagnose the faults in the process and make corresponding treatment. In view of the actual blast furnace production process, the field data acquisition device will collect a large number of data, and most of these data are distributed from non-#china_person0#. Therefore, on the basis of Independent component Analysis (ICA), a data-driven approach is proposed. The research on blast furnace fault is carried out. However, the ICA algorithm has some defects and is not suitable for blast furnace fault, so the ICA algorithm is improved first. Starting with the number and importance of independent elements, an improved ICA algorithm is proposed to improve the accuracy of fault detection to some extent. Secondly, aiming at the characteristics of slow change of blast furnace faults. The ICA algorithm is combined with the multivariate exponential weighted average multivariate Exponent Weighted Moving Average. Combined with ME WMA method, the influence of historical data is taken into account better by weighted method, and the detection effect of blast furnace fault is improved. Finally, considering the blast furnace smelting process. Operations from the point of view of security, often before the possibility of failure to take appropriate measures. Therefore, the support vector machine support machine (SVM) algorithm for small samples is combined with the improved ICA algorithm. Once the improved ICA algorithm is used to detect the fault, the SVM algorithm is used to diagnose the fault type.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TF576.7

【相似文献】