基于混合高斯—朴素贝叶斯的高炉炉况判别系统

发布时间：2018-04-21 09:41

本文选题：高炉炉况判别 + 主成分分析　；参考：《浙江大学》2017年硕士论文

【摘要】：钢铁冶炼在我国国民经济中占有十分重要的地位,自2014年起,我国的钢铁产量占世界的比重已经高达40%。高炉炼铁是现代炼铁工艺的主要方法,高炉形体庞大、内部反应复杂、生产能耗极高。保证高炉在生产过程中的平稳运行是保持高产率、高能量利用率的基本条件,具有十分重要的意义。本文针对高炉炉况平稳性判别展开研究。目前,对高炉异常炉况的诊断已经有较多工作。先后出现了专家系统、人工神经网络和支持向量机等研究方法。但是,这些工作的关注点主要针对异常炉况发展的后期,是在异常炉况已经较为严重时对异常的原因进行诊断。而早期的炉况平稳性监测在实际生产中依然依赖工程师的个人经验。并且,这些模型的训练方法多为有监督训练,而实际生产中炉况的异常标签很难获得。针对这些问题,本文提出了一套基于混合高斯-朴素贝叶斯的无监督分类模型,着眼于早期的炉况平稳性监测,并且解决了数据无标签或者标签不可信的问题。同时,基于本文提出的炉况判别模型,设计并实现了一套高炉炉况智能判别系统,利用该系统可以针对高炉炉况数据进行可视分析、模型训练和实时监测。首先,本文对所用数据集进行了介绍与分析,对数据集中的异常值和缺失值进行了剔除、插补等处理。通过相关性分析发现了炉况参数数据的第一主成分与炉况的平稳性密切相关,据此选定了第一主成分的均值、标准差和最大值三个表示炉况平稳性的特征。然后,本文构建了一套基于混合高斯-朴素贝叶斯算法的、用于判别炉况平稳性的无监督分类模型,该算法可以在无监督的情况下进行训练,解决了传统模型面临的无标签或标签不可信的问题。本文将模型的性能与决策树、逻辑回归和支持向量机进行对比。本文模型在综合错误率、第一类错误率和第二类错误率等指标上均优于对比模型。最后,本文基于提出的模型,设计并实现了一套高炉炉况智能判别系统,可以用于高炉历史数据可视分析、模型训练和现场数据的实时监测。
[Abstract]:Iron and steel smelting plays a very important role in the national economy of our country. Since 2014, the proportion of iron and steel production in the world has reached as high as 40%. Blast furnace ironmaking is the main method of modern ironmaking process. It is of great significance to ensure the smooth operation of blast furnace in the process of production, which is the basic condition for maintaining high yield and high energy utilization ratio. In this paper, a study on the stability of blast furnace condition is carried out. At present, more work has been done in the diagnosis of abnormal blast furnace conditions. Several research methods such as expert system, artificial neural network and support vector machine have emerged. However, the focus of these works is mainly on the late stage of abnormal furnace condition development, which is to diagnose the cause of abnormal furnace condition when the abnormal furnace condition is already serious. The early smooth monitoring of furnace conditions still relies on the personal experience of engineers in actual production. Moreover, most of the training methods of these models are supervised training, but the abnormal labels of furnace conditions in actual production are difficult to obtain. In order to solve these problems, an unsupervised classification model based on mixed Gao Si and naive Bayes is proposed in this paper, which focuses on the early stationary monitoring of furnace conditions, and solves the problem of data without labels or labels being unreliable. At the same time, based on the furnace condition discriminant model proposed in this paper, a set of blast furnace condition intelligent discriminant system is designed and implemented. The system can be used for visual analysis, model training and real-time monitoring of blast furnace condition data. First of all, this paper introduces and analyzes the data sets used. The outliers and missing values of the data sets are eliminated and interpolated. Through the correlation analysis, it is found that the first principal component of the furnace condition parameter data is closely related to the steadiness of the furnace condition. Based on this, the mean value, standard deviation and maximum value of the first principal component are selected to represent the stationary characteristics of the furnace condition. Then, an unsupervised classification model based on mixed Gao Si and naive Bayes algorithm is constructed for judging the stability of furnace conditions. The algorithm can be trained without supervision. It solves the problem that the traditional model has no label or label is not trusted. In this paper, the performance of the model is compared with decision tree, logical regression and support vector machine. The model is superior to the contrast model in the index of synthetic error rate, first type error rate and second type error rate. Finally, based on the proposed model, an intelligent judgement system for blast furnace condition is designed and implemented, which can be used for visual analysis of blast furnace historical data, model training and real-time monitoring of field data.
【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】：TF325.6;TP18

【相似文献】