基于支持向量机的水泥生料粉磨细度软测量方法研究

发布时间：2018-01-28 22:44

本文关键词： 生料细度软测量延时最小二乘支持向量机　出处：《燕山大学》2016年硕士论文　论文类型：学位论文

【摘要】：在水泥生产过程中,生料经粉磨形成具有一定细度要求的粉末对水泥质量有巨大影响。生料细度的检测是生料粉磨系统优化控制的重要环节,国内通常采用人工离线化验的方式得到生料细度筛余百分比,但其检测周期一般为两到四小时;国外通常采用在线分析仪进行生料细度的测量,但其价格昂贵且不易维护。针对生料细度人工检测周期时间长、工作效率低、在线分析仪价格昂贵、不易维护等导致生料细度出现合格率偏低及检测效率和经济效益不能兼顾等问题。本文提出了一种快速有效的生料细度软测量方法,设计并实现了生料细度在线软测量系统,有效的解决了人工检测效率低、在线分析仪价格昂贵等问题。具体研究如下:首先,本文总结了现有软测量建模方法,并深入研究了生料粉磨系统工艺和生料磨工作原理。通过分析生料细度的影响因素确定生料细度软测量系统的辅助变量,提出基于支持向量机和数据驱动的建模方法,实现对生料细度的软测量。其次,针对生料细度软测量建模过程中由于存在延时导致建模失效的问题,构建系统延时(T)与最小二乘支持向量回归(LSSVR)相结合的软测量模型:将系统延时划分为“静态响应延时”和“动态响应延时”;借助于归一化互相关与一阶广义差分手段,得到“静态响应延时”和“动态响应延时”两参数的延时估计数学模型;并利用最小二乘支持向量机对经延时系统处理后的样本建立预测模型。基于此T-LSSVR模型对生料细度进行软测量建模实验,验证了本文所提方法对生料细度软测量建模的有效性及可靠性。最后,设计并实现了生料细度在线软测量系统。通过客户端软测量系统与服务器数据库间的数据交互,完成对生料细度的在线软测量,从而实现了在生料出磨前对生料细度的预测。该系统满足了基本的工业需求,节省了人工与经济成本、提高了工作效率,为生料细度的优化控制创造了条件。
[Abstract]:In the cement production process, raw materials grinding with certain requirements of the fineness of the powder have great influence on the quality of the cement raw meal fineness. Detection is an important step in the optimization control of raw material grinding system, the domestic commonly used artificial way to get offline laboratory raw meal fineness sieve residue percentage, but the detection period is generally two to four hours; measurement foreign commonly used online analyzer of granularity of raw material, but the price is expensive and difficult to maintain. The granularity of raw material manual detection time long cycle, low efficiency, on-line analyzer is expensive and not easy to maintain and lead to low detection efficiency and economic benefits can not take into account issues such as qualified rate granularity of raw material. This paper proposes a raw meal fineness soft measurement method is fast and effective, the design and implementation of the granularity of raw material online soft measurement system, effectively solves the problem of low efficiency of manual detection, The problem of online analyzer is very expensive. The specific studies are as follows: firstly, the paper summarizes the existing soft measurement modeling, and in-depth study of the raw material grinding system technology and raw mill working principle. Through the analysis of factors affecting the granularity of raw material auxiliary variable soft measurement system to determine the granularity of raw material, the support vector machine and modeling method based on data driven to realize the soft measurement of the fineness of raw material,. Secondly, the soft measurement modeling granularity of raw material process due to the existence of delay caused failure of the construction of system modeling, time delay (T) and least squares support vector regression (LSSVR) soft measurement model: combining the delay system is divided into "static response delay" and "dynamic response delay"; by normalized cross correlation and generalized difference method, get the "static response delay" and "dynamic response delay" two parameters of time delay estimation Mathematical model; and the use of least squares support vector machines for the delay system treated sample prediction model is established. The T-LSSVR model of soft sensor modeling based on experiments on the granularity of raw material, verify the proposed method for soft measurement modeling granularity of raw material is effective and reliable. Finally, the design and implementation of online soft measurement system of raw meal fineness through the soft measurement system of the client and the server database data exchange, complete the online soft measurement of granularity of raw material, so as to realize the prediction of raw material fineness of raw material in the mill. The system meets the demand of the industrial base, saves manpower and economic cost, improve work efficiency, creating conditions for optimization to control the granularity of raw material.

【学位授予单位】：燕山大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TQ172.66

【相似文献】