【作者】 赵大伟；

【导师】 艾红；

【作者基本信息】 哈尔滨理工大学 ， 模式识别与智能系统， 2010， 硕士

【摘要】 球磨机由于具有容量大、运行可靠、煤种适应性强、维护简单、检修费用低等优点,成为目前国内燃煤电厂制粉系统中使用最多的磨煤设备。理论研究和实践运行均表明：球磨机制粉系统的最优运行与球磨机筒内料位密切相关。目前,实际生产中主要采用差压法、噪音法、功率法等传统方法来检测料位,但是球磨机是一个多变量、非线性、强耦合、大延迟的对象,各参数之间耦合严重,单独采集其中某个参数来反映料位均无法获得理想的效果。数据融合是一种新兴的信息处理技术,是组合多源信息完成目标检测、关联、状态评估的多层次、多方面的过程。其理论和方法已成为智能信息处理及控制的一个重要研究方向。论文首先深入研究了钢球磨煤机的运行特性,总结了现阶段钢球磨煤机料位检测的现状,比较了常用的料位检测方法的优缺点,通过多方论证,将数据融合技术应用到钢球磨煤机料位检测中。方案将球磨机的磨音信号、出入口差压、出入口温差、入口负压作为融合参量,选择了BP神经网络法,利用其强大的信息综合能力,知识泛化能力及结构的容错性等优势,在数据层用作融合算法,并进行了BP神经网络法的融合仿真。从仿真结果可以看出,采用数据融合技术的钢球磨煤机料位检测方案,能够更加精确地实现对钢球磨煤机料位的检测,从而使磨煤机始终工作在较为高效的工况,降低了球磨机的功耗,提高了其效率,为钢球磨煤机优化控制奠定了基础。更多还原

【Abstract】 Ball Mill has large capacity, reliable operation, adapts to many kinds of coal, repairs easily and inexpensively. It has become the most equipment in the domestic coal-fired power plant milling system. The academic study and practical run both show that the ball mill pulverizing system runs closely with the material level. At present, we mainly uses pressure difference between export and import, grinding sound signal, power and other traditional methods to detect the material level in the actual production, but the ball mill is a multivariable, nonlinear, serious coupling and large delay object, it has a number of parameters, and the coupling between parameters is serious. Only gathering one of the parameters to reflect the material level can not get perfect results.Data fusion is a new information processing technology, it was used in the military sphere first.It is a process to combine the multisource information for target detection, association, state evaluation. Its theories and methods have become in the intelligent information processing and control.The paper first studied the running characteristics of ball mill, summarized the actuality of ball mill material level detection, understood the advantages and disadvantages of the present methods of material level detection, applied the data fusion to the ball mill material level detection through argumentation. The project collected the ball mill grinding sound signals, pressure difference between import and export, temperature difference between import and export, import negative pressure as fusion parameters, selected the BP neural network method, used its powerful information comprehensive capabilities, ability of knowledge generalization, and the fault-tolerance of structure as a fusion algorithm in the data layer, and completed the fusion simulation of BP neural network. From the simulation results we can see that using data fusion technology can achieve more accurate detection of ball mill material level, and make the ball mill to consistently run in the more efficient operating condition, improve the efficiency of ball mill, reduced the ball mill power consumption, in order to establish the foundation of the optimized control of ball mill.更多还原