基于子集融合与规则约简的磨矿过程模糊建模

发布时间：2018-03-15 12:02

本文选题：磨矿分级　切入点：Takagi-Sugeno模型　出处：《大连理工大学》2015年硕士论文　论文类型：学位论文

【摘要】：磨矿工业过程复杂,生产线上的控制变量较多,变量间的强非线性动态关系致使磨矿模型较难用精确的数学模型来描述,一般的控制方法难以对其进行控制,多数磨矿过程的控制一般需要操作员依据自身经验对控制变量进行实时调控,但由于操作员的主观经验限制、工况的复杂和边界条件的多变性,人工控制往往难以达到预期的生产指标。对磨矿过程的有效建模,有助于磨矿过程自动化生产的实现,可以避免人工控制的主观性带来的误操作,提高矿产资源的利用率,降低选矿厂的生产成本,提高产量。以选矿工业过程中的磨矿系统这一复杂过程为背景,针对磨矿系统建模问题,本文提出了一种基于子集融合与规则简约的模糊建模方法。首先根据工业数据,采用一种数据驱动的方法建立初始的磨矿过程Takagi-Sugeno模型,然后针对该模型中隶属度函数过拟合问题,提出一种基于模糊C均值聚类(Fuzzy C-means Clustering, FCM)的方法,对初始规则库中同一变量下的隶属度函数参数进行聚类,得到对不同工况具有代表性的融合后的隶属度函数,来降低建模过程过拟合的影响。最后,针对模糊集融合后规则库中出现的规则冗余问题,本文定义了冗余规则相似度,并根据该相似度,对前件相同的冗余规则进行约简,消除冗余,形成最终的泛化能力较强的离线模糊规则库。为验证本文方法的有效性,以另外几种模糊建模方法作为对比,分别采用经典数据与国内某选矿厂的实际工业数据进行实验验证。两组实验分别从模型的精度和结构方面对多种建模方法进行了对比,结果表明,本文方法较其他几种方法在精度和结构方面均体现出了一定程度的优势。最后,以本文建模方法为后台算法,开发了磨矿智能控制系统,对实际磨矿工业过程实现自动化控制。
[Abstract]:The process of grinding industry is complex, the control variables in production line are many, and the strong nonlinear dynamic relationship between variables makes it difficult to describe the grinding model by precise mathematical model, and it is difficult to control it by general control methods. The control of most grinding processes usually requires the operator to adjust the control variables in real time according to his own experience. However, due to the limitation of the operator's subjective experience, the operating conditions are complex and the boundary conditions vary. The effective modeling of grinding process is helpful to the realization of automatic production of grinding process, which can avoid the misoperation caused by subjectivity of manual control and improve the utilization ratio of mineral resources. Based on the complex process of grinding system in the process of dressing industry, aiming at the modeling problem of grinding system, the production cost of concentrator is reduced and the output is increased. In this paper, a fuzzy modeling method based on subset fusion and rule reduction is proposed. Firstly, according to industrial data, an initial Takagi-Sugeno model of grinding process is established by using a data-driven method. Then, a fuzzy C-means clustering method based on fuzzy C-means clustering (FCM) is proposed for the over-fitting of membership function in this model. The parameters of membership function under the same variable in the initial rule base are clustered. In order to reduce the influence of over-fitting in modeling process, the membership function after fusion is obtained for different working conditions. Finally, the similarity of redundant rules is defined in this paper, aiming at the problem of rule redundancy in the rule base after fuzzy set fusion. According to the similarity, the same redundancy rules are reduced to eliminate redundancy and form an off-line fuzzy rule base with strong generalization ability. In order to verify the effectiveness of this method, several other fuzzy modeling methods are compared. The classical data and the actual industrial data of a domestic concentrator are used to verify the experiment. The two groups of experiments are compared with each other in terms of model precision and structure. The results show that, Compared with other methods, this method has some advantages in precision and structure. Finally, an intelligent grinding control system is developed based on the method of modeling in this paper, which can realize the automatic control of the actual grinding industry process.
【学位授予单位】：大连理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TD921.4;TP18

【参考文献】