基于减法聚类ANFIS的链-回-环球团热状态控制模型

 链箅机-回转窑氧化球团热工制度是影响球团矿产质量指标及生产能耗的关键因素，但由于其工艺特点，热工过程状态参数和操作参数多、设备的耦合性强、操作变量与被控变量关系复杂，难以采用精确的机理模型进行控制。在分析链箅机-回转窑氧化球团热工过程物料流及热风流特点的基础上，建立了基于自适应神经模糊推理系统（ANFIS）的热状态控制模型。采用减法聚类划分模型输入空间，采用最小二乘法及梯度下降法对T-S模型进行辨识。结合VC++和MATLAB混合编程的方法，开发了链箅机-回转窑氧化球团热状态控制指导系统，实现了模型的在线计算，以及操作变量的实时控制指导。采用国内某球团厂的生产数据对模型进行了仿真，模型的平均相对误差均小于5%。

Thermal state control model of grate-kiln pellet based on substractive clustering ANFIS

The thermal operations of pellet induration have great effects on energy consumption， productivity and pellet quality in grate-kiln process. Due to the process characteristics， such as multiple variables， strong coupling and the complicated relations between manipulated and controlled variables， the thermal process can hardly be effectively controlled by mechanism model. Material flow and thermal airflow of grate-kiln process were firstly analyzed， and then the subtractive clustering based ANFIS (adaptive neural fuzzy inference system) model was proposed to control the thermal state. In this control model， subtractive clustering algorithm was adopted to partition the input data space， while recursive least square method and gradient descent method were used to identify both premise parameters and conclusion parameters of the T-S model. Using a hybrid of VC++ and MATLAB， control system of grate-kiln pellet thermal state was developed， which realizes the online model calculation and the real-time control guidance. Model validation was conducted using the production data of a particular pelletizing plant， and the results show that the mean relative error of the model is less than 5%.