磨矿分级作业是选矿生产过程中至关重要的环节,磨矿粒度的好坏直接影响到浮选的精矿品位和回收率;通过分析实际磨矿过程的生产状况和基本的生产数据,磨矿粒度存在在线检测成本高、滞后时间长、实现困难等问题;在分析RBF神经网络结构特点的基础上,提出用RBF网络建立磨矿粒度预测模型,网络中心的选取采用可以在线学习的最近邻聚类算法;仿真结果表明,该网络非线性处理能力和逼近能力强,学习时间短,网络运算速度快,模型精度满足工艺要求。
The particle size of grinding circuit is the important factor to the grade of concentrated ore and metal recovery rate. It is not only hard to be detected on-line, but also has large time-delay and measurement cost. On the basis of analyzing the characteristic of RBF neural network structure, the RBF network is used to establish prediction model for particle size of grinding circuit. Simulation results show that the capability of the model can approach function and treat nonlinear problem by using nearest neighbor-clustering algorithm to select the clustering center. Besides the learning time is short and the prediction accuracy satisfies the technical requirement.