螺旋分级过程数学模型研究及应用

Research and application for spiral classification process mathematic model

磨矿分级过程具有多变量、非线性等特点,基于多元回归理论的数学模型难以满足精度要求,迅速发展的神经网络技术可以建立高精度的变量间的非线性映射模型。在已有螺旋分级机基本模型的基础上,利用RBF神经网络建立螺旋分级机的数学模型,并用遗传算法对神经网络进行优化。用某选矿厂两段磨矿分级回路的实际生产数据进行了仿真实验,仿真结果表明,模型精度满足工艺要求。

The process of grinding circuit is multivariable and non-linear,the mathematic models based on the linear regression are hard to satisfy the requesting of accuracy.The fast developing neural networks technique can build a high precision model for the non-linear reflection of variables.Based on the elementary mathematic models about spiral classifier machine,this paper establishes the mathematic model of the spiral classifier by a Radial Base Function （RBF） neural network.Genetic algorithm is used to optimize the RBF.With the actual process data from a grinding circuit of a mineral factory, the result of simulation shows that the accuracy satisfies the craft request.