Abstract: The particle size of grinding circuit is an important factor to the grade of concentrated ore and metal recovery rate.A Support Vector Machine（SVM） model is developed to deal with the problem that particle size cannot be measured in time and efficiently.Compared with SVM existed，a mixture kernels SVM model is used，at the same time，to resolve the problem how to optimize the parameters of the mixture kernels SVM model efficiently，Genetic Arithmetic（GA） is proposed.Simulation results show that the model of particle size of grinding circuit based on mixture kernels SVM which is more excellent than the model based on RBF kernels SVM，is characterized by better approximation，finer generalization and higher accuracy.

摘要： 选矿厂磨矿粒度是影响精矿品位和回收率的重要因素。针对目前无法对磨矿粒度进行实时有效检测问题，提出了一种基于支持向量机的磨矿粒度预测模型。通过对现有支持向量机建模方法分析比较，选择了新型的混合核支持向量机作为预测模型的建模工具，同时为了解决有效选择混合核参数问题，提出利用遗传算法对模型结构参数进行优化。仿真结果表明，用该方法建立的磨矿粒度预测模型优于基于RBF核支持向量机建立的该预测模型，其具有较好的逼近性能和泛化性能及更高的预测精度。