论文部分内容阅读
以典型的两段磨矿回路为研究对象,针对磨矿粒度在线检测困难而难以满足实时控制的难题,提出一种新的软测量建模方法。应用最小二乘支持向量机(LSSVM)进行磨矿粒度软测量建模,为解决参数设置的盲目性,利用改进的变步长果蝇优化算法(FOA)较强的寻优能力对LSSVM的惩罚系数和核参数进行优化。对该模型进行预测仿真,同时与网格搜索法、粒子群法和未改进的果蝇算法优化的LSSVM模型进行对比实验。结果表明,相对于其他模型,改进的FOA-LSSVM收敛速度快,预测精度最高,较好地实现了对磨矿粒度的实时检测。
Taking a typical two-stage grinding circuit as the research object, aiming at the difficulty of on-line detection of grinding particle size, it is difficult to meet the real-time control problem. A new soft-sensing modeling method is proposed. In order to solve the problem of parameter setting blindness, LSSVM was used to model the particle size of soft sensor. To improve the blindness of parameter setting, LSSVM was punished by the improved ability of optimizing FOB (Frosted Optimization) algorithm (FOA) The coefficients and kernel parameters are optimized. The model was predicted and compared with the LSSVM model optimized by grid search, particle swarm optimization and unmodified Drosophila algorithm. The results show that compared with other models, the improved FOA-LSSVM has the advantages of fast convergence and highest prediction accuracy, and realizes the real-time detection of the grinding particle size.