论文部分内容阅读
精矿品位是直接关系到选矿生产效率和产品质量的重要指标,精矿品位的在线检测对提高赤铁矿选矿生产流程的产品质量和生产效率、减少资源消耗具有重要意义。但是,由于选矿生产的强非线性和强耦合性,使得精矿品位往往难于在线连续测量。为了建立精矿品位与工艺指标之间的模型,以便于调整各工艺指标,从而实现全流程优化,本文采用多层感知器神经网络和系统辨识相结合的方法,运用Matlab系统辨识工具箱和神经网络工具箱,提出了由线性模型和非线性补偿模型组成的混合模型结构的精矿品位预报方法,建立了精矿品位预报模型。该方法提高了对于复杂工业过程中精矿品位指标的预报精度,通过现场数据和一系列对比实验验证了本方法的有效性和优点。
Concentrate grade is directly related to the efficiency of mineral processing and product quality an important indicator of on-line detection of concentrate grade hematite beneficiation process to improve product quality and production efficiency and reduce resource consumption is of great significance. However, due to the strong non-linearity and strong coupling of beneficiation production, it is often difficult to continuously measure the concentrate grade on-line. In order to establish a model between concentrate grade and process index in order to adjust each process index so as to optimize the whole process, this paper adopts a combination of multi-layer perceptron neural network and system identification, and uses Matlab system to identify the toolbox and nerve Network toolbox, the prediction method of concentrate grade based on mixed model structure consisting of linear model and non-linear compensation model is put forward, and the forecast model of concentrate grade is established. The method enhances the prediction precision of concentrate grade index in complex industrial processes. The effectiveness and advantages of this method are verified by field data and a series of comparative experiments.