Data-driven Modeling And Condition Analysis Methods For Copper Flotation Process Based On Complex Networks

Copper flotation is an important process of copper production. Because of the numerous parameters and varied influences on flotation status, it is hard to control the real-time parameters, thus affecting the flotation. As the mass flotation sample data has been accumulated from the field, it would be beneficial to dig up some useful information for production feature analysis, which would be importance in promoting the production efficiency and resource distribution of factory.Based on the complex network, the network data theory can dig up the important information and the mapping from mass data in topology. Network topology, known as a method to reveal the knowledge, will be helpful to explore the potential relationship of data in the flotation process. By building up the flotation network model based on data and analyzing the topology characteristics of the networks of different production status, we will be able to develop the guideline and reference to the adjustment of copper flotation production status. Therefore, the research and initial concepts of this thesis are listed below:(1) As for there exists numerous production parameters which are repeated coupling, if all the parameters are directly used to measure the similarity between samples, the computational complexity is large, considering the parameters’influences on flotation conditions, the author build up the parameter related network to identify the sensitive parameters, instead of using all the relations among parameter samples which requires numerous calculations. It requires several steps to build up the accurate network model:First of all, the Pauta criterion and normalized method are applied for data pre-process; Secondly, the parameter related network is built by the time sequence similarity measure and the minimum spanning tree method, the parameters are defined as the nodes and their relations as linkage; and then, identify the sensitive parameters from the evaluation methods based on the indexes of weighted network aggregation degree and node importance. (2) It requires an accurate measure of the similarity between samples to identify the production condition. It comes up with the concept of the decisive parameters assemblies which is used to measure the similarity between samples, and builds the association network of the decisive parameters assemblies by selecting the proper threshold value to construct the sensitive parameters assemblies’network. Based on the K-means in similarity gathering method to explore the sub-assembly of the network, which helps to identify the various flotation status, thus reveling the correlation and characteristics in copper flotation. The method mentioned above will develop a brand new way to analyze the mass parameters and various status in copper flotation.18figures,7frames and72references are listed in this thesis.