Performance Classification And Recognition Of Copper Flotation Process Based On Consensus Clustering

With the wide application of image processing technology in the field of copper flotation, production field can acquire flotation image visual features, monitor the online production status, identify the performance through analyzing the flotation process operation, and provide the basis for optimization control of the flotation process. However, the current flotation monitoring is vulnerable to subjective factors, which lead to information inaccuracy and lag seriously for online monitoring process, affect the adjustment of flotation conditions and bring about a huge waste of resources and energy.A variety of sensors in the flotation process accumulate a large amount of data, especially in the application of machine vision technology, which obtains real time characteristics of bubble. Clustering is an effective method for classifying data. By analyzing the correlation between classified visual characteristics and flotation production index, establish the performance classification and recognition model, which can find the production state accurately and provide guidance for the production operation. The research content and innovation points in the paper are as follows:(1) The improvement of consensus clustering. In view of problems of building similarity matrix in consensus clustering method, the paper mainly analyzes the influence of the Gaussian function, Minkowski distance, nearest-correlation method and local scale method on the clustering results. In order to solve the problems of determining cluster number, the paper proposes an improving consensus clustering algorithm based on Minkowski distance. The algorithm depicts the similarity between samples by means of Minkowski distance, then gets the different clustering results by adjusting the parameters of the distance formula, finally obtains the accurate information about the clustering number through the comparison among the clustering results.(2) The application of consensus clustering in the classification of the flotation performance. Firstly, cluster the offline data by means of the consensus clustering method according to characteristics of flotation data, then establish the model of all kinds of performance by means of least squares support vector machine(LSSVM) method on the basis of the relationship between clustering data and process index.(3) The application of consensus clustering in the identification of the flotation performance. According to characteristics of flotation time-series data, research the consensus clustering based on sliding window to classify the online data accurately, then realize the recognition of flotation performance by analyzing the estimation of the performance model.