Status Identification Method And Application For Copper Removal Process Based On Trend Analysis

Copper removal process from zinc solution is important to the direct leaching of zinc smelting technology. The operators adjust zinc amout in production by considering the laboratory value of outlet copper iron concentration. However, the detection hysteresis and the subjectivity of human judgement cause large fluctuations in outlet copper iron concentration, so the process stability requirements can not be satisfied. This article studies the identification method of copper removal process based on process data, which can provide guidance for operation and significant for the performance stability.Combined with the characteristics of copper removal process, the important factors of the process were analyzed. The comprehensive status evaluation criteria was proposed, which contains both the value and trend information of the parameters. Then extract and identify the trends of parameters in copper removal process by using qualitative trend analysis. The cusum method was used to split linear fragments, and the fragments was classified into7shapes, finally they were conversed and merged into3trend primitives, so the compelete description of trend was achieved. The result of trend analysis is an essential part of status sample, which can make the classification more accurately and improve the status identification of copper removal process.Considering the uneven distribution, differences of feature importance and noise of status samples, a status identification method for copper removal process based on Fuzzy Support Vector Machine was proposed. The fuzzy membership was calculated by the distance between the center of class and the sample, and the feature weight of the sample was calculated by introducing the concept of information entropy, thus the feature weight matrix was obtained to improve the kernel function. Then the model of status identification was built, which can reflect the difference of feature importance and reduce the influence of unbalance sample and noise, so as to improve the accuracy of status identification. The validity and accuracy of the proposed mehod were proved by comparison results with standard experimental data sets and the industrial data.Finally, taking use of historical data of the production process, the status identification system was designed. Combined with the status identification model, the system architecture design, the implementation steps and the main modules were introduced, the systematic functions were implemented in sofrware. Figure22, table8, reference62.