Optimization Control Of The Zinc Roasting Process Based On Multiple Working Condition Identification

Zinc roasting process is the pre-process of the zinc hydrometallurgy process.Two large-scale roasters are operated in parallel to convert zinc concentrate into zinc calcine and sulfur dioxide gas,providing raw materials for downstream leaching and acid-making processes.Due to fluctuations in feed composition,complex reaction mechanisms and offline detection of specific compositions of zinc calcine,the zinc roasting process is subject to large fluctuations in working conditions,making it difficult to optimize and stabilizing control of roasting temperature.Additionally,the coupling between the two roasters and downstream leaching and acid-making processes further exacerbates the difficulties in optimization control of the roasting process.To address these issues,this thesis conducts in-depth research on the production process and reaction mechanisms of the roasting process and proposes a optimization control method for the zinc roasting process based on multiple working conditions identification.A multi-mode estimation model for the soluble zinc rate is established.On this basis,the operating performance evaluation as well as collaborative optimization and stabilizing control of the roasting temperature are realized.The main innovations of this thesis are as follows:(1)A feature-fusion-based working conditions identification method is proposed: An abnormal working condition identification model based on global-local slow feature analysis is established by studying the characteristics of abnormal working conditions in the roasting process.The global and local slow features are extracted from macroscopic and microscopic perspectives,and a one-dimensional convolutional neural network is used to extract deep features for modeling,achieving accurate identification of abnormal working conditions under the presence of extreme samples and working condition switching.A working condition classification method is proposed based on multi-feature fusion by analyzing key process parameters of the roasting process.Shallow features with physical meanings and deep features with hidden information are extracted based on feature engineering and deep learning.The feature set is clustered using Gaussian mixture models,achieving the multiple working conditions classification of the roaster and laying the foundation for multi-mode modeling and operating performance evaluation.(2)A mechanism-guided data-driven online estimation method of soluble zinc rate is proposed: A mechanism model based on the unreacted shrinking core model and two-phase fluidization model is established by analyzing the reaction kinetics and hydrodynamics of the zinc roaster.The model parameters are identified under different working conditions to obtain the multi-mode mechanism model.On this basis,a mechanismguided data-driven modeling method is proposed.The estimated value of the mechanism model is used as labels to guide the training of the datadriven model,solving the problems of insufficient accuracy of the mechanism model and difficulty in training data-driven methods under multiple sampling rates,and enabling online estimation of soluble zinc rate.(3)An operating performance evaluation and temperature collaborative setting method under coupling indicators is proposed: Several key performance indicators affecting the operating performance are designed by analyzing the coupling relationship between the roasting process and the downstream processes.On this basis,a fuzzy synthetic evaluation model is established.To improve the evaluation results,a variableweight degradation degree is proposed to adjust the importance of indicators.Based on the evaluation results and sensitivity analysis of the soluble zinc rate estimation model,a temperature collaborative setting method based on temperature adjustable margin is proposed.The temperature setting values are reasonably allocated to two roasters based on their operating states,achieving optimized operation of the roasting process.(4)A trend-based event-triggering fuzzy control method for roasting temperature is proposed: A temperature fuzzy controller is established through an in-depth analysis of the operating conditions and process technology of the roasting process.The temperature trend extracted by the qualitative trend analysis is used as the input of the fuzzy controller instead of the temperature gradient,which improves the accuracy of the temperature state evaluation.An event-triggering strategy based on the temperature trend is designed and combined with the fuzzy controller to achieve temperature stabilizing control of the roaster under uncertainty and large time delay.Based on the above researches,a large-scale roaster operation optimization system is developed,which has the following functions: process monitoring,curve display,fault alarm,and optimization control.The system can run stably for a long time in the industrial field,cope with routine field conditions,and ensure temperature stability.It effectively improves product quality and ensures the long-term operation of zinc roasters.