Dynamic Modeling For Zinc Roaster Under The Change Of Raw Material Composition And Its Application In Process Monitoring And Control

As a large-scale equipment in zinc smelting production,the efficient and stable operation of the roaster is an important guarantee for improving the products quality and production efficiency of zinc smelting.However,the complex physical and chemical reactions and self generated heat release mechanism of the roaster result in a significant impact of the raw material composition on its operating state.Establishing a dynamic model of the roaster under the condition of frequent changes in raw material composition and using it to guide production is of great significance for ensuring the products quality and improving production efficiency.Therefore,this thesis takes "data support-dynamic modeling-model application-system development" as the research logic,proposes a real-time estimation method for the soluble zinc rate of the roaster outlet product component information,and establishes a dynamic model of the roaster.Based on this,an anomaly monitoring method and a stable control strategy for quality indicators are designed,and an intelligent monitoring system are developed to ensure the efficient and stable operation of the roaster.The main research work and innovations of this thesis are as follows:(1)To solve the problem of complex distribution of process variable data and abnormal laboratory data in the roaster,a robust estimation model for soluble zinc rate based on variational autoencoder is proposed.Firstly,process variables are reconstructed using variational autoencoder,Gaussian distribution constraints are applied to hidden layer features using KL divergence,and parameter gradient updates and feature extraction are achieved using reparameterization trick.Process variables with complex distributions are mapped to the Gaussian distribution feature space.On this basis,kernel density estimation is used to characterize the training error of batch samples,and a parameter update method based on the optimization of the estimation error distribution is proposed,which reduces the sensitivity of the model to outlier and realizes real-time robust estimation of soluble zinc rate.The proposed method is verified by using industrial data of the roaster,and the results show that the real-time estimation accuracy of soluble zinc rate meets the actual production requirements,providing high-quality data support for subsequent dynamic modeling research.(2)To solve the problem of dynamic changes in the operating conditions of the roaster caused by batch changes of incoming raw materials,a dynamic modeling method integrating the thermodynamics of variable mass systems and decomposition-ensemble of for the roaster is proposed.Firstly,based on the thermodynamic theory of variable mass systems,the coupling relationship between mass changes of various substances and energy changes in the furnace is studied,and a first principle model of the roaster is established;Then,in response to the issue of the impact of changes in raw material components on model accuracy,a dynamic parameter selection and update strategy is proposed.Based on real-time process data and soluble zinc rate estimation data,the parameters of first principle model can be updated online,and a dynamic model library for different raw material components is constructed.In addition,a data-driven compensation model based on decomposition-ensemble is proposed to address the assumptions in the first principle model and the errors caused by the uncertainty of the roasting production process,and incremental updates are made based on real-time data to achieve dynamic compensation for temperature errors.The accuracy of the proposed dynamic modeling method is verified by steady operating conditions experiments under typical raw material components and dynamic operating conditions experiments under batch changes of raw material components.The model can be used to analyze the relationship between raw material components,control parameters,and the operating conditions,providing guidance for the roaster.(3)To solve the problem of frequent changes in the composition of incoming raw materials affecting the quality indicators of the roaster,based on the guidance of dynamic model,an anomaly monitoring method based on knowledge sharing deep generation network and a stable control strategy under temperature setting optimization are proposed.On the one hand,based on Bayesian derivation of conditional logarithmic likelihood,a knowledge sharing deep generation network model is designed.KL divergence is used to realize knowledge sharing of input branches,fusion features of process variables and quality variables are extracted,quality-related information and quality-unrelated information are separated by separating hidden layer features,and real-time quality estimation data is obtained by dynamic model to construct detection indicators.The detection and recognition of quality-related and quality-unrelated anomalies are realized.On the other hand,a stable control strategy under temperature setting optimization for the roaster is proposed in view of the influence of the change of raw material composition.When the raw material batch changes,the model is dynamically retrieved by the similarity of raw material composition.The optimal temperature set value and feed rate under the current raw material composition are obtained based on the optimization of the retrieved model.Then,a fuzzy inference method based on temperature change trends and deviations is proposed to achieve feedback regulation of feed rate,ensuring the stable operation of the roaster under optimal quality,and reducing the influence of raw material composition changes on the operating state and production efficiency of the roaster.(4)On the basis of the above theoretical research,based on the process characteristics and actual needs of the roaster on-site,an intelligent monitoring system for zinc roaster has been developed,achieving functions such as real-time estimation of soluble zinc rate,online monitoring of operating status,optimization of temperature setting values,control of feed rate and historical data query.The practical application of the intelligent monitoring system has effectively improved the product quality and production efficiency of the roaster,providing support for the efficient and stable operation of the zinc smelting production process.