Modeling And Optimization Control Methods For Synthesis Process In Hydrometallurgy And Its Applications

With the resources depletion, environmental protection and metallurgical materials demand for some high technology, hydrometallurgy plays an important role in preparation, separation and purification of metals and their compounds, recovery of valuable metals in raw materials and prevention of environmental pollution. Therefore, hydrometallurgy has been developed rapidly and applied widely. Synthesis process is an important unit operation for separation and purification and preparation of metal compound in hydrometallurgy. The remarkable advantages of synthesis process are easy to operate, low cost and small investment. Although synthesis technic in hydrometallurgy of our country meets advanced world standards, the control of synthesis production process still remains off-line analysis, experience adjustments and manual control, which lead to low efficiency, high cost, bad product quality and high resourses comsumption. Therefore, reduction of raw material consumption and production cost, and improvement of economic efficiency have become serious problems that need to be solved urgently in hydrometallurgy industrial development. The online measurement and optimization control of key variables for hydrometallurgy are the foundation to solve these problems.This paper aims at the implement of on-line measursing product particle size and optimization control in hydrometallurgy synthesis process. Based on deeply analyzing the characteristics of synthesis production process, hybrid modeling method which is composed of mechanism and data modeling methods is used to predict the product particle size, and then an integrated control strategy is developed to improve product quality of synthesis process. The main researches are summarized as follows:1. Through analyzing the basic principle of synthesis and based on the relationship of material balance and population balance equation, the dynamic first principle model of synthesis was established. By model simulation, the dyncamic feature of the synthesis process was revealed to find the major affected factors of the process, to identify the secondary variable of the soft sensor model, and to establish the foundation of soft sensor model.2. Aim at the difficulties of applying the dynamic first principle model to the industrial field directly and process data usually with outliers, a robust soft sensor model of synthesis process was established adopting parallel hybrid model. The model was composed of the simplified first principle model and robust compensated model in parallel, therefore the advantages of different modeling methods can be exerted. The purpose of simplification was to reduce the unmeasured variables in the first principle model, and improve the computational efficiency of the model. The compensated model, which employed the process data, was used to remedy the problem of decline in prediction accuracy brought about by model simplifying. For building robust compensated model, a robust LSSVR approach based on robust learning algorithm was presented. First, the LSSVR model was used to predict process outputs and residuals were obtained between real outputs and predicted outputs. Then robust learning algorithm was carried out to train the weights of LSSVR iteratively. Therefore, a robust LSSVR model was built. By simulation experiment, the efficiency of the hybrid modeling method was verified.3. Aim at the difficulty of applying the soft sensor model constructed based on the offline history data to adapt the time variant characteristic of the process, an intelligent correction approach of soft sensor model based on model performance assessment was proposed. Firstly, in order to evaluate the performance of the soft sensor model, an error Gaussian mixture model (GMM) was developed to describe the probabilistic characterization of model prediction error. From the error GMM, error variance can be obtained to assess the reliability of model output prediction, then the assessed model performance results which was used to determine whether the model needs to be correted or not were given. Finally, the model correction strategy was carried out using either the short-term or long-term correction, depending on the assessed model performance results. Model performance assessment method can avoid the soft sensor model being corrected blindly by the traditional correction method. Employing the proposed robust hybrid method and correction strategy to predict the industrial data of cobalt oxalate synthesis process, satisfied results were obtained.4. Aim at the difficulty of optimization control for synthesis process in hydrometallurgy, an integrated control strategy based on robust hybrid model was proposed. It combined batch-to-batch control and midcourse correction (MCC). On the one hand, batch-to-batch control strategy was used to exploit the repetition nature of synthesis process and update the manipulated variable trajectory for new batch run using information from previous batch runs. It can overcome the model plant mismatches and unmeasured disturbances. On the other hand, MCC within a batch calculated the future manipulated variable values using all past information up to the current point in time and taked control action to reduce the effects of disturbances and bring the off-spec product quality back to the target. The integrated control strategy can combine the advantages of both methods to enhance the optimal operation of synthesis process. By simulation experiment, the efficiency of the control strategy was verified.5. By combining the proposed soft sensor modeling and integrated control approaches, optimal operating system software of synthesis section was designed and developed. Soft sensoring for product particle size of synthesis process was achieved, and the optimal operating guide for this process was provided. The results of applying the software to a synthesis section in a cobalt hydrometallurgy plant verified its efficiency.