Optimization Under Uncertainty For The Blending Process Of Alumina Production And Its Application

The raw material blending(RMB) is the first process of alumina production by sintering method. The quality of the slurry produced by the RMB process makes direct influence on the product quality and energy consumption of the downstream processes. However, the RMB is a complex process with strong coupling, unknown chemical reactions and complex condition changes. Especially, the composition fluctuation of raw materials and the difficulty of on-line measurement result in the uncertainty in process parameters, which makes it very difficult for human operation and traditional optimization technologies to effectively realize the optimal control of the blending process.Considering the uncertainty in raw material composition parameters, the optimization strategy under uncertainty is proposed to enhance the robustness of the optimization system for the blending process of alumina production. And the industrial application results prove the effectiveness of the proposed scheme. The major innovation research achievements include:(1)After investigating the RMB process, the optimization strategy under uncertainty, including three modules of the mathematical description of uncertain parameters, the structure of optimization model with uncertain parameters and the solution algorithm, is proposed for the blending process. The optimization strategy firstly obtains the confidence interval of uncertain parameters on the basis that the blended materials are divided into two kinds of raw materials and returned materials. And the interval uncertainty degree and the interval cover rate are defined to evaluate the confidence intervals. Then, according to the production requirements of slurry quality and the time-delay of off-line measurement, the structure of the optimization model with uncertain parameters is researched for the RMB process. And, the classificatory knowledge base and the expert reasoning strategy, which match with the structure of the optimization model, are designed to effectively solve the optimization problem of the RMB process.(2)For the returned material with the large composition fluctuation and the strong nonlinearity, the confidence intervals of the compositions are predicted based on the generalized phase space reconstruction technology. Firstly, the correlation dimension and the largest Lyapunov index are adopted to analyze chaotic characteristics of time series of compositions of returned material. Then, by using rough set(RS) theory, the generalized phase space of multi-component time series of returned material is reconstructed to extract training samples and LS_SVM was used to describe the relationship between input and output variables. Finally, the confidence intervals of composition parameters of returned material are obtained by compensating certain deviations for the prediction results of LS_SVM. For raw materials with small fluctuation of compositions, the statistical method is used to determine the confidence intervals of uncertainty parameters, and the rolling update strategy is introduced to on-line optimize the interval. Data verification results show the effectiveness of parameter intervals.(3)Based on the uncertainty, time-delay, coupling and interval requirements, the lexicographic interval goal optimization model with uncertain parameters, which includes the prediction model for slurry quality, is built for the RMB process of alumina production. In the prediction model, a mechanistic model based on material balance principle is established as the master-rule model, and an intelligent compensation model combining neural networks(NNs) with empirical knowledge is integrated with the mechanistic model to enhance the prediction precision. Simulation experiments show that the prediction results of integrated model are better than that of single mechanism model as well as that of NN model. So, the integrated model can deal with the large time-delay of off-line measurement for slurry quality. Based on the quality prediction model, the lexicographic interval goal optimization model with uncertain parameters is built, the objective of which is to minimize the violation to the intervals of quality indexes. This model can accurately describe the optimization problem of alumina blending process due to taking into account uncertain parameters, strong coupling and interval requirements of quality indexes.(4)Considering the structure of the optimization model, the lexicographic reasoning strategy based on the hammersley sequence sampling(HSS) technology and classificatory knowledge base is designed to solve the optimization problem with multiple objectives, interval requirements and uncertain parameters. In the design of knowledge rules, the knowledge rules of blending process are classified into different groups according to the first optimization objective, and sorted by precedence in every group. This kind of classificatory knowledge base with precedence matches the optimization model and the reasoning strategy, and the effective solution is realized.(5) On the basis of the researches amentioned above, the optimization system under uncertainty is developed for the RMB process of alumina production. In the sytem, the OPC is used to realize the real-time communication between the optimization computer and the DCS, and the analyzed data management system is connected to the optimization system by the intranet to acquire automatically the analyzed data from the offline laboratory. And the optimization system realizes the functions including process monitoring, ratio optimization, data leading-in, data management and so on. The results of industrial application show that the proposed strategy can reduce the burden of operators, enhance the robustness of the system, improve the quality of slurry and stablize the whole alumina production. It provides a good optimization mode for other blending processes with uncertainty.