The Application Research Of The Cerebellum Model Network In Alumina High Pressure Dissolution And Settlement Separation

Alumina production process is a complicated continuous chemical industrialproduction process. High pressure dissolution and settlement separation are two keyworking procedure of alumina production process. Caustic ratio is an important controlindex of the high pressure dissolution process.Dilute solid content is an important controlparameter of settlement separation process.They not only have important guiding role tothe alumina production, and reflect the product quality of alumina industry. At present,however, the detections of caustic ratio and dilute solid content was obtained by thedirectly calculation through the chemical analysis. The measured result existed the largelag and couldn’t reflect production conditions in the real time. Therefore, caustic ratio anddilute solid content soft sensor model based on the CMAC neural network wasproposesd,through CMAC network structuring the relationship model of secondaryvariables which can be measured and leading variables,indirectly estimating the value ofleading variable, realizing online real-time detection of caustic ratio and dilute solidcontent.The main work contents of paper are as follows:Firstly, according to the CMAC mapping concept algorithm putted forward by Albusexisting address uneven distribution defect, introduced the optimal offset vector algorithmbased on heuristic method, made address space distribution even more uniform andimproving the accuracy and generalization performance of CMAC network modeling;Secondly, according to traditional cerebellum model neural network’s studyalgorithm-Least Mean Square algorithm itself exists conflict between convergence speedand steady state performance, adopted variable step long adaptive LMS algorithm based onthe hyperbolic secant function as CMAC learning algorithm,gave consideration toconvergence rate and steady-state disorders two indexes, and improved the robustness andthe stability of CMAC network modeling;Thirdly, analysed process mechanism of the alumina high pressure dissolution andsettlement separation process, selected secondary variables which can be measured ofcaustic ratio and dilute solid content soft measurement model, and adopted the partial leastsquares algorithm to reduce the dimensional of secondary variables, simplified input spaceof caustic ratio and dilute solid content soft measurement model, reduced the complexity ofthe soft measurement model, improved convergence speed and precision of the model; Fourthly, designed and realized caustic ratio and dilute solid content soft sensorsystem based on the CMAC network, realizing online real-time detection of caustic ratioand dilute solid content, compared the performance of soft sensor system adoptingtraditional CMAC network with the performance of soft sensor system adopting improvedCMAC network, the results showed caustic ratio and dilute solid content predicted by softmeasurement system based the improved CMAC network was very close to actual producevalue, accuracy was higher, and prediction stability was better.Based on mechanism analysis of the alumina high pressure dissolution and settlementseparation process, researched application of caustic ratio and dilute solid content in highpressure dissolution and settlement separation process, adopted caustic ratio and dilutesolid content of real-time detection to guide high pressure dissolution and settlementseparation process, controlled alumina production conditions, and improved the benefit ofalumina industry.