Soft Sensing Modeling Study Of Alumina Concentration In The Aluminum Electrolysis Process Based On LSSVM

Alumina concentration is an extremely important parameterin the process of aluminum electrolytic, which is an important basis of blanking intervals for aluminum cell,and it affects the material balance of aluminum cell.Due to the high temperature, strong magnetic field, corrosive and other features of aluminum cell,the alumina concentration is very hard to be testedonline,which seriouslyinfluenced the control of aluminum electrolytic process.To this end, this paper based on LSSVM aluminum electrolytic process relationship model between alumina concentration and other parameters to achieve a soft alumina concentration measurements.This paper firstly analyzed the process mechanism of aluminum electrolytic and the main factors influencing the alumina concentration.For the characteristics ofblanking alumina-dissolving-consumption process,the paper emphatically discussed the relationship betweenalumina concentration and groove resistance and the slope of the resistance. Forthe hard problem of expressing the state of the electrolysis bath,we usedwavelet packetto decompose and reconstruct the bath voltage signalto extract the characteristics of voltage signal, then effectively describedthe state ofthe electrolysis. bathwith the extracted signal.Analyzed the characteristics of electrolytic process data,on the basis of the data preprocessing, selected the auxiliary variables of aluminum soft measurement model.For the hyper-parameter values of LSSVM soft measurement model would affect itsaccuracy and generalization ability, in this paper, usd artificial colony selection algorithm to optimize the hyper-parameter of LSSVM.Because of artificial bee colony algorithm easily fall into local optimum, this paper used chaotic search strategy to replace the artificial colony algorithm gaining new food source, the method of the artificial colony algorithm was improved,simulation results showed the effectiveness of the improved algorithm. Improved artificial colony algorithm was used to optimize hyper-parameters of LSSVM,then established alumina concentration of soft measurement model.Industrial field operating data demonstrated that the soft measurement model established by this paper has higher precision and generalization ability, which laid a foundation for the control of the process of alumina.