| 摘要 |
FIELD: non-ferrous metallurgy, particularly electrolytical production of aluminium. SUBSTANCE: method comprises steps of measuring current values of electric current I and voltage Uel of cell; calculating with use of measured values reduced voltage Vr and concentration of alumina Cal in electrolyte; comparing current values Ur and Cal with predetermined ones and sustaining Ur and Cal in preset range by moving anode and controlling quantity of alumina charged to cell; additionally correcting current values Ur and Cal with use of neural-network programming; compensating sudden changes of Ur caused by movement of anode; filtering Ur by means of smoothing filter created on base of first artificial neural network learned by means of sampling preliminarily smoothed value Ur; receiving filtered value U as output signal; determining changing rate dU/dt of filtered voltage in time; converting said value to first derivative of filtered voltage dU/dCal at using as conversion coefficient value dCal/dt; determining second and third derivatives of voltage d2U/dC2al and d3U/dC3al; using received values of first, second and third derivatives of voltage U as input signals for second artificial neural network learned on base of analytical relations between first, second and third derivatives of voltage U and concentration of alumina in electrolyte; using current value Cal of alumina concentration as outlet signal. EFFECT: enhanced technical and economical factors of aluminium production process. 3 cl, 5 dwg, 2 tbl |
