US3812024A - Control of an aluminum reduction cell - Google Patents

Abstract

A PROCESS FOR CONTROLLING THE FEEDING OF ALUMINA TO A REDUCTION CELL COMPRISING DETERMINING A SMOOTHED CELL RESISTANCE, COMPARING THE CELL RESISTANCE WITH THE LOWEST BASE LEVEL RESISTANCE AND, WHEN THE DIFFERENCE EXCEEDS A PREDETERMINED AMOUNT $R*, FEEDING ALUMINA TO THE CELL. THE VALUE OF $R* IS VARIED ACCORDING TO THE NOISE LEVEL OF THE CELL.

Description

May 2l, 1,14 w. H, GQODNQW ET AL 3,812,024

\ CONTROL 0F AN ALUMINUM REDUCHQN CELL Filed Mar'ch 20 1972r 2 lSweeps-Sheet 1 -FQOM @A r//oof PeEcEea/NG CELL 4 2o L12 k2o TO ANODE lBUS BR OF NEXT CELL TIE- :1 v

May 2l, 1974 w. H. GooDNow ErAL 3,812,024

CONTROL OF AN ALUMINUM REDUCTION CELL Filed March 20, 1972 2 Sheets-Sheet z l /09 y /oo V fcALcuL A re BASE LE VEL 65T /N/r/A L kL/AL U65 PEL'A/VCE 'Ralf/Q0. v L pzoRi-j/nn CA LCULA rE Ala/5E /NDEX /Ol nmz l Alf): IDELTA J Vn READ V, 1lf=0 Por VOL rs v H2 c AMPS. "l A /102 cALcL/LA rE CELL L RES/SMA/CE No EL.: w A l u I" @LCL/LATE G0 T0 ,04 A E 103 AR* a +b/VI No K# K 1 114 l CALCUL A rE A /2 VES 0 l 5 AR =QI ,QBL CALCUL/:f5 60 T0 B G0 ro A sMooTf/L-'D 6 ,1955/5 m /vcE 115 f El =f(7i Ri-1 "Rin) E5 Acr/ VA rE /O FEED /A/o/CA 70x25 cALcL/Larf 0f V/A r/o/v No nar/4f RI /z @o ro B [108 I IE. a

ABSTRACT orl 'DISCLOSURE y c A process for controlling the feeding of alumina to a reduction cell comprising determining la smoothedcell resistance, comparing the cell resistance -with the lowest base level resistance and, when the differenceY exceeds a predetermined amount ARI, feeding alumina to the cell. The value of AR* is varied according to the noise level of the cell.

BACKGROUND OF THE -INVEN'IION This invention relates to the 'control of an aluminum reduction cell. The production of valuminum by the electrolysis of alumina dissolved in a molten salt electrolyte such asV cryolite is an old and well-known process commonly termed the Hall-Heroult process. The alumina, which is dissolved in an electrolyte, breaks down into its components, the oxygen being liberated at the anode and the metallic aluminum being deposited in a pool of molten metal which forms in the cavity of the electrolytic cell.

United Sw Patent The liberated oxygen combines with the carbon `.of the anodes to form a mixture of carbon dioxide and carbon monoxide which evolves from the cell.v T he pool of molten metal at the cavity of the cell in etect constitutes the cathode of the cell. y

The conventional aluminum reduction cell comprises a steel shell, a lcurrent-carrying carbonaceous lining disposed therein, and one or more movable anodes Adisposed within the cavity defined by the carbonaceous lining. The carbonaceous cathode lining lmay be a monolithic lining, which is tamped into place and baked in during the operation of the cell, or it may bel composed of carbonaceous 'blocks which have 'been bakedprior to installation of the cell. A plurality of conductor bars are embedded in the cathode lining. Frequently, insulating material such as granular alumina is disposed between the steel shell vand the carbonaceous lining to conserve ythe heat generated during the electrolytic process.

Between about 80 and 90% of theelectrolyte composition normlly consists of a cryolite which is a mixture of sodium and aluminum fluoride (NaaAlF). Excess sodium fluoride results in a loss of efliciency and, therefore, it

is the usual practice to add an excess of aluminum fluoride to prevent this loss of efficiency. Due to its volatility, however, aluminumy fluoride is usually added in restricted production, increasing anode carbon burnoif and genamounts, for example, up to, 10% in excess of the p v crust is usually covered over with additional alumina to preheat thealumina prior to its introduction into the electrolyte and to conserve heat.

As the electrolysis proceeds, alumina is consumed in direct proportion to the metal production. As the alumina concentration in the electrolyte is reduced, a point is reached where a phenomenon known as an anode effect occurs. The voltage drop across the cell, which normally stoichiometric amount of .aluminum fluoride in cryolite.

3,812,024 Patented May 21, 1974 ICC s 2 runs between 4 and 5 volts, rises rapidly to a level of about 40 volts or more. This effect is generally attributed to a low concentration of alumina in the electrolyte. The actual concentration of alumina in the electrolyte at which this effect occurs seems to depend upon the temperature and composition of the electrolyte, the current density, and the like, but it usually occurs at an alumina concentration below about 2% by Weight of the electrolyte. The anode effect is generally terminated by the addition of more alumina. This may be done manually or automatically. The frozen crust, which covers the electrolyte, is broken by a suitable crust breaker and the alumina, which has been previously distributed to the top of the crust, falls into the electrolyte. lFrequently, the addition' of alumina is accompanied by vigorous stirring of the electrolyte to remove the gas film which forms on anode surfaces. This technique quickly causes the anode effect to disappear. The disadvantages of the anode effect are many and varied. One of the major disadvantages is upset of one cell affects the other cells in the series.

Although the anode effect presents serious problems in the control of aluminum reduction cells, this phenomenon is generally a less serious commercial prob- 'lem' than 'the severe overfeeding of alumina to the cell.

A severely overfed cell is commonly termed sick pot or sick cell. If a cell is severely overfed, the alumina does not dissolve in the electrolyte and tends to settle at the bottom of the cell seriously disrupting the current distribution, reducing current efficiency, reducing metal erally reducing cathode life. While it may take a short period, e.g. 5 to 10 minutes, to manually terminate an anode etfect, a sick cell takes considerable time to return to normal operations. Thus it is the usual industrial practice to operate at the lower part of the alumina concentration range to specifically avoid the pot condition. Indeed, many automatic control strategies attempt to promote at least one anode effect per day, specifically to prevent the overfeeding of alumina to the cell.

AOne of the most efficient control techniques which has been developed for aluminum reduction cells is that` described in the copending application of D. R. Bristol et y al., Ser. No. 731,901, now U.S. Pat. No. 3,625,842. In

where R, is the resistance in ohms,

V1 is the measured cell voltage, Ii is the measured cell current, and

Ais the back electromotive force.

-From a series of R1 values a base level resistance RB is periodically determined by suitable smoothing or averaging techniques. Periodically, an averaged or smoothed cell resistance RI is derived from. several sequential R,

V VV'3,812,924I y values. The smoothed cell resistance RI is periodically. Lchanging acellparameter, and in particular alumina concompared with the lowest previously determined RB values for the particular cycle, and when this difference, herein termed AR, exceeds a predetermined value, herein termed AR*, alumina is added to the cell. Although analog-type lcomputing devices can be employed to calculate `and average or smooth the various resistance values, digital computing devices have been used almost universally. Although the back electromotive force, A, varies somewhat according to the alumina concentration in the electrolyte, it is presumed to be constant for purposes of the above calculation. A typical value of A is 1.6 volts.

The control techniques described by Bristol et al. were a significant advance in the art of controlling an aluminum reduction cell, Abut serious problems were found in the commercial embodiment of this technique. It was found that despite filtering the voltage and current measurements, as well as the various smoothing or averaging techniques used in calculating the various resistance values used in the control program, the control program became confused during periods when the noise level of the cell was higher than normal. The control strategy tended to overfeed the cell during periods of unstable operation which are characterized by a high noise level. An overfed cell tends to become increasingly noisy, and therefore the problem of overfeeding is quickly magnified. Unless the cell operators take timely corrective action, the continuous overfeeding of an unstable cell willrapidly create la sick cell. Therefore, in the commercial application of this control technique, it was found necessary to remove the cell from automatic feed control when the noise level of the cell became extremely high. In one commercial potline in which 144 cells are connected in a series, vit was found that from up to 40% of the cells were off automatic control at any one particular time due to the high noise level of the cell.

As used herein, the expression noise, and terms of similar import, refers to the deviation or fluctuation of the electrical characteristics of the cell over a period of time. These fluctuations are commonly termed the A.C. component. These characteristics of the cell can be current, voltage or resistance. A high noise level can result from many factors other than overfeeding such as im- DESCRIPTION OF THE INVENTION This invention is directed to an improvement in 'the automatic control of an aluminum reduction cell wherein the cell resistance is calculated from the cell current and cell voltage, the cell resistance is compared with a base level resistance, and if the difference is greater than AR*, cell parameters such as alumina concentration are method of determining the set point for the difference in resistance to provide improved cell control. The present invention minimizes the tendency of overfeeding of alumina to the cell when the noise level of the cell becomes higher than normal.

changed. More particularly, this invention relates to a The inventors have found that by varying the setpoint AR*, the criterion for feeding alumina to the Cel1,in'ac cordance with the noise level of the cell, that the tendency to overfeed is significantly reduced. It has been found that, as the noise level increases, the AR* set point must be increased accordingly.

In accordance with this invention the set point for centration, AR*, or the demand feed criterion, is determined from the general equation:

AR*--=a+b -N.)m 2) where l 'a and Ib arelpredetermined constants, and

The constants a and b arid the exponent m be Achosen according to the characteristics of the Vcell such as cell size, cell design, and the like. If it is desired to have a greater sensitivity to noise, it may be desirable to use 5a larger number for the exponent m. In controlling an experimental 170,000 ampere cell in accordance with this invention, the constant a was set at 0.25, the constant b was set at 0.36, and the exponent was set at l. The Noise Index varied from 0.5 micro-ohms for a stable condition to 50 micro-ohms for a highly unstable condition.

In the preferred embodiment of the invention, the reduction cell is scanned periodically (e.g'. every 2 to 12 seconds) to measure the cell voltage and the cell current, and, using these valuesin equation (1;), calculating resistance Ri. A base level resistance RB is obtained by fa smoothing technique from a series of R1 values which have been calculated over a period of time, (e.g. 2 minutes). A smoothed cell resistance RI is obtained also by a smoothing technique from a series of Rl values which have been calculated over a second interval of time, nor- `mally shorter than the first interval (e.g. 16 seconds vor demand feed criterion AR*, then alumina is fed to the cell. An alternate, but equivalent, procedure involves deriving the RB and R1 from equation (l) utilizing smoothed voltage and current values. However, this procedure-'is less desirable because-it increases data storage requirements and, because -both voltage and current values must be smoothed, involves twice as many smoothing operations as the preferredmethod described above.

As usedherein, the' term smoothing Vor words of simif'lar import,refer to` the electrical, mechanical, electromechanical and mathematical techniques for determiningv a representative value from a number of-such values or predicting the next value Afrom a series of such values. Included in vthis terminology are mathematical methods 'such as averaging and predicting techniques whichl are described in Data Smoothing and Prediction by RL B. Blackman (1'965). Agpredictin'gv technique is preferred in determining RI."A` predicted value will generally be slightly highervthan` an averaged value because the resistance is gradually rising during most of the controlcycle. Thus the AR* when averaged values are utilized will be smaller than when a predicted'value is utilized.

Periodically, the criterion for feeding alumina, AR*, is updated and changed in accordance with equation (2) or its equivalent. The noise index NI in equation (2) is a measure of the deviation or A.C. component of the electrical characteristics. It may bethe absolute maximum deviation of R1 values over a period of time, it may be the maximum deviation of R, values over a period of time from an averaged or smoothed value such as RI, it may be a summation or average of the absolute value of these deviations over a period of time, or it may be a filtered deviation of R1 values from vR'x values. Similar measures of the variation of voltage or current could be used. However, the constant b would have to be adjusted to compensate for the `different units. l

With reference to PIG. 1, which shows a schematic diagram `of a computer controlled reduction cell, it maybe seen that the cell 10 comprises a metal shell 11, an nsulating layer 12 which can be of any desired refractory material, usually alumina, disposed wtihin said shell and a conductive lining 13 disposed therein, preferably carbonaceous in nature. Within the cavity defined by conductive lining 13 are a pool of molten metal 14 at the bottom of said cavity, a body of molten electrolyte 15 disposed thereon, and a crust 16 composed of frozen electrolyte and alumina disposed on top of the molten electrolyte. Partially immersed in the electrolyte are carbonaceous anodes 17 which are electrically connected to the bus bar 19 by anode rods 18 and supported thereby. (Means to connect the anode rod to the bus bar are not shown.) Disposed within the carbonaceous cell lining 13 are cathodic current collecting bars 20, the outer ends of which project out of the shell 11. The exposed ends of the conductor bars 20 are electrically connected to the anode bus bar of the next cell by suitable means not shown. The anode bus bar 19 is electrically connected to the cathodic conductor bars of the prior cell in the pot line by means not shown.

As shown in FIG. l, the cell is provided with a current sensing device 21 which is suitably connected to the anode bus bar 19. A voltage sensing device 22 is connected between the anode bus bar 19 and the cathodic conductor bars 20. The current sensing device and voltage sensing device are operably connected to a computer device 23. The computing device may lbe an analog or digital device, although the digital computing means is preferred because of its speed and data storage capacity. The current and voltage sensing means 21 and 22 may be of any suitable means for performing these functions. However, if the computing device 23 is digital in nature and the output of the sensing devices is analog, suitable analog to digital converters are employed to render the devices compatible. A multiplexing unit (not shown) may be employed to monitor the voltage and current sensing means of each cell of the pot line. The computing device 23 may be provided with suitable input and readout devices.

The computing device may be, and usually is, programmed to provide many functions in addition to those described herein. For example, the computer program may have subroutines to control the anode to cathode distance, optimize the back of electromotive force value, optimize current distribution in the anodes, and the like. However, in the present invention, the computing device has two prime functions and these are to generate a signal for feeding alumina when the cell resistance rises to the point when an anode effect is approaching, and to adjust the criterion for feeding alumina in accordance with the noise level of the cell.

FIG. 2 is a simplified flow diagram of the various steps involved in the operations of a typical computer program for the present invention for a single reduction cell. This type of ow diagram is commonly prepared by programmers prior to writing the detailed computer program. In this ligure the rectangles are used to designate a calculating, or operating, step and the diamond shape indicates a comparing step. The program begins at point A. In block 100 the counter K is set at a value of 1 and the lowest base level resistance RBIJ is set at an abitrary value of 100, a lvalue considerably higher than normal resistance observations. In the operation of block 101 the cell is scanned and the voltage V1 and the current I1 are measured. These values are used in block 102 to determine the cell resistance R1. K is then compared with a predetermined value n in diamond 103, and if K is less than n, K is increased by l as shown in block 104. The cell is again scanned, to measure current andvoltage and the cell resistance is again calculated. This loop continues until the number of scans n is reached. When K is equal to or greater than n, the program proceeds to block S wherein the smoothed cell resistance RI is calculated by a suitable smoothing technique from the last n-Rl values. The deviation of the resistance during the scanning period, DELTAi, is calculated in block 106 by subtracting the most recently determined cell resistance iR, from the smoothed resistance RI. Supplement to calculating DELTAl, the counter K is compared in diamond 107 with nn, and if K is no t equal or greater than nn, as indicated in block 108, K is increased by one and the program reverts to point B. If K is equal to or greater than nn, the program proceeds to block 109 wherein the base level resistance -RB is calculated from the last nn values of smoothed resistances RI. The noise index NI is calculated in block 110` by averaging the absolute values of the DELTAi previously calculated during the scan period. In the comparison step of lll, if the base level resistance RB is less than the lowest base level resistance RBL, then the old value of RBL is replaced by the calculated base level resistance and the program reverts to point B. If the base level resistance RB is greater than the lowest base level resistance RBL, then the program proceeds to block 113 wherein the demand feed criterion AR* is calculated. In block 114 the increase in resistance AR is calculated by substracting the lowest base level resistance RBI, from the smoothed Yresistance RI. As shown in diamond 115, if AR is greater than the demand feel criterion ARZ, then the program proceeds to block 116 where the feed indicator is activated for the feeding of alumina to the cell. The program then reverts to point A and K is reset at l. If AR is not greater than the demand feed criterion, the program reverts to point B. In a typical program, n is normally set at about 8 and the value n=n is set at about 128. The cell is scanned normally about every two to twelve seconds to measure voltage and current values.

Usually a delay of up to l0 minutes is programmed to occur after the feed indicator is activated to allow time for the alumina to dissolve in the electrolyte and for the anode-cathode distance to be adjusted.

The process of the present invention lowers the average alumina concentration during upset cell conditions by increasing the average time between feeds and, because of this, the anode effect frequency may increase slightly. However, the small increase in anode effect frequency does not significantly alect the overall efficiency of the control technique.

It is obvious that various modifications and improvements can be made in the above-described invention without departing from the spirit thereof and the scope of the appended claims.

What is claimed is:

1. In the process of controlling the feeding of alumina to an aluminum reduction cell wherein a base level resistance RB and a smoothed cell resistance RI are periodically determined from the electrical parameters of the cell, a lowest RB is determined from a series of RB values, the diiference AR between the lowest RB value and RI is periodically determined and said difference AR is compared with a criterion AR* for feeding alumina to the cell and when AR equals or exceeds AR*, alumina is fed to the cell, the improvement comprising (A) sensing the fluctuations of an electrical parameter of the cell selected from the group consisting of cell resistance, cell voltage and cell current over a preselected period of time, and

(B) determining AR* from the equation where a and b are predetermined constants, where N1 is a measure of said fluctuations of an electrical parameter, and m is a number from 1 to 2.5. 2. The process of claim 1 wherein the RB and R1 values are determined from the equation.

where `R is a resistance in ohms,

3,812,024 7 V is a smoothed cell voltage, I is a smoothed cell current, and A is a predetermined back electromotive force.

3. The process of claim 1 wherein the RB and RI values of time. are determined from a series of RI values 'which are 5 determined from the equation 3,625,842 R=V1I- A 3,660,256 3,712,857

where 10 R1 is the resistance in ohms, V1 is the measured cell voltage,

I! is the measured cell current, and

A is a predetermined back electromotive force. 4. The process of claim 1 wherein NI is a measure of the uctuation of cell resistance over a preselected period References Cited UNITED STATES PATENTS HOWARD S. WILLIAMS, Primary Examiner D. R. VALENTINE, Assistant Examiner

Images