FIELD OF INVENTION

[0001] The invention relates to a method for predicting the length of submerged arc furnace (SAF) electrodes and for the automatic adjustment and control thereof, and a system that employs same.

BACKGROUND OF INVENTION

[0002] A SAF typically contains a plurality of electrodes, normally three or six, concentrically mounted in a mantel and capable of independent vertical movement, up and down, by means of a respective electrode positioning mechanism. This vertical positioning parameter is a function of electrical current flow within the furnace, with each electrode controlled to move up or down depending upon the current value.

[0003] Each electrode is powered by electrical input from the furnace transformers via a high current line which terminates at a contact clamp on the electrode.

[0004] In addition to the electrode positioning, the electrode tip length (defined as the length of the electrode from the contact clamp to the electrode tip, and hereinafter referred to as “electrode length”) may be increased by means of a hydraulic or electric electrode slip mechanism. [0005] Electrode length and, more importantly, uniformity in this length or equidistant electrode-tip depth across the electrodes, is one of the most important parameters in optimally operating the furnace from an energy and material loss point of view. Uniformity in the electrode lengths provides a substantially planar reaction zone within the furnace, maximising energy efficiency and material productivity.

[0006] This uniformity is difficult to maintain as the length of each electrode will change relatively to the others as a result of:

• a different rate of consumption within the furnace due to the paste quality of each electrode or variance in the composition (carbon balance) of the burden surrounding the respective electrode;

• a different power input to each electrode; · a different burden discharge rate into the furnace causing irregular build-up of burden about the electrodes; or

• a different slippage rate.

[0007] Continuous direct measurement of the electrode lengths to determine the extent of this electrode length unevenness is practically impossible, not only for the reason that the electrodes operated in an environment of extreme temperature, but that the electrode tip is submerged in the furnace burden.

[0008] One method of determining the electrode length is to measure the electrode weight with load cells. This method is complicated by the fact that numerous forces are acting on the electrode, including forces caused by the pressure ring and contact shoes, roof seal friction and direct load applied to the electrode tip by the burden in the furnace. As these forces are not constant, any electrode length calculation employing this method will be inaccurate,

[0009] Another method employs mathematical models, inputting current and voltage measurements, to provide an electrode length estimation. However, a smelting operation is characterized by multiphase reactions and complex chemistry, as well as ill-defined momentum, heat and mass transfer phenomena. This makes the development of accurate mathematical models highly challenging.

[0010] Thus, real-time knowledge of the electrode length in a submerged arc furnace, the single most significant parameter to optimal performance, currently cannot be continuously and accurately reported.

[0011] The present invention at least partially solves the aforementioned problem.

SUMMARY OF THE INVENTION

[0012] The invention provides a method for predicting a length of each of a plurality of electrodes in a submerged arc furnace and for making an adjustment to the length of an electrode based on the length as predicted, which method includes the steps of:

(a) entering into a processor which runs a program which includes an artificial intelligence algorithm a dataset which includes a measurement of the length of each electrode (“the measured length”) and information on one or more of the following which is associated with the measured length: the temperature of a hearth of the furnace, the temperature of a sidewall of a furnace, the electrical power supply to the electrodes, the rate at which the electrode moves through a holder of each electrode and the vertical position of the holder;

(b) entering into the processor data from one or more of the following: a first temperature sensor for measuring the temperature of the hearth, a second temperature sensor for measuring the temperature in the sidewall, a power measuring device for measuring the electrical power supply to the electrodes, a slip meter for measuring the rate at which the electrode moves through the holder, and a position sensor for measuring the vertical position of the holder; (c) obtaining from the processor based on the dataset and the data an estimate of the length of each electrode (“the estimated length”); and

(d) adjusting the length of an electrode based on the estimated length of each electrode by changing the rate at which the electrode moves through the holder or by changing the vertical position of the holder. [0013] The dataset may be a historical batched dataset. The data may be a current stream of data.

[0014] The method includes the additional steps of analysing the chemical composition of a batch of ore material (charge) which will be fed into the furnace and entering the results of the analysis to the processor concurrently with the data pertaining to the batch of charge when processed or smelted within the furnace. [0015] More specifically, the method includes the step of analysing, at least, the carbon content of the batch of charge which will be fed into the furnace and feeding the results of the carbon content analysis into the processor concurrently with the data pertaining to the batch charge when processed or smelted within the furnace. [0016] Additionally, the dataset and the data may contain information on the furnace pertaining to the power at which the furnace is operating, the temperature of the off gas, the chemical composition of the off-gas, the chemical composition of the slag, the metal alloy and the charge.

BRIEF DESCRIPTION OF THE DRAWINGS [0017] The invention is further described by way of an example with reference to the accompanying drawings in which:

Figure 1 diagrammatically illustrates a submerged arc furnace to which the method of the invention is employed; and

Figure 2 is a flow diagram illustrating the method of the invention. DESCRIPTION OF PREFERRED EMBODIMENTS

[0018] Figure 1 illustrates a submerged arc furnace (SAF) system 10 to which the method of the invention is applied.

[0019] The system includes a SAF 11 which has a plurality of electrodes, respectively illustrated 12A and 12B (the third electrode in the plurality is not illustrated). Each of these electrodes is a typical Soderberg type composite or graphite electrode which will not require further description. The electrodes are vertically mounted, in a radial array, with a lower portion 14 extending into a furnace 16. The lower portion terminates at a tip 16.

[0020] The furnace 17 is defined by a hearth 18, a sidewall 20 and a water-cooled hood 22. The hearth and the sidewall have a refractory lining. A bath 24 is contained within the furnace. The bath is comprised of a number of layers, including an upper solid or semi-solid furnace burden 26 and a lower liquid layer of molten and slag metal 28. The lower layer periodically is tapped through a tap hole 30.

[0021] As the ore in the burden is smelted down, and the lower liquid layer tapped off, the burden is recharged with a charge 31 (ore, reductant, fluxing agent), which is input through a feeder system 32 via charge input chutes 34.

[0022] The gaseous by-product of the smelting process is removed from the furnace 16 via an off-gas duct 36.

[0023] Each electrode is powered by electrical power from an electrical power source 40. The source is connected to the respective electrode via a high current line which terminates at a contact shoe 42. The contact shoe is enclosed by a contact clamp 44 which clamps the contact shoe into electrical contact with the electrode.

[0024] To ease the explanation that follows, reference is made to a single electrode.

[0025] An electrode slip mechanism 46 (ESM) is engaged with the electrode 12 to hold the electrode vertically in suspension, penetrating the hood 22 and positioning the tip 16 within the burden 26. The mechanism comprises a bottom and a top clamping ring (48A and 48B) and a hydraulic actuator 50. The actuator causes the clamping rings to compress or release their clamping grip on the electrode. [0026] Another actuator, the electrode positioner mechanism (EPM) 52, is engaged with the electrode 12 to move the electrode vertically upwardly or downwardly within a finite set vertical range. The EPM includes a hoist platform 54, which is fixedly connected to the ESM, and a pair of co-operating hydraulic hoists 56 which are interposed in respective suspension arms 58 which connect between an upper level floor 60 of the furnace housing (not shown) and the hoist platform.

[0027] The parameter that is estimated and controlled by the method of the invention, as described below, is a length of the electrode, being a portion of the electrode from a bottom edge 62 of the contact clamp, to the tip 16 and designated X on Figure 1 (“electrode length”).

[0028] The electrode length is operationally managed by altering the rate at which the ESM allows the electrode to slip relatively to it. The slip rate is a variable of the clamping force imposed on the electrode by the ESM. Adjusting the slip rate, by actuating clamping rings, adjusts the rate at which the electrode is fed into the furnace. The slip rate, along with the factors described in the background above, will vary the electrode length.

[0029] To estimate the electrode length for each electrode, in an attempt to ensure that there is substantial uniformity in this length for the reasons explained above, the method of the invention is employed. [0030] A plurality of sensors is required to receive and relay information on operational parameters in accordance with the method as described below. To this end, the furnace system 10 is fitted with: a plurality of thermocouples (64.1, 64.2, 64.3 ...) located within the refractory layer of both the hearth 18 and the sidewall 20; a thermocouple 66 which is associated with a transformer 67 of the electrical power source 40; a thermocouple 68 and an off-gas analyzer 70 which is associated with the off-gas duct 36; a power measuring means 72, for example a voltage transformer, a current transformer or a kilowatt-hour meter, electrically associated with the power source; an electrode position sensor 74 to measure the relative position of the hoist platform 54 of the EPM 52; and a encoded slip meter 86 associated with the ESM 46.

[0031] Each of these sensors is connected to, or in communication with, a processor 76 (see Figure 2) which runs a program which includes an artificial intelligence (Al) algorithm. The processor, in turn, communicates with both the ESM

46 and the EPM 52.

[0032] A machine learning step precedes real-time predictive and responsive steps. In this preceding step, a series of historical datasets are compiled of the operational parameters of the furnace system 10 including: temperature (in at least the hearth 18 and the sidewall 20, but also preferably in the transformer, the off-gas duct; hoist platform (holder) 54 position; power input provided by the power supply 40; chemical make-up of the off-gas; chemical analysis of the ore material (charge) input, the tapped metal and the slag; and the electrode slip rate. Regarding the latter, this rate can be measured manually alternatively to being determined by the slip meter 86. [0033] Associated with each dataset is an actual measurement of the respective electrode 12 which measurement is carried out after a melt-down process known in the art. The datasets and associated measurements are stored in a historical database 78 and fed to the processor (76.1) to “train” the Al algorithm by associating information derived from the operational parameters with electrode lengths,

[0034] Once the algorithm is trained, real-time process data-streams can be received by the processor (76.2) from one or more of the sensors (64, 66, 68, 70, 72, and 74) to enable the processor to make a predictive calculation on the length of the electrodes 14 at any given moment.

[0035] Supplementary to these real-time data-streams is the data input derived from conducting a chemical analysis of the ore, metal and slag. These steps are respectively designated 80, 82 and 84 on Figure 2. The chemical analysis, done on the ore, metal and slag (charge 31), is date-stamped. This step is employed to ensure that the chemical data is input to the processor concurrently with receipt of the real-time data-streams emanating from the processing of the batch of ore to which the chemical data pertains. In addition, manually calculated data pertaining to a slip rate 86 is also input to the processor 76 with the contemporaneous real-time data-streams and the chemical data.

[0036] The processor, running the Al algorithm, is able to output an accurate prediction of the length of the electrodes 88. This predictive calculation becomes increasingly accurate the more information the historical database contains.

[0037] The inventor conducted a trial to determine the accuracy of the predictive outcomes of the method of the invention when compared to actual electrode lengths.

Preceding the trial, nine months of actual data was used to train and test the Al algorithm. In conducting the trial, two months of data were used in the method of the invention, without an associated measured electrode length, to predict the electrode lengths for the three electrodes 12 of the furnace system 10 (see “Predicted length” rows). Thereafter, these predicted lengths were compared against actual measurement lengths (see “Actual measured length" rows). The actual measurement lengths were never introduced to the Al model to ensure the integrity of the results.

Table 1 below tabulates the results of the trail.

TABLE 1

[0038] The results proved the method of the invention to be very accurate in predicting the length of the electrodes with a maximum variance (between predicted and actual) of only 1.96%.

[0039] With the accuracy of the output assured, the predicted electrode lengths can be acted upon remedially by the actuation of the ESM 46 and/or the EPM 52, to align the electrode tips of each electrode 14 substantially in planar orientation. This can be achieved automatically (as illustrated in Figure 2 whereby the processor automatically actuates the ESM and/or the EPM) or by a human operator,