BACKGROUND

[0001] Embodiments of the present specification relate generally to smelting, and more particularly to systems and methods for controlling aluminum smelting pot line.

[0002] The process for producing aluminum consists of mining and purification of bauxite, production of alumina ore and electrolytic reduction (or smelting) of alumina to obtain aluminum. In the conventional reduction operation, alumina ore is dissolved in a molten salt electrolyte such as cryolite contained in a plurality of reduction cells (also referred herein as pots) connected in series forming a pot line. Each of the plurality of pots of the pot line includes an anode and a cathode. Molten aluminum is deposited at the cathode by passing a very large current, on the order of thousands of amperes of current, between the anode and cathode electrodes. The production of aluminum is a function of the relative ion concentration as measured in terms of the percentage of alumina dissolved in the molten electrolyte under specific conditions of cell temperature and current density. As the apparent alumina concentration in the electrolyte decreases below a particular level, a continuous gas film is rapidly formed on the anode surface causing an "anode effect" characterized by increased resistance between anode and cathode. The anode effect reduces production in the entire pot line since the increased resistance reduces current flow through the pot line. The increase in cell resistance also causes generation of heat raising the temperature of the electrolyte and causing loss of fluoride from the electrolyte. Addition of alumina to the electrolyte causes the gas film on the anode to dissipate, returning the cell resistance to a normal value. While anode effects are of primary concern in the operation of a pot line, excess alumina in the cell electrolyte must also be avoided since excess alumina contaminates the metallic aluminum, and adversely affects the normal operation of the cell.

[0003] The control of aluminum reduction cells is important to optimize economics of the smelting process. The temperature and AlF3 concentration of electrolyte are very important factors for aluminum production cells. Temperature should be kept in a reasonable range, usually within 10C (degree Celsius) of a target temperature. The target temperature typically varies from 940-970°C (degree Celsius) depending on technology used for constructing the Aluminum smelting pot. According to an estimate, for an increase in temperature about 10 degrees beyond the required value, the current efficiency decreases by 2% and the energy consumption increases by 3%. The AlF3 concentration of electrolyte is one of the most important factors affecting the temperature of the alumina reduction. The suitable operating temperature should be stable as lower operating temperatures do not allow Al203 to dissolve and reduce the stability of operation of cells. The operating temperature is based on the composition of the electrolyte. The concentration of AlF3 and Al203 in electrolyte usually changes faster than others with time. Typically, the concentration of Al203 is controlled automatically by the cell controller to be kept between 1.5 and 2.5%. The concentration of AlF3 is an important parameter to affect the operating temperature.

[0004] The industrial aluminum production is a dynamic process which has some repeated operations, such as extraction of aluminum on a daily basis, changing of anode, and avoiding the anode effect. The smelting process is a multivariate, nonlinear process. It would be advantageous to have an effective model for characterizing the smelting process in a timely manner in spite of the complex nature of the process.

BRIEF DESCRIPTION

[0005] In accordance with one aspect of the present specification, a method, is disclosed. The method includes receiving smelting data corresponding to a pot line in operation. The smelting data includes a pot parameter, an electrolyte parameter and a timing parameter. The method further includes determining a temperature estimate, an estimate for excess of aluminum fluoride, and an estimate of aluminum fluoride consumed based on the smelting data corresponding to a smelting pot. The method also includes determining an estimate of aluminum fluoride dose to be supplied to the smelting pot based on the temperature estimate, the estimate for excess of aluminum fluoride, and the estimate of aluminum fluoride consumed. The method includes feeding the smelting pot with the estimate of aluminum fluoride dose for continued optimal operation of the pot line. [0006] In accordance with another aspect of the present specification, a control system is disclosed. The control system includes a data acquisition unit configured to receive smelting data corresponding to a pot line in operation. The smelting data includes a pot parameter, an electrolyte parameter and a timing parameter. The control system further includes a machine learning unit communicatively coupled to the data acquisition unit and configured to determine a temperature estimate, an estimate for excess of aluminum fluoride, and an estimate of aluminum fluoride consumed based on the smelting data corresponding to a smelting pot. The control system also includes a recommendation unit communicatively coupled to the machine learning unit and configured to determine an estimate of aluminum fluoride dose to be supplied to the smelting pot based on the temperature estimate, the estimate for excess of aluminum fluoride, and the estimate of aluminum fluoride consumed. The control system includes a processor unit communicatively coupled to the recommendation unit and configured to feed smelting pot with the estimate of aluminum fluoride dose for continued optimal operation of the pot line.

[0007] In accordance with one aspect of the present specification, an aluminum smelting pot line system is disclosed. The aluminum smelting pot line includes a plurality of reduction cells electrically connected in series. The aluminum smelting pot line further includes a control sub-system communicatively coupled to the plurality of reduction cells and configured to receive, using a data acquisition unit, smelting data corresponding to a pot line in operation, wherein the smelting data comprises pot parameters, electrolyte parameters and timing parameters. The control sub-system is further configured to determine, using a machine learning unit communicatively coupled to the data acquisition unit, a temperature estimate, an estimate for excess of aluminum fluoride, and an estimate of aluminum fluoride consumed based on the smelting data corresponding to a smelting pot among the plurality of reduction cells. The control sub-system is also configured to determine, using a recommendation unit communicatively coupled to the machine learning unit, an estimate of aluminum fluoride dose to be supplied to the smelting pot based on the temperature estimate, the estimate for excess of aluminum fluoride, and the estimate of aluminum fluoride consumed. The control sub-system is further configured to feed, using a processor unit communicatively coupled to the recommendation unit, smelting pot with the estimate of aluminum fluoride dose for continued optimal operation of the pot line.

DRAWINGS

[0008] These and other features and aspects of embodiments of the present invention will become better understood when the following detailed description is read with reference to the accompanying drawings in which like characters represent like parts throughout the drawings, wherein:

[0009] FIG. 1 is a diagrammatic illustration of an aluminum smelting pot line in accordance with an exemplary embodiment;

[0010] FIG. 2 is a system for controlling the operation of the aluminum smelting pot line in accordance with an exemplary embodiment;

[0011] FIG. 3 is a block diagram illustrating training and deployment of machine learning model in the aluminum smelting pot line of FIG. 1 in accordance with an exemplary embodiment;

[0012] FIG. 4 is a block diagram illustrating use of virtual sensors in estimating dosage of aluminum fluoride in accordance with an exemplary embodiment;

[0013] FIG. 5 is a schematic of a pot line having a plurality of pots with a corresponding dosing recommendation engine in accordance with an exemplary embodiment; and

[0014] FIG. 6 is a flow chart of a method of controlling a smelting pot line in accordance with an exemplary embodiment.

DETAILED DESCRIPTION

[0015] Systems and methods for Aluminum smelting are described herein. More particularly, the systems and methods are configured for control of an aluminum smelting pot line. [0016] The term“smelter” used herein refers to a system designed for smelting metals such as aluminum in an industrial setup. The term“pot line” is equivalently and interchangeably used to refer to a smelter in the present specification. Specifically, the pot line includes a plurality of pots coupled with one another to enhance the capacity of the smelter. The pots of the smelter are also referred as“cells” or“reduction cells” in the subsequent paragraphs. The term“bath” refers to a vessel used for smelting in a smelting pot. The term“alumina” refers to aluminum ore and chemically denoted as AI2O3 (aluminum oxide). The term“anode” refers to positive terminal of the pot line and the term“cathode” refers to negative terminal of the pot line. The anode and cathodes are made of carbon and the anode carbon is consumed during the smelting process requiring periodic replacement. The term“anode change” refers to replacement of anode in a smelting pot typically once in a few weeks. The term“virtual sensor” used herein refers to a technique of sensing a physical parameter by using a mathematical or simulation model the employs one or more parameters sensed by available physical sensors.

[0017] FIG. 1 illustrates a system 100 representative of an aluminum smelting pot line in accordance with an exemplary embodiment. The system 100 includes a plurality of pots 102, 106, 110, 114, 118 configured to process Alumina obtained from the refinery to generate liquid Aluminum. It may be noted that the Alumina used in the plurality of pots is obtained from the refinery by processing Bauxite obtained from mining. Each of the plurality of pots 102, 106, 110, 114, 1 18 facilitates chemical reaction of reduction of Bauxite to liquid Alumina in a controlled manner. Typically, the system 100 is installed in long buildings with the plurality of pots 102, 106, 110, 114, 118 connected in series. In one embodiment, each pot such as 110 is typically a steel box with an example dimension of about twenty feet long, six feet wide and three feet deep lined with carbon acting as a cathode 138. Further, the pot also includes a carbon electrode acting as an anode 124. In a typical smelting pot line, more than hundred pots are electrically connected in series.

[0018] As illustrated in FIG. 1, the system 100 includes a power source 144 providing power to the plurality of pots 102, 106, 110, 114, 118 connected in series. The power source 144 is coupled to anode 156 of the pot 102. The cathode of pot 102 is coupled to anode 122 of the pot 106. The cathode 154 of the last pot 118 is connected to the negative terminal of the power source 144. In some cases, the plurality of pots 102, 106, 1 10, 1 14, 118 are housed in several buildings and are electrically connected in series to form a single smelting pot line.

[0019] The alumina, dissolved in a molten salt called cryolite and aluminum fluoride, is used as electrolyte 140 for reduction of aluminum in the pot line. The electrolyte 140 introduced into each of the plurality of pots 102, 106, 110, 114, 118 through an electrolyte inlet 148. Further, aluminum fluoride is introduced into the plurality of pots 102, 106, 1 10, 114, 118 through an aluminum fluoride inlet 150. The system 100 also includes a control subsystem 146 configured to control the quantity of electrolyte, the quantity of aluminum fluoride introduced into each of the plurality of pots 102, 106, 1 10, 114, 1 18. The aluminum fluoride helps in controlling the temperature of the plurality of pots. However, optimal quantity of aluminum fluoride is required to be added to each smelting pot to control the operational cost. The operating temperature within a predefined range and the excess of aluminum fluoride are used as key process parameters in the smelting process. The control subsystem 146 is configured to optimize energy consumption, current efficiency and consumption of raw materials such as aluminum fluoride as key performance indicators (KPIs).

[0020] The control subsystem 146 is further configured to control electrical operating parameters such as, but not limited to, a voltage applied across the pot line and a current flowing through the electrolyte of the pot line. In one embodiment, a direct electrical current is passed through the electrolyte 142 having molten alumina from anode to the cathode. In one embodiment, a current magnitude between fifty thousand amperes and one hundred and fifty thousand amperes and a voltage magnitude of four volts is applied to each pot to perform electrolysis. Electrolysis in each pot of the smelting pot line reduces alumina molecules into molten aluminum 142 and oxygen. The oxygen is deposited on the carbon anode and forms carbon dioxide. The aluminum 142 settles to the bottom of the pot. The molten aluminum may be siphoned through nozzles 104, 108, 112,116, 120 into crucibles at regular intervals typically once a day. More electrolyte 140 is added to the pots 102, 106, 110, 114, 118 through the electrolyte inlet 148 and the electrolysis is continued without interruption. The operating parameters of the pot line is controlled by controlled addition of the aluminum fluoride through the fluoride inlet 150. The molten aluminum 142 extracted from the smelter pot line is cast into billets, ingots, T-Bar or Sow.

[0021] Optimal quantity of aluminum fluoride required by a smelting pot also depends on parameters such as, but not limited to, raw material quality, age of pot, efficiency of gas treatment center. The dosing approach proposed in embodiments disclosed herein is based on coupled machine learning models for temperature, excess AlF3, and AlF3 consumption. The coupled machine learning models involve multiple machine learning techniques having interdependency among them.

[0022] In one embodiment, the temperature of the electrolyte within a smelting pot is determined by a multivariable thermal balance of the energy added and lost from the pot. The energy balance is a strong function of the process conditions and operations on the pot. Higher value of height of metal in pot results in more heat loss from the pot resulting in drop in temperature. Higher value of bath height in pot results in less heat loss resulting in rise in temperature. Higher value of pot voltage corresponds to increased energy added to the pot which results in rise in bath temperature. With the increase in aluminum fluoride doses, aluminum fluoride concentration in the bath increases and temperature of liquid in the pot deceases. With the decrease in alumina concentration, Anode effect sets in resulting in significant amount of energy added to the pot resulting in temperature increase of the bath. During anode change, cold anode is immersed in the pot resulting in temperature drop. Metal tapping operation and bath removal extracts heat from the pot resulting in drop in temperature.

[0023] In some embodiments of the disclosed technique, excess Aluminum Fluoride concentration of the electrolyte in the aluminum smelting pot is modelled as a function of the temperature of the pot. Higher pot temperatures result in lower values of excess AlF3. Increases in AlF3 doses result in increases in excess AlF3. The age of the pot and the age of cathode has an impact on excess AlF3 changes for a given AlF3 dose addition. In some smelting pot lines, a secondary alumina may be regenerated by a gas treatment center and reintroduced into one or more of the smelting pots. In such embodiments, the control subsystem 146 is further configured to consider different fluoride levels of the secondary alumina.

[0024] In one embodiment, the control subsystem 146 is configured to receive a plurality of smelting parameters and determine optimum quantities of electrolyte and aluminum fluoride to be introduced to each of the plurality of pots 104, 106, 112,116, 120. Further, the control subsystem 146 is configured to estimate requirements of electrolyte and aluminum fluoride at frequent intervals. In one embodiment, the control subsystem 146 is configured to employ machine learning techniques to process a plurality of smelting pot parameters and determine the dosing requirements at shorter intervals such as one minute or one hour. The control subsystem 146 disclosed herein avoids costlier and cumbersome physical measurements of the smelting parameters to determine the quantity of electrolyte and the aluminum fluoride required for smelting.

[0025] FIG. 2 is a schematic 200 illustrating working of the control subsystem 146 of FIG. 1 for controlling the operation of the aluminum smelting pot line in accordance with an exemplary embodiment. The schematic 200 includes a block diagram 202 representative of the control subsystem 146, smelting data 204 provided to the control subsystem 146 and a plurality of smelting parameter estimates 206 generated by the control subsystem 146. The block diagram 202 includes a data acquisition unit 224, machine learning unit 228, a recommendation unit 226, a processor unit 230 and a memory unit 232. The units 224, 226, 228, 230, 232 of the control sub system 146 are communicatively coupled with each other through a communications bus 234. In one embodiment, the smelting data 204 includes a plurality of pot parameters 208, a plurality of electrolyte parameters 210, a plurality of process timing parameters 212 and other parameters 214 corresponding to the aluminum smelting pot line. In one embodiment, the plurality of smelting parameter estimates 206 includes an estimate of aluminum fluoride dose 222 to be supplied to one of a smelting pot among the plurality of pots in the pot line, a temperature estimate 216 corresponding to the smelting pot, an estimate for excess of aluminum fluoride 218 introduced to the smelting pot, and an estimate of aluminum fluoride consumed 220 at the smelting pot. In the illustrated embodiment, the plurality of smelting parameter estimates 206 corresponds to one of the plurality of pots in the pot line. However, the plurality of smelting parameter estimates 206 may also include parameters corresponding to multiple smelting pots of the pot line. In one embodiment, the control subsystem 146 represented by the block diagram 202 may be a microprocessor, a controller, a general purpose unit, or a special purpose computational unit interfaced with sensors and storage devices to receive the smelting data 204. The sensors may include, but not limited to, a temperature sensor, a pressure sensor, a timer, a flow sensor, a volume sensor or a weight sensor. In one embodiment, the control subsystem 146 represented by the block diagram 202 is realized using a plurality of processor units such as the processor unit 230 co-operatively coupled to realize the data acquisition unit 224, the machine learning unit 228, and the recommendation unit 226. The plurality of processor units may be present in different locations to form a distributive computational system or may be co located in a same site. It may also be envisioned that the control subsystem 146 or a part of it may be realized as a computation as a service in a cloud based systems.

[0026] The data acquisition unit 224 is communicatively coupled to the plurality of sensors configured to acquire one or more parameters of the smelting data 204 and configured to pre-process the smelting data 204 for further processing. The data acquisition unit 224 includes circuitry required to transmit or receive data from the sensors and may additionally include memory storage and processing power to store and process the smelting data. The data acquisition unit 224 is configured to pre-process the smelting data and in one embodiment, the pre-processing functions include, but not limited to, noise reduction, normalization and outlier filtering.

[0027] The machine learning unit 228 is communicatively coupled to the data acquisition unit 224 and configured to perform one or more machine learning techniques to derive one or more parameters of the smelting parameter estimates 206. Specifically, in one embodiment, the machine learning technique may be a deep learning technique. In another embodiment, the machine learning technique may be a classification method or an estimation technique. In one embodiment, the machine learning unit 228 includes a neural network based estimator for realizing the learning technique. The machine learning unit 228 may include more than one neural network for estimating a plurality of parameters. In one embodiment, one neural network may be used to determine the temperature estimate 216 corresponding to a smelting pot. Further, a second neural network is configured to estimate the excess of aluminum fluoride 218 introduced to the smelting pot. A third neural network of the machine learning unit 228 may be configured to determine the estimate of aluminum fluoride consumed 220 at the smelting pot. The machine learning unit 228 is also configured to train the first neural network, the second neural network and the third neural network based on historical labelled data. The machine learning unit 228 may also employ any other machine learning techniques to estimate one or more of the smelting parameters estimates 206.

[0028] The recommendation unit 226 is communicatively coupled to the machine learning unit 228 and configured to determine the estimate of aluminum fluoride dose 222 to be provided to a smelting pot among the plurality of pots in the pot line. In one embodiment, the estimate of aluminum fluoride dose 222 to be supplied to the smelting pot is based on the temperature estimate 216, the estimate for excess of aluminum fluoride 218, and the estimate of aluminum fluoride consumed 220. The recommendation unit may determine the estimate of aluminum fluoride dose 222 using a mathematical model, a machine learning technique, a Monte Carlo simulation technique or a combination thereof. The recommendation unit 226 is also configured to process the smelting data 204 in generating the estimate of aluminum fluoride dose 222.

[0029] The processor unit 230 is communicatively coupled to the recommendation unit 226 and configured to control feeding of the smelting pot with the estimate of aluminum fluoride dose for continued optimal operation of the pot line. It may be noted that the smelting pot is working continuously and periodical feeding of alumina and aluminum fluoride dose at optimal quantities is required for efficient operation of the smelting pot line. The processor unit 230 acts as a controller to accurately control the feeding of the aluminum fluoride to each of the smelting pots at a frequent interval of an hour to several hours which is equivalent to a fraction of a day. It may be noted that the processor unit 230 may include a plurality of processing elements and each of the processing element may also be configured to perform the operation of one or more of the data acquisition unit 224, the machine learning unit 228 and the recommendation unit 226. The processor elements may be general purpose microprocessor or controller. In an alternate embodiment, the processor elements may be a special purpose processing element such as a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), or any type of special purpose application specific hardware.

[0030] The memory unit 232 is communicatively coupled to the processor unit 230 and configured to store smelting data, parameters of machine learning techniques, or parameters of any mathematical or neural network model used to determine one or more parameters of the smelting parameter estimates 206 or the smelting parameter estimates 216, 218, 220, 222 corresponding to each of the pots in the pot line. The memory unit 232 may be a single memory storage unit or a plurality of smaller memory storage units coupled together through the communications bus 234 and coordinated by the processor unit 230. In one embodiment, the memory unit may be a random-access memory (RAM), read only memory (ROM), or a flash memory. The memory unit may also include, but not limited to, discs, tapes, or hardware drive based memory units. In one embodiment, the memory unit 232 may be a non-transitory memory encoded with instructions

[0031] FIG. 3 is a block diagram 300 illustrating training and deployment of machine learning model in the aluminum smelting pot line of FIG. 1 in accordance with an exemplary embodiment. The block diagram 300 illustrates training phase with the blocks 302, 304, 306, 308, 310, validation phase with the blocks 312, 314 316 and deployment phase with the blocks 318, 320, 322. It may be noted herein that the plurality of machine learning models used in determining the plurality of smelting parameter estimates 206 in FIG. 2 may be trained, validated and deployed using a scheme illustrated in the block diagram 300. In accordance with the illustrated embodiment, the training phase includes acquisition of raw data 302 and pre-processing of data in 304 for preparing the data for further processing. While a plurality of sensors may be used to sense the raw data in block 302, the data acquisition unit 224 is used to pre-process the raw data at the data preparation block 304. Further, a subset of the acquired data having suitable labels may be used as test data 306. The training is performed in block 308 by the machine learning unit 228 to generate a trained model 310. In the validation phase, a subset of acquired may be used to generate validation data by the data acquisition unit 224 at the block 312. The validation of the trained model is performed by the machine learning unit 228 at the block 314 to generate a validated model 316. The validated model 316 is used at block 320 using realtime data 318 to generate a parameter estimate 322 which is one of the plurality of smelting parameter estimates 206.

[0032] FIG. 4 is a block diagram 400 illustrating use of virtual sensors in estimating dosage of aluminum fluoride in accordance with an exemplary embodiment. The block diagram includes a first virtual sensor 402 configured to sense a temperature value in a smelting pot, a second virtual sensor 404 configured to sense a percentage excess of aluminum fluoride in the smelting pot, a third virtual sensor 406 configured to sense consumption of aluminum fluoride in the smelting pot. The block diagram 400 also includes an aluminum fluoride dosing recommendation engine 414 configured to receive the temperature value, percentage excess value of aluminum fluoride in a smelting pot and a consumption quantity of aluminum fluoride during smelting process and determine a quantity of aluminum fluoride dose required for optimal operation of smelting in the smelting pot. It may be noted herein that the first virtual sensor 402, the second virtual sensor 404 and the third virtual sensor 406 may be an estimator or a predictor employing a mathematical model, or a simulation model based on the smelting data. Further, the three virtual sensors 402, 404, 406 are interdependent as represented by arrows 418, 420, 422, 424, 426, 428. Specifically, the temperature value predicted by the first virtual sensor 402 may employ the second virtual sensor 404 and the third virtual sensor 406. Further, the excess aluminum fluoride value predicted by the second virtual sensor may use the first virtual sensor 402 and the third virtual sensor 406. Similarly, the aluminum fluoride consumption in a smelting pot determined by the third virtual sensor may be based on the first virtual sensor 402 and the second virtual sensor 404. In one embodiment, the aluminum fluoride dosing recommendation engine is designed based on an equation given by:

^F t = Fp + k ₁ (Tp— T _t ) + K ₂ (E _p - E _t ) wherein, Ft is representative of target dose of aluminum fluoride, F _P is predicted fluoride consumption, k ₁ , k ₂ constants, T _p is predicted temperature from the first virtual sensor, E _p is predicted value of percentage excess of aluminum fluoride, 7) is target temperature and E _t is target percentage excess aluminum fluoride.

[0033] It may be noted herein that net AlF3 consumption in each pot depends on the pot line and pot level mass balance. Quality of Alumina which is used as raw material, strongly impacts the AlF3 consumption. Higher value of sodium oxide or calcium oxide content in Alumina increases AlF3 consumption. Age of the pot/age of cathode has an impact on % Excess AlF3 changes for a given AlF3 dose addition. ALF3 consumption is also a strong function of fluorine evaporated from the pot which in turn are dependent on the temperature and % Excess AlF3 of the pot. Efficiency of Gas treatment center affects the fluorine% in the secondary alumina delivered to the pot which would impact the net AlG3 consumption. Higher the AlF3 purity, lower is the need of AlF3 to maintain the %Excess ALF3. Net AlF3 consumption is a complex function of these processes and operation variables and can be expressed as

where, is representative of purity of aluminum fluoride, and A1, A2, A3, A4, are

expressed as:

A ₁ 34.3 x W _Na20 + 19 x W _Ca0 + 0.49 X W _Na20 x E _AIF3

A ₂ = x 1 x e ^-age/x 2

[0034] The parameters x1, x 2, x3,x4,x5 are constants specific to a smelting pot. In one embodiment, these parameters may be obtained through historical data corresponding each smelting pot among the plurality for pots in the ppt line. The variable W _Na20 represents weight of Sodium Oxide in Alumina added to the smelting pot, W _caO is weight of Calcium Oxide in Alumina added to the smelting pot, E _AIF3 is percentage excess of aluminum fluoride, age is the age of the cathode in the pot, T is temperature and WFlorine is weight of Fluorine in the secondary alumina coming from gas treatment center.

[0035] In one embodiment, the first virtual sensor 402, the second virtual sensor 404, the third virtual sensor 406, the dosing recommendation engine 414 as described by nonlinear set of equations may be realized by neural network based machine learning techniques. The neural networks may be designed and deployed using training phase, validation phase and the deployment phase as explained previously with reference to FIG. 3. The output of virtual sensors is generated by processing a plurality of present and/or previous parameter values obtained from physical sensors and/or virtual sensors by the trained neural networks. The output of the trained neural network is considered as the present virtual parameter value sensed by a corresponding virtual sensor.

[0036] FIG. 5 is a schematic 500 of a pot line having a plurality of pots with a corresponding dosing recommendation engine in accordance with an exemplary embodiment. The schematic 500 includes a plurality of smelting pots 502, 504, 506, 508 connected in series. Each of the smelting pots 502, 504, 506, 508 are fed independently with Alumina and aluminum fluoride for optimal operation. Each of the smelting pots 502, 504, 506, 508 includes a corresponding dosing recommendation engine among a plurality of dosing recommendation engines 510, 512, 514, 516. The recommendation engine 510 associated with the smelting pot 502 uses parameters estimated by the first virtual sensor 518, the second virtual sensor 520 and the third virtual sensor 522. The recommendation engine 512 associated with the smelting pot 504 uses parameters estimated by the first virtual sensor 524, the second virtual sensor 526 and the third virtual sensor 528. The recommendation engine 514 associated with the smelting pot 506 uses parameters estimated by the first virtual sensor 530, the second virtual sensor 532 and the third virtual sensor 534. The recommendation engine 516 associated with the smelting pot 508 uses parameters estimated by the first virtual sensor 536, the second virtual sensor 538 and the third virtual sensor 540. Each of the dosing recommendation engines 510, 512, 514, 516 determine a recommended dose corresponding to the respective smelting pots among the plurality of smelting pots 502, 504, 506, 508. [0037] FIG. 6 is a flow chart 600 of a method of controlling a smelting pot line in accordance with an exemplary embodiment. The method of 600 includes receiving smelting data corresponding to a pot line in operation at step 602. The smelting data includes a pot parameter, an electrolyte parameter and a timing parameter. In one embodiment, the pot parameter includes at least one of pot dimensions, pot age, coordinates of pot, pot voltage, pot temperature, height of metal, instability in voltage, height of metal, or instability in voltage. Further, the electrolyte parameter includes at least one of a height of bath, a mass of bath, a mass of crust added to pot, a percentage excess aluminum fluoride, a dose of aluminum fluoride, and a dose of alumina. The timing parameter includes at least one of a start time of anode effect, an end time of anode effect, a start time of metal tapping, an end time of metal tapping, a start time of anode change, or an end time of anode change. The timing parameters are used in determining the temperature value predicted by the first virtual sensor 402, the excess aluminum fluoride value predicted by the second virtual sensor and the aluminum fluoride consumption in a smelting pot determined by the third virtual sensor. Further, the timing parameters may also be used in determining the consumption rate of carbon in the anode and thereby indicating the time of anode change during the course of smelting.

[0038] The method 600 further includes determining a temperature estimate, an estimate for excess of aluminum fluoride, and an estimate of aluminum fluoride consumed based on the smelting data corresponding to a smelting pot at step 604. In one embodiment, the temperature estimate, the estimate of excess aluminum fluoride and the estimate of aluminum fluoride consumed are determined using machine learning technique. It may be noted that any machine learning technique such as, but not limited to, a neural network based deep learning network may be used for estimating one or more of the temperature, aluminum fluoride consumption and the percentage excess of the aluminum fluoride. Specifically, in one embodiment, determination of the temperature estimate is based on a first machine learning model trained to provide a temperature estimate by processing the smelting data. Similarly, determination of percentage excess aluminum fluoride is based on a second machine learning model trained to provide a quantity measure of aluminum fluoride based on the smelting data. Further, in another embodiment, determination of consumption of aluminum fluoride may be provided by a third machine learning model trained to provide a quantity of aluminum fluoride based on smelting data.

[0039] The method 600 further includes determining an estimate of aluminum fluoride dose to be supplied to the smelting pot at step 606 based on the temperature estimate, the estimate for excess of aluminum fluoride, and the estimate of aluminum fluoride consumed. In one embodiment, the temperature estimate, the estimate for excess of aluminum fluoride, and the estimate of aluminum fluoride consumed are determined by machine learning techniques. Further, the determination of the estimate of aluminum fluoride is also performed using a machine learning technique. A neural network based model may be determined in a training phase using the smelting data and the outputs of three virtual sensors configured to estimate the temperature, excess aluminum fluoride and aluminum fluoride consumption as explained previously with reference to FIG. 3.

[0040] The method 600 also includes feeding the smelting pot with the estimate of aluminum fluoride dose for continued optimal operation of the pot line at step 608. It may be noted that, in embodiments disclosed herein, a separate estimate of aluminum fluoride dose is determined for each of the smelting pots in the pot line. The frequency of estimation may be once in every few hours, once in an hour and once in a minute depending on the requirements of the smelting process.

[0041] Embodiments for the present specification is able to perform aluminum smelting with minimized energy consumption and optimal use of aluminum fluoride and other input components. The technique of machine learning for determining smelting parameter estimates avoids physical measurements of temperature and excess of aluminum fluoride in a smelter which are time consuming and labor intensive. Further, the present technique is able to provide parameter estimates at reduced time intervals in real-time without requiring to wait for completion of analyzing the samples in laboratory. Further, the present technique is able to provide key process parameters for each of the plurality of smelting pots without using generalized rules and/or lookup tables.

[0042] The above-described advantages should be regarded as illustrative rather than restrictive. It is to be understood that not necessarily all such objects or advantages described above may be achieved in accordance with any particular embodiment. Thus, for example, those skilled in the art will recognize that the systems and techniques described herein may be embodied or carried out in a manner that achieves or improves one advantage or group of advantages as taught herein without necessarily achieving other objects or advantages as may be taught or suggested herein.

[0043] While the technology has been described in detail in connection with only a limited number of embodiments, it should be readily understood that the specification is not limited to such disclosed embodiments. Rather, the technology can be modified to incorporate any number of variations, alterations, substitutions or equivalent arrangements not heretofore described, but which are commensurate with the spirit and scope of the claims. Additionally, while various embodiments of the technology have been described, it is to be understood that aspects of the specification may include only some of the described embodiments. Accordingly, the specification is not to be seen as limited by the foregoing description but is only limited by the scope of the appended claims.