Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
METHOD OF CONTROLLING A GAS FURNACE FOR MELTING METAL AND CONTROL SYSTEM THEREFOR
Document Type and Number:
WIPO Patent Application WO/2019/115655
Kind Code:
A1
Abstract:
A method of controlling a metal melting process in a gas furnace (3) with a control system (2), the gas furnace comprising a furnace chamber (9) enclosed by walls (5a, 5b, 5c) and a door (7), gas burners (11) configured to heat the metal (1) loaded in the chamber, a flue gas exit (13) and a plurality of sensors (15) including at least one temperature sensor at a top wall (5a) of the furnace chamber and at least one temperature sensor (15d) to measure the metal temperature in its melted state. The method comprises: a) receiving in the control system during the melting process, process parameter measurements from said sensors including at least one temperature sensor, b) simulating in the control system a physical heat exchange in the furnace including radiation between walls and the metal, conduction though the metal and walls, and convection in the furnace chamber including calculating a temperature at a position in the furnace corresponding to said at least one temperature sensor, c) comparing the data obtained from the physical heat exchange model with data obtained from the sensor measurements, d) tuning at least one adjustable parameter of the physical heat exchange model to reduce the difference between compared data, e) based on the tuned physical heat exchange model, calculating in the control system the change over time of the process parameters, including at least one process parameter which is not directly measured up to the end of the cycle, f) transmitting to an operator interface (14) of the control system, information on the change over time of predicted process parameters and at least one recommendation for executing a future operation in relation to the furnace at a recommended time.

Inventors:
ROSTAMIAN, Amin (Route de Vallaire 4a, 1025 St-Sulpice, 1025, CH)
RAPPAZ, Michel (Rue du Simplon 3A, 1006 Lausanne, 1006, CH)
Application Number:
EP2018/084647
Publication Date:
June 20, 2019
Filing Date:
December 12, 2018
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
NOVAMET SARL (Route de Vallaire 4A, 1025 St-Sulpice, 1025, CH)
International Classes:
F27B3/28; F27D19/00; F27D21/00
Other References:
JØRGEN FURU: "An Experimental and Numerical Study of Heat Transfer in Aluminium Melting and Remelting Furnaces", 31 January 2013 (2013-01-31), XP002778819, ISSN: 1503-8181, ISBN: 978-82-471-4152-6, Retrieved from the Internet [retrieved on 20180305]
None
Attorney, Agent or Firm:
REUTELER & CIE SA (Chemin de la Vuarpillière 29, Nyon, CH-1260, CH)
Download PDF:
Claims:
Claims

1. A method of controlling a melting process of a batch of metal in a gas furnace (3) with a control system (2), the gas furnace comprising a furnace chamber (9) enclosed by walls (5a, 5b, 5c) and a door (7), gas burners (11 ) configured to heat the metal (1 ) loaded in the chamber, a flue gas exit (13) and a plurality of sensors (15) including at least one temperature sensor at a top wall (5a) of the furnace chamber and at least one temperature sensor (15d) to measure the metal temperature in its melted state, the method carried out in real time comprising:

a) receiving in the control system during the melting process, process parameter measurements from said sensors including at least one temperature sensor,

b) simulating, in the control system, during the melting process of said batch of metal, physical heat exchange in the furnace with a numerical physical heat exchange model including radiation between walls and the metal, conduction though the metal and walls, and convection in the furnace chamber including calculating a temperature at a position in the furnace corresponding to said at least one temperature sensor,

c) comparing the data obtained from the physical heat exchange model with data obtained from the sensor measurements during the melting process of said batch of metal,

d) tuning at least one adjustable parameter of the physical heat exchange model to reduce the difference between compared data during the melting process of said batch of metal, e) based on the tuned physical heat exchange model, calculating in the control system, during the melting process of said batch of metal, the change over time of the process parameters, including at least one process parameter which is not directly measured up to the end of the cycle of melting said batch of metal,

f) transmitting to an operator interface (14) of the control system, during the melting process of said batch of metal, information on the change over time of predicted process parameters and at least one recommendation for executing a future operation in relation to the furnace at a recommended time during the melting process of said batch of metal.

2. The method according to any preceding claim, wherein the process parameters measured and received by the control system and included in the numerical model further include :

- a flue gas exit temperature ( Tfume ),

3. The method according to any preceding claim, wherein the process parameters measured and received by the control system and included in the numerical model further include :

- a burner gas inlet temperature ( Tgas )

- a gas flow rate {Gfiow

4. The method according to any preceding claim, wherein the process parameters measured and received by the control system and included in the numerical model further include :

- an air inlet temperature (Tair)

- an air flow rate (7 ).

5. The method according to any preceding claim, wherein the process parameters measured and received by the control system and included in the numerical model further include any one or more of:

- a chemical composition of at least one component (such as C02) of the flue gas,

- a gas pressure inside the furnace chamber,

- a furnace bottom wall temperature.

6. The method according to any preceding claim, wherein the simulated physical heat exchange model includes calculating in the control system the heat exchange when the door is opened and including the opening of the door in the evolution of the process parameters.

7. The method according to any preceding claim, wherein the recommended future operation includes any one or more of the operations of:

- loading solid metal in the furnace,

- skimming dross on the surface of the molten metal,

- stirring the molten metal in the furnace,

- introducing a thermocouple (15d) in molten metal,

- changing the heat output of the gas burners (11 ),

- introducing chemicals in the furnace,

- introducing alloys in the furnace,

- stopping the melting process.

8. The method according to any preceding claim, wherein the information on the change over time of predicted process parameters transmitting to the operator interface (14) includes a plurality of recommendations for executing said future operations at recommended times up to an end of the melting process.

9. The method according to any preceding claim, wherein generating the recommended future operation and recommend timing thereof includes accessing a database (10) storing past melting process parameters and executing a learning algorithm on said past melting process parameters.

10. The method according to any preceding claim, wherein operator feedback on operations on the furnace are input into the operator interface and transmitted from the operator interface to a server of the control system.

1 1. The method according to the preceding claim, wherein operator feedback includes information on whether or not operational procedures were executed at the recommended times or postponed.

12. The method according to any preceding claim, wherein dynamic efficiency indicators, including a ratio of energy used to heat and melt the metal, over the energy provided to the furnace, is computed and outputted by the control system during said melting process of said batch of metal.

13. The method according to any preceding claim, wherein a dynamic defect detection indicator, including a ratio of undesirable and unknown energy contribution over the energy provided to the furnace is computed and outputted by the control system during said melting process of said batch of metal.

14. The method according to any preceding claim, wherein process performance indicators, including a final melting rate, or a final gas energy consumption, are calculated and outputted on-line by the control system before the end of said melting process of said batch of metal.

15. The method according to any preceding claim, wherein efficiency indicators, defect detection indicators and performance indicators computed and outputted by the control system are used to detect on a short-term basis, during said melting process of said batch of metal, any abnormality such as bad door sealing or unknown insulating ceramic breakage.

16. The method according to any preceding claim, wherein the efficiency indicators are used on many melting cycles to detect drifts of the furnace, such as a clogged heat-exchanger or an excessively eroded ceramic lining of the furnace.

17. The method according to claim 12, 15 or 16, wherein the efficiency indicators are used to compare various furnaces on a statistical basis or to train learning algorithms.

18. The method according to any preceding claim, wherein the amount of dross formation is estimated from a variation of emissivity parameters.

19. Control system (2) for a gas furnace for melting metal, adapted to carry out the process according to any one of the previous claims, comprising a melt optimization computing system (6) including:

- a melt optimization control program stored in a memory of the computing system and executable by a processor of the computing system, the melt optimization control program comprising an acquisition module configured to receive process parameters from the sensors, including at least one top wall temperature measurement and at least one molten metal temperature measurement, and a simulation module (18a) configured to calculate a physical heat exchange model in the furnace including radiation between walls and the metal, conduction though the metal and walls, and convection in the furnace chamber including calculating a temperature at a position in the furnace corresponding to said at least one temperature sensor, the melt optimization program configured to compare acquired process parameters with process parameters calculated by the numerical physical heat exchange model and to tune the model by adjusting at least one parameter thereof in order to reduce the difference between measured and calculated process parameter values; and

- an operator interface (14) comprising a display, configured to receive and display information outputted by the melt optimization program, including information on the change over time of predicted process parameters and at least one recommendation for executing a future operation in relation to the furnace at a recommended time.

20. Control system according to the preceding claim, wherein the operator interface is arranged to display a furnace status panel (16a) providing information on the status of process parameters and a recommendation panel (16b) providing information on the recommended future operations and the timing thereof.

21. Control system according to the preceding claim wherein the operator interface is arranged to display a control and further actions panel (16c) providing information on the next recommended operation and the timing thereof.

22. Control system according to either of the two directly preceding claims, wherein the furnace status panel comprises indicators, for instance in the form of gauges, of estimated predicted values including any one or more of:

- Normalized predicted melting rate (tonnes of solid load melted / hour);

- Normalized predicted gas consumption (normalized cubic meter of gas/ tonne of molten metal);

- Normalized predicted efficiency up to the current time (power effectively going into the metal, divided by total power); Cumulated heat loss from door openings.

23. Control system according to any preceding claim, further comprising a database (10) in which process parameters of past melting processes are stored, and a learning module (18b) receiving data from the database (10).

24. Control system according to any preceding claim configured to execute the method of any one of claims 1 to 18.

Description:
Method of Controlling a Gas Furnace for Melting Metal and Control System therefor

Field of the Invention

This invention relates to a method of controlling a gas furnace for melting metal and a control system therefor. The metal may in particular be aluminium.

Background of the Invention

The current world production of aluminium is around 120 million tpa (tonne per annum), with about 70 million tonnes produced by secondary production (remelting of aluminium). The secondary production includes melting of aluminium ingot and scrapes in furnaces followed by casting. The melting process carried out in melting furnaces is a costly process which consumes significant amounts of energy and generates significant amounts of C0 2 emissions (approx. 160 kg/tonne).

In a gas furnace, combustion with oxygen in the air of a fossil gas injected in gas burners located in the furnace supplies the enthalpy to the top surface of aluminium (e.g. scrap), previously loaded into the furnace. Thereby, the metal (a load) is heated and melts by both convective and radiative heat transfer mechanisms from the flame to the melt, but also indirectly by radiation from the heated ceramic of the furnace walls and vault. During a melting cycle, the temperature inside the furnace is regulated by adjusting the power output of the burners in such a way as to maintain a nearly constant temperature of the ceramic vault or the flue gas. A thermocouple can be introduced into the melt only once the solid load has been melted, in which case the melt temperature can be used to control the power of the burner.

During a melting cycle, the liquid metal and the solid scrapes can be oxidized to form a dross layer over the melt. The formation of a dross layer reduces the heat transfer from the burners to the metal, and consequently reduces the melting efficiency, but also generates heat by the oxidation reaction.

The efficiency and cost of the melting process thus highly depends on the behavior of the furnace operators. The melting process which includes several steps are performed in a certain time sequence that the operators determine based on their experience and feeling. This however introduces huge variations in the process and increases the costs of the operations, as well as energy consumption and C0 2 emissions.

Conventional methods actually implemented in industry for the evaluation of the efficiency of a batch cycle heating system are based on two global values: The normalized energy consumption, i.e., the total energy used to achieve the cycle normalized by the weight of the load,

• The normalized processing rate, i.e., the cycle duration normalized by the weight of the load.

The total cycle can be determined from the end of loading up to reaching a desirable temperature or finishing the desirable thermal cycle.

Example: A melting furnace consumed 20Ό00 kWh energy during 5 hours to melt 20 tons of solid aluminum ingot. In this case:

• The normalized gas consumption for a melting cycle is ~ 1000 [kWh/t]

• The normalized process rate (melting rate) is 4 [tonnes/h]

The efficiency of the furnace and a cycle can be assessed based on either normalized energy consumption or normalized process rate. A lower normalized gas consumption and a higher process rate indicate generally a higher thermal efficiency of the process cycle and furnace. Most industrially implemented thermal processes such as melting or heat treatment cycles are based on batch processing in industrial furnaces, i.e, each load batch undergoes this melting of thermal treatment on individual cycle. Each cycle sequence starts from loading the material into the furnace, followed by heating the system to increase and/or maintain the load temperature at a desirable temperature and to provide the latent heat of fusion if melting is concerned. The cycle can take some hours and ends by unloading the furnace.

Using only these two indicators has the following major limits:

• As the indicators can only be calculated at the end of the cycle, it is not possible to evaluate the efficiency during the cycle. In this case, it is impossible to monitor the evolution of the process efficiency during the cycle and to detect deviation from a good performance, malfunctioning or drift of the whole system during a cycle.

• These indicators represent the overall efficiency over the whole cycle. That means that one or two values summarize the effect of all the parameters. When deviations from optimal conditions during a cycle occur, it is therefore very difficult to separate the effects of each parameter and to determine the main cause of energy waste (such as door opening, load type, burner and heat exchanger condition,...).

• It is also impossible to identify within a cycle the period during which the efficiency was reduced.

• Since the indicator is influenced by many parameters, a lower efficiency cannot be directly related to a long-term drift or malfunctioning of the facility, e.g., associated with a deficiency of the heat exchanger, a deficiency of the burner, a reduced thickness of the insulating ceramics over time, etc. Summary of the Invention

In view of the foregoing, it is an object of this invention to provide a method and a control system for a gas furnace for melting metal that optimizes the melting process, in particular to reduce energy consumption and process time.

It would be advantageous to provide a method and a control system for a gas furnace for melting metal that provides information on optimal time sequences of process steps on a real- time basis to an operator.

It would be advantageous to provide a method and a control system for a gas furnace for melting metal that enables early detection of abnormalities and faults.

It would be advantageous to provide a method and a control system for a gas furnace for melting metal that provides recommendations for rapid corrective actions during production to operators.

It would be advantageous to provide a method and a control system for a gas furnace for melting metal that provides performance information such as performance statistics.

It would be advantageous to provide a method and a control system for a gas furnace for melting metal that enables retrofit implementation in existing gas furnaces.

Objects of this invention have been achieved by providing the gas furnace control system for melting metal according to claim 19, and a method according to claim 1.

The gas furnace control system and method according to the invention may advantageously take into account the various parameters that are susceptible to influence the efficiency of each cycle, in particular:

• Load amount / load type: the batches may contain different amount of load with various typologies. For example, for a melting furnace, various amount of solid scrapes with different geometry and typology may be loaded into the furnace.

• Positioning of the load in the furnace: for heat treatment furnaces and melting furnaces, it is very important how different loads / parts are positioned in the furnace. This influences the furnace efficiency.

• Environmental parameters: ambient temperature, air flow, initial load temperature, temperature of the combustion gas, temperature of cooling water, moisture, all these parameters may affect a process efficiency. • Human effects: many of the thermal processes are conducted by a human operator, who conducts the operations such as loading, positioning of load in the furnace, door opening, unloading, etc.

• Calorific value of the fuel or gas: the chemical composition of the gas/fuel is not a fixed value (natural gas may range from 11 to 15 kWh/kg) and can vary over time.

• Furnace conditions: refractory layer thickness and nature, usage, door sealing condition, burners regulation, heat source controller, efficiency of heat, efficiency heating power generated from electrical coil, electrical resistance or electrode in arc melting furnaces, etc.

The method and system according to the invention enables to monitor and to evaluate the efficiency of the cycle at any time during the cycle. This method can be used on-line (i.e. in real time) during the process. It can also be applied off-line to better analyze past process cycles. This can be very useful for statistical analyses of large number of cycles, to train predictive models based on artificial intelligence and machine learning. This can also be applied for the evaluation of the efficiency of a new facility or a facility after modification and to compare it with existing ones.

The method according to embodiments of the invention may be based on the calculation (computation) of various heat exchange contributions that occur in a furnace, based on a physical model that accounts for the major phenomena (conduction, radiation, convection, combustion). Such an approach allows to separate the heat flow contributions and losses.

An example of heat balance according to various heat flux contributions may for instance include the following power inputs and outputs:

• P1 = SUM of supplied heat power contributions = chemical enthalpy of the burning gas, exothermic oxidation of metal, electrical energy, heat of combustion of other materials such as organic coatings on metal scrapes

• P2 = Sum of the net heat power absorbed into the load = heat power entering into the load from radiation (heating elements, flame, furnace walls), combustion-convection (flame), induction, conduction from refractories, minus that leaving the load from radiation, conduction and convection. This power can be calculated by integration over space and time of the specific heat increase of the metal, including the latent heat of fusion if melting occurs.

• P3 = SUM of unknown heat power = heat power loss related to unknown refractory damage, bad door sealing, etc. • P4 = SUM of heat power loss through flue gas = heat power contribution escaping the furnace directly through the chimney or after heat exchangers

• P5 = SUM of heat power loss into the furnace refractory = all heat fluxes entering into furnace wall, roof, door and bottom refractory

• P6 = SUM of heat loss power through the door openings = all radiative and convective heat powers through the opened door

• P7 = SUM of heat power loss through any cooling system of the furnace

Which may define the following heating power balance:

• P1 (source) = P2 (desirable) + P3 (undesirable, unknown) + (P4 + ... + P7) (undesirable, calculated)

From these various contributions, one can define various criteria to better characterize a thermal heating system, in particular the efficiency, short term problems due to defects, and longer term drift, as discussed below:

Efficiency criterion for energy

May be determined by the ratio P2 / P1. Alternative definitions of the efficiency criterion (indicator) may be:

• time integration of P2 / time integration of P1 over the whole cycle

• dynamic integration of P2 up to a given time / duration from beginning of the cycle

• dynamic integration of P2 / dynamic time Integration of P1 up to a given time

These definitions of efficiency are fairly natural, since the purpose of a thermal process is heating the load to the desirable temperature (or state) compared to the total heat provided to achieve this goal. Thus P2 is the“useful” energy and this must be compared to P1. A fully efficient system should have P2 = P1.

Defect detection

May be determined by the ratio P3 / P1. Alternative definitions of the defect detection criterion (indicator) may be:

• time integration of P3 / time integration of P1 over the whole cycle

• dynamic integration of P3 up to a given time / duration from beginning of the cycle

• dynamic integration of P3 / dynamic time Integration of P1 up to a given time

Since the model takes into account all the contributions except P3, the difference between P1 and all the calculated contributions P2, P4, P5, P6, P7 gives access to P3. An increase of P3 means that the system is less efficient due to an unknown reason (bad door sealing, ceramic insulation defect, etc). Longer-term drift

This control parameter is an evaluation made over many cycles of the average energy efficiency criterion (indicator) defined above. The model uses values of a“perfect” furnace, with brand new ceramic insulation layers, well-functioning burner and heat exchangers, etc. Over time, clogging of the heat exchangers by fumes or thinning of the ceramics by interaction with the metal will gradually degrade the efficiency of the cycles. Making an average of P3/P1 over many cycles can detect these long-term drifts.

Disclosed herein is a method of controlling a metal melting process in a gas furnace with a control system, the gas furnace comprising at least one furnace chamber enclosed by walls and a door, gas burners configured to heat the metal loaded in the chamber, a flue gas exit and a plurality of sensors including at least one temperature sensor at a top wall (vault) of the furnace chamber and at least one temperature sensor to measure the metal temperature in its melted state, the method comprising:

a) receiving in the control system during the melting process, process parameter measurements from said sensors including at least one temperature sensor,

b) simulating, in the control system, physical heat exchange in the furnace with a numerical physical heat exchange model including radiation between walls and the metal, conduction though the metal and walls, and convection in the furnace chamber including calculating a temperature at a position in the furnace corresponding to said at least one temperature sensor,

c) comparing the data obtained from the physical heat exchange model with data obtained from the sensor measurements,

d) tuning at least one adjustable parameter of the physical heat exchange model to reduce the difference between compared data,

e) based on the tuned physical heat exchange model, calculating in the control system the change over time of the process parameters, including at least one process parameter which is not directly measured up to the end of the cycle,

f) transmitting to an operator interface of the control system, information on the change over time of predicted process parameters and at least one recommendation for executing a future operation in relation to the furnace at a recommended time.

In an advantageous embodiment, the process parameters measured and received by the control system and included in the numerical model further include a flue gas exit temperature. In an advantageous embodiment, the process parameters measured and received by the control system and included in the numerical model further include a burner gas flow rate.

In an advantageous embodiment, the process parameters measured and received by the control system and included in the numerical model further include a burner gas inlet temperature.

In an advantageous embodiment, the process parameters measured and received by the control system and included in the numerical model further include an air inlet temperature.

In an advantageous embodiment, process parameters measured and received by the control system and included in the numerical model further include any one or more of:

- a chemical composition of at least one component (such as C0 2 ) of the flue gas,

- a gas pressure inside the furnace chamber,

- a furnace bottom wall temperature.

In an advantageous embodiment, the simulated physical heat exchange model includes calculating in the control system the heat exchange when the door is opened and including the opening of the door in the evolution of the process parameters.

In an advantageous embodiment, the recommended future operation includes any one or more of the operations of:

- loading solid metal into the furnace,

- skimming dross on the surface of the molten metal,

- stirring the molten metal in the furnace,

- introducing a thermocouple in the molten metal,

- changing the heat output of the gas burners,

- introducing chemicals into the furnace,

- introducing alloying elements into the molten metal,

- stopping the melting process.

In an advantageous embodiment, the information on the change over time of predicted process parameters transmitting to the operator interface includes a plurality of recommendations for executing said future operations at recommended times up to an end of the melting process. In an advantageous embodiment, the step of generating the recommended future operation and recommend timing thereof includes accessing a database storing past melting process parameters and executing a learning algorithm on said past melting process parameters.

In an advantageous embodiment, operator feedback on operations on the furnace are input into the operator interface and transmitted from the operator interface to a server of the control system.

In an advantageous embodiment, operator feedback includes information on whether or not operational procedures were executed at the recommended times or postponed.

In an advantageous embodiment, dynamic efficiency indicators such as a ratio of energy used to heat and melt the metal, over the energy provided to the furnace, is computed and outputted by the control system during said melting process of said batch of metal.

In an advantageous embodiment, a dynamic defect detection indicator, such as a ratio of undesirable and unknown energy contribution over the energy provided to the furnace is computed and outputted by the control system during said melting process of said batch of metal.

In an advantageous embodiment, process performance indicators, such as a final melting rate, or a final gas energy consumption, are calculated and outputted on-line by the control system before the end of said melting process of said batch of metal.

In an advantageous embodiment, the efficiency indicators, defect detection indicators and performance indicators computed and outputted by the control system may be used to detect on a short-term basis, during said melting process of said batch of metal, any abnormality such as bad door sealing or unknown insulating ceramic breakage.

In an advantageous embodiment, the efficiency indicators may be used on many melting cycles (i.e. many batches) to detect drifts of the furnace, such as a clogged heat-exchanger or an excessively eroded ceramic lining of the furnace.

In an advantageous embodiment, the efficiency indicators may be used to compare various furnaces on a statistical basis. In an advantageous embodiment, the amount of dross formation may be estimated from a variation of emissivity parameters.

Also disclosed herein is a control system for a gas furnace for melting metal, adapted to carry out the process according to any one of the previous claims, comprising a melt optimization computing system including:

- a melt optimization control program stored in a memory of the computing system and executable by a processor of the computing system, the melt optimization control program comprising an acquisition module configured to receive process parameters from the sensors, including at least one vault temperature measurement and at least one molten metal temperature measurement, and a simulation module configured to calculate a physical heat exchange model in the furnace including radiation between walls and the metal, conduction though the metal and walls, and convection in the furnace chamber including calculating a temperature at a position in the furnace corresponding to said at least one temperature sensor, the melt optimization program configured to compare acquired process parameters with process parameters calculated by the numerical physical heat exchange model and to tune the model by adjusting at least one parameter thereof in order to reduce the difference between measured and calculated process parameter values; and

- an operator interface comprising a display, configured to receive and display information outputted by the melt optimization program, including information on the change over time of predicted process parameters and at least one recommendation for executing a future operation in relation to the furnace at a recommended time.

In an advantageous embodiment, the operator interface is arranged to display a furnace status panel providing information on the status of process parameters and a recommendation panel providing information on the recommended future operations and the timing thereof.

In an advantageous embodiment, the operator interface is arranged to display a control and further actions panel providing information on the next recommended operation and the timing thereof.

In an advantageous embodiment, the furnace status panel comprises indicators, for instance in the form of gauges, of estimated predicted values including any one or more of:

Normalized predicted melting rate (tonnes of solid load melted / hour);

Normalized predicted gas consumption (normalized cubic meter of gas/ tonne of molten metal); Normalized predicted efficiency up to the current time (power effectively going into the metal, divided by total power, and integration over time of such a ratio);

Cumulated heat loss from door openings.

In an advantageous embodiment, the control system further comprises a database in which process parameters of past melting processes are stored, and a learning module receiving data from the database.

Further objects and advantageous aspects of the invention will be apparent from the claims, and from the following detailed description and accompanying figures.

Brief Description of the drawings

The invention will now be described with reference to the accompanying drawings, which by way of example illustrate the present invention and in which:

Figure 1a is a schematic simplified cross-sectional diagram of a typical gas furnace for melting metal, for instance for melting aluminium scrapes and ingots;

Figure 1 b is a schematic simplified diagram of a gas furnace for melting metal used to describe an example of a numerical model used in a heat exchange model according to an embodiment of the invention;

Figure 1 c illustrates viewing factors calculation between two surfaces used in a heat exchange model according to an embodiment of the invention;

Figures 2a and 2b are schematic simplified overview diagrams of implementation of a control system for a gas furnace for melting metal according to an embodiment of the invention;

Figure 2c shows an example of an implementation structure of a control system for a gas furnace for melting metal according to an embodiment of the invention,

Figure 3 is a schematic simplified diagram of a gas furnace for melting metal connected to a control system for the gas furnace according to an embodiment of the invention;

Figure 4 illustrates schematically examples of data collected by a control system for a gas furnace for melting metal according to an embodiment of the invention;

Figure 5 illustrates schematically a graph of measurements during a melting process collected by a control system for a gas furnace for melting metal according to an embodiment of the invention;

Figure 6a illustrates a schematic graphical operator interface of a control system for a gas furnace for melting metal according to embodiments of the invention;

Figure 6b shows an example of the graphical interface showing a timeline indicating actions that may be selected by an operator; Figures 6c and 6d illustrate pop-up windows of the graphical operator interface displaying information on the loading actions (fig. 6c) and requesting operator input on the actions carried out (fig. 6d);

Figure 7 is a flowchart illustrating steps of a metal melting process with a gas furnace according to an embodiment of the invention.

Figure 8 illustrates schematically a control system for a gas furnace for melting metal according to an embodiment of the invention, showing an example of data and workflows in the system structure.

Figure 9 illustrates a workflow for a solve (mode) function of the control system of figure 8.

Detailed description of embodiments of the invention

Referring to the figures, a gas furnace system for melting metal 1 , for instance for melting aluminium, comprises a gas furnace 3 and a control system 2. The gas furnace 3 comprises a furnace chamber 9 enclosed by walls 5 and a door 7, gas burners 1 1 configured to heat the metal 1 in the chamber, a flue gas exit 13 and a plurality of sensors 15. The furnace may further comprise a stirrer within the chamber configured to stir the metal in its melted state, in particular an electromagnetically driven stirrer. The door 7 is typically a sliding door through which the metal is loaded. The vault and lateral walls of the furnace are lined with heat resistant ceramic material and the gas burners may typically be mounted on the sidewalls of the chamber above the level of the liquid metal at full capacity. The gas fumes (combustion product) are recovered through the flue gas exit 13 which may be connected to a heat exchanger in order to recover heat energy.

The various heat and mass transfers occurring within a furnace are illustrated schematically in figure 1 , such heat transfers including radiation among the walls and metal 1 , convection within the chamber, conduction through the metal and the walls, convection and radiation between the flame and walls and metal, and when the door is opened, heat loss by radiation and convection through the opening.

The sensors 15 may include :

a- Sensor 15a indicating that the furnace door is fully open

b- Sensor 15b indicating that the furnace door is fully closed

c- Sensor to measure the inner furnace pressure

d- Thermocouple 15d to measure the metal temperature

e- Thermocouple 15e to measure the temperature on a top side of the chamber (for instance between 3 to 7 thermocouples depending on the furnace)

f- Thermocouple 15f to measure the temperature on the bottom side of the furnace (to monitor the refractory damage and metal infiltration risks)

g- Thermocouple 15g to measure the temperature of the exit flue gas h- Thermocouple to measure the flue gas temperature after the heat exchanger i- Thermocouple to measure the flue gas temperature before the heat exchanger k- Thermocouple to measure the inlet gas temperature

I- Thermocouple to measure the inlet air temperature

m- Gas/fuel flow rate measurements

n- Air/oxygen flow rate measurements

o- Optionally, sensor indicating furnace tilting and furnace position

p- Sensor indicating the presence of flame

q- Optionally, sensors indicating the furnace weight (on hydraulic pistons)

r- Electromagnetic stirrer power, frequency

s- Optionally, sensor to analyze the chemical composition of the flue gas

The gas furnace control system 2 comprises a furnace control unit comprising inputs connected to the sensors 15 and outputs connected to automated furnace actuation devices such as the stirrer induction coil (if available), door actuation motor (if available), fuel and air control valves to control the gas burner heat output, flue exit vent actuators (if available) and other actionable motors and actuators controlling elements of the furnace.

The furnace control unit further comprises a data output connected to a melt optimization computing system 6. The melt optimization computing system may either be integrated into the furnace control unit, or provided as a computing system separate from the furnace control unit. In the illustrated embodiments, the computing system is illustrated as a separate system from the furnace control unit. In the latter embodiment, the melt optimization computing system 6 may be in the form of a local computing system installed at the site of the furnace, or be in the form of a distributed computing system remote from the site of the furnace and connected to the furnace control unit 4 via a communication network 12, for instance the internet, or a combination of both a local computing system and a remote computing system.

The control system 2 comprises a melt optimization control program installed and executable in the melt optimization computing system 6, and a database 10 including data stored therein from previous melting cycles.

The data of a melting cycle, present or past, may include data from some or all of the sensors 15 and operational data, which may include:

• Type of metal load (scrapes types and morphology) and loading time sequence. These data can possibly be obtained through another program, which calculates the loading composition from available scrapes, or the operator can manually provide these data to the program

• Time of the first loading

• Time to start melting

• Time of transfer (after melting, the liquid metal may be transferred to a holding furnace for further treatment)

The control system 2 further comprises an operator interface 14 for displaying information on the melting process and optionally for allowing the operator to input operational data on the melting process and instructions for operation and control of the furnace. The operational data on the melting process, such as the start or end times, loading of metal into the furnace chamber, operation of the stirrer, or other operations controlled manually by the operator during the melting process may be transmitted to the computing system for analysis by the melt optimization program and/or for storage in the database 10. The instructions for operation and control of the furnace may be inputted into the furnace control unit to control the automated furnace actuation devices.

The operator interface 14, in an embodiment illustrated in figure 6, may include a furnace status panel 16a, a recommendation panel 16b, and a control and further actions panel 16c.

The furnace status panel 16a may comprise various information on the status of the melting process, such as the position of the door (open or closed), the start time, the run time, the temperature of the melt, the estimated fuel or energy consumption, the percentage of melt vs solid (solid fraction), and other status parameters useful for the operator to know. As shown in figure 6b, the furnace status panel 16a may also comprise indicators, which may for instance be conveniently provided on the graphical operator interface in the form of gauges, of estimated predicted values such as:

1- Normalized predicted melting rate (tonnes of solid load melted / hour)

2- Normalized predicted gas consumption (normalized cubic meter of gas/ tonne of molten metal)

3- Normalized predicted efficiency up to the current time (power effectively going into the metal, divided by total power)

4- Cumulated heat loss from door openings

The melt optimization computing system according to the present invention enables not only an accurate prediction of these indicators, but also a data-driven on-line evaluation of these parameters. As described herein, optimal values for these indicators depend on various parameters and operational conditions, such as furnace initial temperature and initial liquid remaining in the furnace, load type and loading sequence, and other environmental conditions. The optimal (e.g. of blue colour), standard (e.g. of yellow color) and underperforming (e.g. of red color) value zones for these indicators may be obtained based on a data analytics, in particular machine learning or artificial intelligence systems, based on a plurality of accumulated past production data as well as on simulation data obtained from the physical model. This enables furnace operators and engineers to have a clear evaluation during operation of the melting cycle, based on a combination of data-driven predictive results provided by the physical model and on-line evaluation of the conditions based on artificial intelligent and big data analysis technologies.

The recommendation panel 16b includes various information up to the current time, e.g., predicted and measured vault temperatures, predicted and measured melt temperature, as well as predictions for the future operations together with predicted vault and melt temperatures up to the end of the melting cycle. This information displayed along a timeline in the melting process shows to the operator when are the optimal times to execute or to control various operations, such as when to actuate the stirrer, when to introduce the thermocouple into the melt, when to skim the melt, when to adjust the composition of the melt or when the melting cycle has reached the desired melt temperature.

The control and further actions panel 16c may summarize the timeline indications shown in panel 16b, with information on the remaining time up to the next action and following ones.

The melt optimization control program comprises a data acquisition module configured to receive data from at least some of the sensors 15, including temperature sensors, a simulation module that is configured to calculate a physical model of heat exchange in the furnace, and a learning algorithm. The calculation of the physical model of heat exchange in the furnace may be used to simulate the melting process up to the end of the melting cycle, in particular to estimate parameters such as the solid fraction, melt temperature, and fuel or energy consumption over time. The physical model is adjusted during the melt cycle based on physical measurements of the present state provided by the sensors 15 and optionally by operator inputs, in order to refine the predictions of parameters as the melting cycle progresses based on past information. The learning algorithm processes data from previous melt cycles from the database 10 to determine the efficiency or effectiveness of past actions and based thereon to adjust recommendations for future actions. The prediction and learning algorithms process and output time-based recommendations on actions to be undertaken during the melting process for the operator to execute manually or for automatic execution by the furnace control unit 4. The data acquisition module captures the measurements of at least some of the sensors 15, especially temperatures, during the melting cycle continuously or intermittently, for instance at predetermined intervals. In parallel to data acquisition, the physical model of heat exchange in the furnace, over a future span of time, is calculated by the simulation module, based on a calculation of heat exchange between the gas burners and the furnace chamber, mutual radiation between the various surfaces, thermal conduction within the various domains, convection within the metal and the air, and oxidation at the top surface of the metal load. The model also takes into account several operations such as: door open or closed, addition of solid or liquid metal, liquid melt stirring, surface oxidation (exothermic) and surface oxide skimming.

The predicted data at the location of sensors from the physical model is compared with measured data from the sensors 15 and allows to adjust some of the parameters such as heat transfer coefficients between the burner and the various surfaces. The adjustments may be made at each data acquisition time or at intervals spanning a plurality of data acquisitions. As the melting cycle progresses, the adjustment becomes more and more accurate as the number of comparison data increases.

Advantageously, the physical model is based on heat exchanges and mass transfer within the furnace and metal load that encapsulates the most important physical phenomena (radiation, conduction, convection). While taking into account the geometry of the furnace, its implementation into a numerical code is made in such an efficient way that the simulation may be calculated with a CPU time that is faster than one step of data acquisition, which may occur typically between every 1 to 10 seconds, and thus also much faster than the whole melting cycle which may typically be a few hours. An example of a one dimensional or two dimensional physical model used by the simulation module according to an embodiment of the invention is described hereafter.

The physical model, on which the invention is based, is sufficiently fast to allow numerical simulation of the whole melting process over a time that is faster than the data acquisition system. Yet, it includes the main physical phenomena involved in the process: combustion of the gas and heat exchange of the fumes, heat exchange between the burner and the various surfaces (furnace walls and vault, metal surface), mutual radiation between them, thermal conduction in the various media below these surfaces, convection in the metallic melt, (exothermic) oxidation at the melt surface, phase transformation and compaction within the metal load during melting. The model accounts for the geometry of the furnace, in particular of the non-parallelepiped pool containing the metal. At the core of the model is the mutual radiation between the various surfaces which is calculated according to the laws of radiation with appropriate viewing factors. Opening of the door, which influences mutual radiation of the furnace and metal surfaces, is accounted for.

The implementation of the model is made according to an implicit time-stepping scheme which removes the stability condition of explicit schemes and allows to have time steps that follow accurately those of the data acquisition system.

Temperature-dependent materials properties controlling radiation (emissivity) and conduction (thermal conductivity) are taken from the literature, while those which are not as well known (e.g. heat transfer coefficients from the burner flames to the various surfaces) can be adjusted on-line during data acquisition with a retro-fit method.

An important advantage of the method of the invention is a unique combination of data acquisition and simplified modelling that is reliable and fast. As the numerical model is sufficiently fast, yet physically based, this allows to make some inverse modelling on line during the melting cycle in order to optimize some of the not so well-known parameters of the process. As this fine tuning of parameters is done during data acquisition, this allows to become increasingly predictive in the future of the melting process. The method thus accumulates during one melting cycle the past experience to predict the future and to minimize in this way the melting cycle, the consumed energy, the amount of C0 2 release. It becomes a guide to the operator for what is happening in the furnace, without opening the front door, and for when making necessary operations such as reloading some more metal and scrapes, skimming the oxide skin, stirring the melted metal, inserting the thermocouple into the melt without breaking it or terminating the melting cycling within the right time and without excess superheat.

The control system according to the invention allows to determine the optimal process sequences and parameters evolving over time to minimize the total cycle time, total energy consumption and dross generation.

An example of a numerical model to simulate a physical model of heat exchanges in a melting furnace according to an embodiment of the invention is described below with reference to figures 1 a, 1 b, 1 c and 3 in particular. The geometry of the furnace being modelled is substantially parallepipedic, however other furnace geometries may be considered and modelled. In order to optimize melting or remelting furnaces, in particular for the Al industry, a numerical model based on sound physical aspects, has been developed. It is fast and efficient and directly coupled with in situ data acquisition, it allows reverse engineering and optimisation to be made on line.

The model takes into account that the furnace comprises a bottom portion containing the metallic melt 1 , a moving front door 7 which allows to feed the metal into the furnace in one or several batches, an open space above the metal further delimited by a lateral surface (side wall 5c) and a vault (top wall 5a). Temperature T vau u is measured in the top wall, and optionally in other positions on the lateral surfaces. Heat is provided by combustion of a gas through one or several gas burners 11. The gas flow G fiow and gas temperature T gas are measured. In many furnaces, fumes resulting from combustion exit though a heat exchanger 17, allowing to preheat fresh air flow A ho „ to an air temperature T air which is also measured. The temperature of the metal T metai can be measured with a thermocouple 15d which can be inserted into the melt once it is sufficiently liquid. The model includes three main components:

• A model of combustion

• A model of heat transfers

• A coupling with on-line data acquisition

Combustion model

This part of the model describes the power input from the reactants and the losses from the combustion products during the operation of the furnace.

Knowing the excess air ratio with respect to the gas l (l = 1 corresponds to no excess) and assuming full combustion of the gas, the stoichiometric reaction of combustion for a gas such as methane may be calculated according to:

CH 4 + 2l (0 2 + 3.76N 2 ) = C0 2 + 2H 2 0 + (l - 1 ) 0 2 + 2l (3.76N 2 ) (Eq. 1 )

The power exchange Q ' burn between the combustion products and the furnace can be readily obtained from the equation:

where ή r and n r are the molar fluxes of products and reactants, respectively, while Iiz(t r ) and h (J r ) are the molar enthalpy of formation of each component, taken at their corresponding temperatures, T p and T r . In the example of Eq. (1 ), the reactants are CH 4 , 0 2 , and N 2 , while the products are C0 2 , H 2 0, 0 2 and N 2 . The enthalpy of formation of the various chemical species h{ (7)) (/ ' = rfor reactants and / = p for products), are calculated from the enthalpy of formation h{ ( T 0 ) at a reference temperature

T 0 , according to the equation: (Eq. 3) where c™; is the temperature-dependent specific heat of the considered species / ' , per unit mole. The temperatures 7) of the various species in Fig. 1 b are: T air for air, T gas for gas, and T fume for the exit fumes.

The molar fluxes of the various species are calculated from the measured flow rates of gas G fiow and air 7 , while their temperatures T air , T gas are directly estimated from temperature sensors in the respective air and gas conduits. The fraction of fumes going through the heat exchanger 17 is accounted for in the heat balance.

The obtained heat power exchange Q burn is then used as an input in the numerical simulation of conduction and radiation heat transfers.

Heat transfer model

Contribution from the burners

The total heat power exchange Q burn is attributed to the various furnace surfaces according to adjustable parameters a,(T f , T flam e). In the case of a parallelepipedic furnace, the index“j” can take 6 values corresponding to the metal surface, the vault, the door and the three other lateral surfaces. A surface (j) has a temperature T at a given time f, and T flame is the temperature of the flame of the burner. The parameters a,(T f , T flam e) can be optimized during the process by inverse modelling for reverse engineering) in order to best fit the measured values (see next section).

Taking into account these coefficients a,(T f , T flam e), the total flux Q urn [W] received by surface S j is given by: (Eq. 4) or a flux density en by:

Radiation among the furnace surfaces At the core of the heat transfer model is a calculation of radiation between the various surfaces (melt, vault, door, lateral walls), radiation being the dominant heat transfer process in a remelting furnace for Al alloys. The model uses a net radiation method, assuming each surface (j) to be grey, with an emissivity e/7}), and diffuse. In order to speed up the

calculations, the average viewing factors being surface (j) and surface (k) (see figure 1c) are calculated either by analytical integration for a parallelepipedic furnace or by numerical integration for a more complex furnace shape. It is recalled that the viewing factors F jk are given by:

where (¾,/¾) are two parametric parameters of surface Sy (e.g, (x,y) for Si in Fig. 2), ( ,¾) are two parametric parameters of surface S k (e.g, (y’,z) for S 2 in Fig. 1 c), / ( ¾, ,¾) is the distance separating the two surface elements, and cos Q and cos 6 k are the two cosines of the angles between r^ and the outward-pointing normal to each surface (see Fig. 1 c). Once these geometrical viewing factors are known, the“radiosity” method allows to calculate the heat exchanged by radiation between the various surfaces. Adding the burner energy input, this allows to obtain the heat flux condition at each surface, at each time step, knowing the surface temperatures Tf of each surface. One has the boundary condition for a surface (j):

where K, is the thermal conductivity of the material below surface (j), the summation is carried out over the N s surfaces of the furnace and melt (6 in the case of a parallelepiped), s B is Stefan-Boltzmann’s constant, and q T s k is the outcoming radiative flux of facet (k). These fluxes are found by solving the system of equations:

*t \4

A MTH = B e t Ί " (Eq. 8) where the matrix elements A^are given by:

A jk = 1 Wj = k ; A jk = -(1 - ej)F Jk if j ¹k (Eq. 9)

This situation corresponds to a closed door of the furnace, i.e., the cavity of the furnace corresponds to a closed system for radiation. When the door is open, its radiation is considered separately from the other surfaces of the furnace and the corresponding opening is replaced by a medium of emissivity 1 and a temperature equal to the ambient. Heat diffusion from the surfaces

Knowing the heat flow contribution from the burner at each surface (Eq. (5)) and their mutual radiation by solving Eqs. (8), the boundary conditions of Eqs. (7) allow then to solve the heat diffusion in each medium below the N s surfaces: (Eq. 10) where z is the coordinate parameter in the material perpendicular to the surface (j), T/z,ί) is the local temperature at location z-at time t, and h j is the volumetric enthalpy. For the ceramics below the surfaces of the furnace, the volumetric enthalpy is simply given by:

where (pc p ) is the volumetric specific heat of the material. For the metal itself, the enthalpy accounts also for the latent heat of fusion L f\

where g ( · is the local volume fraction of liquid metal. Equation (10) (Eq. 10) is solved using an implicit time stepping (also called backward Euler) scheme with the boundary condition at the surface given by Eq. (7) and a boundary condition at the outer surfaces of the furnace given by an imposed temperature.

It may be noted that the diffusion equation may be solved below each surface, taking into account the various thicknesses of different ceramics, with their corresponding properties. For the bottom surface (metal), the diffusion equation within the metal and that for the ceramics below are coupled via a heat transfer coefficient at the metal-ceramic interface.

Coupling with data acquisition

Since the radiation and diffusion calculations are very fast and efficient, with a time step that can be set equal to that of the data acquisition system, the temperatures 7' 7, (r fc ,f) measured at the N tc locations r^ of the thermocouples can be compared with calculated values at the same locations and time Since these later values are function of some parameters, in particular the coefficients a, giving the distribution of the burner power, an inverse algorithm can be used to optimize their choice. More precisely, one tries to minimize the function:

where N t is the number of time steps used in the optimisation procedure. This minimum is given by:

where the X k tc,p are the sensitivity coefficients of temperature with respect to variations of parameters <¾:

These coefficients can be calculated numerically by small variations of the parameters <¾. Turning now to the overall melting process, a melting cycle may comprise the following operations:

Furnace cleaning

Depending on the operational conditions and the amount of deposited dross on the furnace walls and furnace bottom, a cleaning step may be applied before a melting cycle. This cleaning step is typically required at least twice per week for a furnace in service. In the case that the subsequent melting cycle includes materials with incompatible chemical compositions from materials of the previous melting cycle, a full cleaning procedure may be applied to remove all residual metal from the furnace chamber. Otherwise, it is common that a new melting cycle includes some residuals (for instance about two tonnes of liquid residual metal) from the previous cycle.

Loading

A typical melting cycle starts with loading. Scrapes and ingot and some master alloys can be loaded directly into the furnace. If the principal metal is aluminium, master alloys may for instance include Cu and Zn. In the case of using scrape with low compacity, the loading cannot usually be carried out in one step. In this case the loading is split into several load packages (for instance between 2 to 6). Each load package will be charged into the furnace during the melting cycle, as soon as free space in the furnace allows for further loading. Load packages may contain liquid aluminium obtained from smelting or recycling process. There are several safety issues regarding the loading procedure. For example it is forbidden to load big ingot and scrapes into the molten pool (to avoid splashes). The loading of scrapes which are not pre-heated also requires several precautions because humidity may create water vapor explosions.

In order to optimize the melting process, a global objective is to load metal into the furnace as soon as possible, as much as possible, and as quickly as possible. The simulation program according to embodiments of the invention predicts the volume change of scrape during melting and thus predicts the time when there is sufficient space inside the furnace allowing for further loading. This helps to minimize the number of door openings, and thus unnecessary energy loss.

Stirring

Stirring can be applied during a cycle in order to homogenise the temperature inside the furnace. A stirring is applicable and efficient only if there is enough liquid metal in the furnace.

In order to optimize the melting process, the simulation program according to embodiments of the invention predicts whether or not a stirring is required during a melting cycle and what the optimal time is for stirring based on temperature gradients in the metal load, calculated oxide quantities and calculated coefficients of heat absorption.

Thermocouple introduction

The liquid metal temperature is one of the most important parameters of the melting process. In order to measure this liquid metal temperature, a thermocouple is introduced into the molten metal pool. The thermocouple should not be introduced if the amount of remaining solid scrapes is too large due to a risk of breaking the thermocouple. After introducing this thermocouple, the gas burners power can be regulated based on the measured liquid metal temperature.

In order to optimize the melting process, it is desirable to introduce the thermocouple as soon as possible. The simulation program according to embodiments of the invention predicts the optimal time for thermocouple introduction based on the estimation of average metal temperature and liquid fraction over time.

Skimming

Skimming is applied to remove the oxides that form at the upper surface of the molten pool. The formation of oxides at the upper surface of the pool generates heat but has also an important impact on the heat transfer from burners to metal. Skimming has therefore a significant impact on the furnace efficiency. A normal melting cycle includes at least one skimming step. It is normally performed when the metal is fully molten. In some furnaces, depending on the type of scrapes, skimming might have to be avoided. In the case of formation of significant oxide during the cycle, an early skimming may be applied to remove this oxide layer at the middle of a melting cycle. In a reverberatory gas/fuel furnace for aluminium melting, it is important to optimize the timing for removal of the dross layer on the metal melt by skimming. The dross layer is made of aluminium oxide which has a high melting point. The dross layer will not melt further, but acts as a heat insulator. If it is allowed to grow too thick, it will insulate the metal melt from the burner flame. The dross will be more heated and more metal will be oxidized. Note that the formation of the oxide skin is exothermic, i.e., generates heat which is included in the physical model.

In order to optimize the melting process, skimming should thus be applied when a significant amount of oxide has formed and decreases substantially the heat transfer from burners to metal load. In addition, skimming is only possible if there is significant amount of liquid available. The simulation program according to embodiments of the invention predicts the optimal time for skimming based on an estimation of the fraction of liquid metal over time.

Sampling and alloying

In some metal melting processes, a sample may be taken from the furnace to determine the chemical composition of the melt. Master alloys may be added to the melt to obtain the desired chemical composition.

End of cycle

The melting cycle ends when the liquid metal has achieved the desired temperature and chemical composition.

In order to optimize the melting process, increasing the melt temperature above the desired temperature (typically 720°C for aluminium) should be avoided for several reasons: excess of energy consumption, increase of metal oxidation, longer melting cycle, and evaporation of some solute elements (e.g. Zn). Simulation of the melting process with the simulation program according to embodiments of the invention in addition to data acquisition when the thermocouple is inserted into the melt allows to achieve the optimal time-temperature cycle.

A melting process according to an embodiment of the invention is illustrated in figure 7. The melting process may include the following steps:

1- Collecting data from sensors 15 - examples of data measured during a melting cycle are illustrated in figures 4 and 5 2- Calculating inlet power of the burners with fuel, gas, air and oxygen flow rates and their respective inlet temperatures

3- Calculating flue gas chemical composition and flow rates with using data fuel/gas/air /oxygen flow rates with respect to combustion reaction

4- Obtaining flue gas temperature from the sensors

5- Calculating in real-time the gas burners power based on gas/fuel/air/oxygen flow rates, temperatures, using numerical description of combustion

6- Calculating power loss through flue gas using flue gas flow rate and temperature

7- Numerical simulation of the furnace from the beginning of the melting cycle up to the actual time based on the physical model described hereinabove

8- Adjusting the physical model on a real time basis by comparing simulated and measured values such as furnace vault temperature to tune some heat transfer coefficients and some adjustable parameters in the model on a real time basis

9- Determining some process values such as metal temperature, solid, liquid and oxide quantities with the tuned physical model

10- Determining metal load volume and free available space in furnace with the tuned physical model

1 1- Determining the heat flux (heat transfer coefficients) to the metal load and to the various surfaces of the furnace, with the tuned physical model

12- Calculating heat diffusion in the various walls of the furnace as well as in the metal load (in this case, taking into account mass transfer, e.g. creep of solid aluminium and flow of liquid aluminium)

13- Performing simulations for future times during the melting cycle using the tuned model

14- Predicting the gas / air flow rates evolution for future times with an algorithm with tuned burners power to achieve / maintain either vault or metal desirable temperatures ( for example 1 100°C for vault before thermocouple introduction into the melt and 750°C for metal temperature.)

15- Predicting the evolution of solid, liquid, oxide up to the end of the melting cycle

16- Determining the optimal times for split loading (based on the prediction of free space evolution in the furnace)

17- Determining the optimal time for thermocouple introduction (based on obtaining over 90% liquid fraction or average temperature of 600°C in the metal)

18- Determining the optimal time for stirring (based on temperature gradient in the load and coefficient of absorption)

19- Determining the optimal time for oxide skimming, based on oxide quantities and coefficient absorption and liquid metal quantity 20- Repeating some or all of the above steps at predetermined or regular intervals, for instance every 10 to 60 seconds, in order to update values

Figure 5 shows a time-sequence of temperatures actually measured in a furnace by the data- acquisition system and simulated with the physical model. It shows the temperatures of the furnace vault and of the metal bath. Please note that the temperature of the bath is measured only when the thermocouple is plunged into the melt. The events associated with the opening of the furnace door are clearly visible. During the door opening, the thermocouple is removed from the melt, thus explaining the fast temperature decrease, in order to enable mechanical stirring.

EXAMPLE

An example of implementation and configuration of the melt optimization computing system 2c is described below.

1. Melt optimization computing system structure

The melt optimization computing system includes following main components as schematically illustrated in figure 2c:

• A user interface

• A data acquisition module

• A common database

• A Process Optimizer Module (POM) that constitutes the melt optimization control program.

The Process Optimizer Module (POM), user interface, and data acquisition modules communicate directly and independently to the common database.

The global workflow in the melt optimization computing system may include the following steps which may be repeated regularly, for instance within 10 to 20 second time intervals:

Step l : Data acquisition: The data acquisition module obtains a new set of data from sensors, IT network or operator interface and stores these data in the common database.

Step 2: Data processing by POM: POM uses this new set of data including all the previous ones up to the current time, to update predictions and recommendation. The obtained results are stored in the common database.

Step 3: Interface update: The user interface module communicates these updated recommendations to user / operator. 2. Data structure in the database

The data structure of the system may be as follows in this implementation example. That may include collected data related to previous melting cycles, for instance at least two previous cycles (if available), as well as the data of the current melting cycle up to the latest data acquisition time (current process time).

2.1 Sensor data

These data include a selected list of process data obtained from the furnace control unit. The values are either measured by sensors or other values calculated by the furnace control unit. This data structure is updated preferably at each data acquisition time step (for instance typically each 10 to 20 seconds).

2.2 Action data

This data structure includes information related to start and end times of the set of actions such as loading, stirring, skimming, alloying, as carried out by the operator during the process.

2.3 Load data

This data structure includes loading data corresponding to each melting cycle. The loading material type, loading weight and loading time are stored in this data structure.

2.4 POM internal variables data

This data structure contains geometrical information, thermophysical properties of the materials implemented in the furnace, numerical parameters and initial conditions that are required to start a simulation for a dedicated furnace.

2.5 POM results data

The results obtained from POM are stored in this data structure. These data includes calculated temperature field and predicted optimal times for actions, The user interface communicates these data to the operator.

3. Process Optimizer Module (POM)

The POM structure and flow of data is depicted in figure 2c and is detailed in figure 8. It includes the main components as follows:

3.1 Data Analytics The data analytics component checks that the various data coming from the database are consistent as input for the physical model. The data analytics component integrates a learning module to detect abnormalities and drifts.

3.2 Physical Model

Based on the data available in the database up to the current time, the physical model enables POM to obtain accurate temperature field inside the furnace for the past and future times. This model includes conduction, convection, radiation, combustion and oxidation models to calculate the future temperature field in the furnace based on predicted optimal actions.

3.3 Inverse modeling

The inverse modeling component, compares on-line the POM results data and sensor data up to the current time. That allows to adjust on-line some of the POM internal variable data to reduce the differences between some calculated and measured values.

3.4 Planification module

The planification module predicts the optimal times for actions such as thermocouple insertion into the melt, stirring, skimming, alloying, from the current time up to the end of the cycle. The planification module is based on several components: internal rules of the company operating the furnace, data obtained from the physical model, analysis of past cycles based on machine learning algorithms.

An example of detailed implementation of the POM is illustrated in figure 8. It includes the following main steps:

Step 1 : Obtaining data from the database: Input Sensor data, Load data, Action data and Configuration data from the database.

Step 2: Update Action (UA): Update actions data according to the new set of Sensor data.

Step 3: Update Load (UL): Update the load according to the updated action data.

Step 4: Solve steady-state: Obtain an arbitrary steady state temperature distribution based on configuration data.

Step 5: Store temperature field: Keep the temperature field at the end of Step 4. This initial condition is required for future calculations as well as for the implementation of the inverse method which tunes some of the physical model variables.

Step 6: Solve past & Tune adjustable variables: Obtaining the temperature distribution for all past times up to the current time according to the available data for the past times, updated actions and load data. At this step, the optimization method is used to tune several of the physical model variables, based on inverse modelling to have the best agreement between past measured and calculated temperatures Step 7: Solve future & plan action / plan Load: Obtaining the temperature distribution for all future times up to the end of the melting cycle. Plan future actions based on a set of rules: past experience and rules used in the company running the furnace, analysis of past melting cycles based on machine learning, monitoring and optimisation of the performance indicators of the current melting cycle.

Step 8: Store Output in the database: Simulated Temperature data, Predicted Performance Indicators, Predicted Actions, Load data are stored in the database.

The solve (mode) function may have the following features

Solve (steady-state)

The goal is to start a simulation with a relevant temperature distribution in the furnace walls. The solver uses an arbitrary duration (e.g. 5h) for an empty furnace, with the following conditions to obtain steady-state:

• Doors are closed

• Total load weight of a certain mass of liquid (T = 750°C) e.g. 2 tonnes

• Burners are working and the burners power are calculated by using PID (Proportional Integral-Differential) functions to maintain the vault temperature at the vault target

temperature. As no sensor data is used to calculate the furnace power, this mode is very similar to future mode (described hereafter).

• No new actions during the calculation. At the end of this step, the furnace load is set to 0 (empty furnace) and will be ready for new loads.

• T_vault = target temperature of the vault

• Simulation time = e.g. 5h to establish the steady state condition

• Simulation time step = e.g. ~10 s

Simulation time can be set to zero after reaching the steady state condition.

The steady state calculation is not mandatory, if the program can restart from the end of the previous simulation.

Solve (past)

After obtaining the suitable temperature distribution as an initial condition, the simulation is performed to obtain the temperature field for all past times steps from the time corresponding to the beginning of the Input Sensor data up to the current time (end of the data). If no recording of the temperature field from previous cycles is available, the beginning of the past time is arbitrary set to the beginning of the 2 previous cycles (around 15 - 24h of simulation time). However, if the temperature field related to the last melting cycle is available, the simulation can start directly from the beginning of the current cycle. The solve (steady state) is not necessary in such condition. The calculation conditions for the past mode are as follows:

• Door opening conditions according to the door opening sensor signals. A dedicated function gets the door opening condition for the past times according to the available measured sensor data.

• Furnace power is calculated as a function of gas / air flow rates and temperatures.

• Actions such as loading, stirring, skimming, transfer including corresponding action times are obtained from the Action data.

• Loading data indicating type and weights, sequence, etc., is obtained from the Load data. Solve (future)

The solve (future) calculation mode determines the temperature field from the current time for all future times up to the end of the cycle.

The calculation conditions set for the future mode are as follows:

• Door opening condition according to the predicted actions. The Action data can always determine if the doors are open or closed according to the start_time and end_time of the planned actions.

• Burners power is calculated using PID functions to reach and maintain the vault or metal temperatures at the vault / metal target temperatures.

• Actions such as loading, stirring, skimming, transfer including corresponding action times are obtained from planned Action data.

• Loading data indicating type and weights, sequence, etc. will be obtained from the Load data.

List of references:

Metal 1

Melt

Melt surface 1 a

Solid

Ingots

Scrape

Furnace 3

Walls 5

Top wall 5a

Bottom wall 5b

Side walls 5c

Door 7

Sliding door

Furnace chamber 9

Gas burners 1 1 Flue gas exit 13

Heat Exchanger 17

Stirrer 17

Control System 2

Sensors 15

Temperature sensors

Thermocouple 15d to measure the metal temperature

Thermocouple 15e to measure the temperature on a top side of the chamber

Thermocouple 15f to measure the temperature on the bottom side of the furnace

Door sensor

Sensor 15a indicating that the furnace door is fully open

Sensor 15b indicating that the furnace door is fully closed

Sensor to measure the inner furnace pressure

Flue gas sensor

Thermocouple 15g to measure the temperature of the exit flue gas Thermocouple to measure the flue gas temperature after the heat exchanger

Thermocouple to measure the flue gas temperature before the heat exchanger

Gas sensor

Furnace control unit 4

Inputs

Outputs

Melt optimization computing system 6

Inputs

Outputs

Server 8

communications network 12

Database 10

Melt optimization control program (Process Optimization Module) POM Simulation module 18a

Learning module 18b

data acquisition module

data storage module

database 10

Operator interface 14

Furnace status panel 16a

Recommendation panel 16b

Control and further actions panel 16c