Control system for a forehearth of a glass production system

Abstract

Control system (50) for controlling heating of a forehearth (30) of a glass production system (10), configured to: receive a target temperature outcome (ETAR); iteratively generate candidate temperature setpoint sets (TSET1, TSET2, …, TSETN); for each candidate temperature setpoint set (TSET1, TSET2, …, TSETN), iteratively generate a plurality of respective candidate control parameter sets (Q1, Q2, …, QN); associate a respective simulated temperature set (T1, T2, …, TN); for each candidate temperature setpoint set (TSET1, TSET2, …, TSETN), determine one converging control parameter set (Q1, Q2, …, QN) among the candidate control parameter sets (Q1, Q2, …, QN); generate a respective simulated temperature outcome (ESIM) for each converging candidate control parameter set (Q1, Q2, …, QN); based on the comparison between the simulated temperature outcomes (ESIM) and the target temperature outcomes (ETAR), select one of the respective converging candidate control parameter sets (Q1, Q2, …, QN) as an actual control parameter set (Q1, Q2, …, QN) for controlling the heat exchange means (38, 40) of the forehearth (30).

Description

[0001] "CONTROL SYSTEM FOR A FOREHEARTH OF A GLASS PRODUCTION SYSTEM"

[0002] Cross-Reference to Related Applications

[0003] This Patent Application claims priority from Italian Patent Application No . 102024000023247 filed on October 18 , 2024 , the entire disclosure of which is incorporated herein by reference .

[0004] Technical Field

[0005] The present invention relates to a control system for a forehearth of a glass production system . It also relates to a forehearth comprising the control system, a glass production system comprising the forehearth, a correspondent control method and a relative computer program product .

[0006] Background of the Invention

[0007] As known, glass containers are made by means of a glass production system, namely comprising a batch house , a hot end and a cold end .

[0008] The batch house allows to prepare and mix into batches the raw materials required for glass production ( typically sand, soda ash, limestone , feldspar, cullet and other raw materials ) .

[0009] The hot end melts the batched materials into molten glass and produces segments of molten glass ( referred to as glass " gobs" ) that are then moulded into glass containers .

[0010] The cold end inspects the glass containers to ensure that they are of acceptable quality .

[0011] Typically, the moulding phase of the hot end of the manufacturing process is performed in an Independent Section ( IS ) machine , which comprises a plurality of identical sections ( e . g . , between one and twenty) , each of which is adapted to produce one or more containers simultaneously (e . g. , up to four) .

[0012] In particular, the hot end operations are performed by a furnace, a forehearth, a gob feeder apparatus, a shearing mechanism, a gob distribution arrangement and delivery system and the sections of the IS machine.

[0013] In the furnace, the batched materials are melted into molten glass and supplied to a feeder bowl of the gob feeder apparatus. In particular, a forehearth connects the furnace and the gob feeder apparatus and is used to transfer the molten glass from the furnace to the gob feeder apparatus. The forehearth can be operated to control the conditions of the molten glass flowing through it, in particular its temperature, for example by means of burners and fans placed in the forehearth.

[0014] Then, in the gob feeder apparatus, streams of molten glass flow from the feeder bowl through one or multiple outlets, towards the sections of the IS machine. Each of the streams of molten glass is cut with the shearing mechanism, located below the feeder bowl, into roughly cylindrical segments of glass (i.e., the gobs) , which fall by gravity and are guided by means of the gob distribution arrangement and delivery system (e.g., comprising scoops, troughs and deflectors) into their respective moulds in the respective section of the IS machine. Since each section of the IS machine can have a plurality of sets of moulds which can operate simultaneously, a correspondent number of cut molten glass streams can be simultaneously supplied to these respective sets of moulds.

[0015] As evident, an accurate control of the molten glass in the forehearth and of the design itself of the forehearth is required for an optimal glass production, to prevent temperature and viscosity problems of the molten glass in the feeder bowl .

[0016] Known solutions use pre-analysis tools , for example based on computational fluid dynamics ( CFD) models , to help designing the forehearth . However, such simulations currently can only be used in the design phase and are independent from the real-time functioning of the forehearth .

[0017] Moreover, during use , the forehearth functioning can be controlled by de fining setpoints for the burners , to be controlled in closed-loop . However, currently the setpoints can be set either manually by a skilled human operator, expert in the functioning of the glass production system, or based on empirical curves , obtained by recording and analysing the past operations of the forehearth . As evident, these known approaches can lead to severe issues , for example caused by insuf ficient expertise of the operator or by inaccuracy or partiality of the obtained empirical curves .

[0018] Furthermore , after the setpoints are defined, currently the functioning of the burners is adj usted in closed-loop by acquiring feedbacks from sensors in the forehearth and by comparing such feedbacks with the setpoints . This known control implies a static control which cannot take into consideration real-time changes of the operative parameters of the forehearth or of the surrounding environment .

[0019] Therefore , known solutions do not allow for any kind of automatic feedback between the feeder and the forehearth .

[0020] Document BYRSKI WITOLD et al . , "Adaptive identi fication method for simulation and control of glass melting process" , 2019 24 TH INTERNATIONAL CONFERENCE ON METHODS AND MODELS IN AUTOMATION AND ROBOTICS (MMAR) , IEEE , 26 August 2019 , discloses an adaptive identi fication method for simulation and control of glass melting process that is based on using ID modelling techniques for the forehearth and arbitrarily selecting the temperature control setpoints .

[0021] Summary of the Invention

[0022] The aim of the present invention is to provide a control system, a forehearth, a glass production system, a control method and a computer program product that overcome the issues mentioned above .

[0023] According to the present invention, a control system, a forehearth, a glass production system, a control method and a computer program product are provided, as defined in the annexed claims .

[0024] Brief Description of Drawings

[0025] For a better understanding of the present invention, preferred embodiments thereof are now described, purely by way of non-limiting example and with reference to the attached drawings , wherein :

[0026] - Figure 1 is a block diagram schematically showing components of a glass production system;

[0027] - Figure 2 is a schematical perspective view showing a forehearth of the glass production system of Figure 1 , according to an embodiment of the present invention; and

[0028] - Figure 3 is a block diagram schematically showing a control system for the forehearth of Figure 2 , according to an embodiment of the present invention .

[0029] In particular, the figures are shown with reference to a triaxial Cartes ian reference system defined by an X axis , a Y axis and a Z axis , orthogonal to each other .

[0030] In the following, elements common to the di f ferent embodiments have been indicated with the same reference numbers .

[0031] Detailed Description of the Invention

[0032] Figure 1 schematically shows a glass production system 10 for the production of glass articles such as glass bottles .

[0033] The glass production system 10 comprises a batch house 12 , a hot end 14 and a cold end 16 .

[0034] The batch house 12 , of per se known type , allows to prepare and mix into batches of raw materials 11 required for glass production ( typically sand, soda ash, limestone , feldspar, cullet and other raw materials ) .

[0035] The hot end 14 melts the batched materials 13 into molten glass 16 and produces segments of molten glass ( also referred to as glass gobs 20 ) that are then moulded into glass containers 24 .

[0036] The cold end 16 inspects the glass containers 24 to ensure that they are of acceptable quality .

[0037] Typically, the moulding phase of the hot end 14 of the manufacturing process is performed in an Independent Section ( IS ) machine 14 ' .

[0038] The IS machine 14 ' comprises a furnace 15 , a forehearth 30 , a gob feeder apparatus 17 , a shearing mechanism 19 , a gob distribution arrangement and delivery system 21 and a plurality of identical sections 23 ( e . g . , between one and twenty) , each of which is adapted to produce one or more containers simultaneously ( e . g . , up to four ) .

[0039] In the furnace 15 , the batched materials 13 are melted into molten glass 16 and supplied, through the forehearth 30 , to a feeder bowl 17 ' of the gob feeder apparatus 17 .

[0040] In particular, the forehearth 30 couples the furnace 15 and the gob feeder apparatus 17 and is used to transfer the molten glass 16 from the furnace 15 to the feeder bowl 17' .

[0041] Then, in the gob feeder apparatus 17, streams 18 of molten glass flow from the feeder bowl 17' through multiple outlets, towards the sections 23 of the IS machine 14' . Each of the streams 18 of molten glass is cut with the shearing mechanism 19, located below the feeder bowl 17' , into roughly cylindrical segments of glass (i.e., the gobs 20) , which fall by gravity and are guided by means of the gob distribution arrangement and delivery system 21 (e.g., comprising scoops, troughs and deflectors) into the moulds in the respective sections 23. Since each section 23 of the IS machine 14' can have a plurality of sets of moulds which can operate simultaneously, a correspondent number of cut molten glass streams 18 can be simultaneously supplied to these respective sets of moulds.

[0042] As evident, other components of the IS machine 14' or, more in general, of the hot end 14 and cold end 16 can be present even if they are not here described in detail, since they are of per se known type and they are not useful for the understanding of the present invention.

[0043] Figure 2 shows the forehearth 30 in further detail.

[0044] The forehearth 30 has an external body 32 that defines and delimitates an internal volume (or channel) 34 extending through the external body 32 from an entry end (or entry opening) 32' and an exit end (or exit opening) 32" of the external body 32. The ends 32' and 32" are opposite to each other with respect to a main direction (or longitudinal direction) of the external body 32.

[0045] The molten glass 16 enters the forehearth 30 through the entry end 32' and exits it through the exit end 32", during use.

[0046] The forehearth 30 comprises a plurality of portions (or trunk) 36, housed in the channel 34 and consecutive among them through the main direction. In the example of Figure 2, two portions 36 are exemplarily shown; however, a different number of portions 36 can be analogously considered (e.g., five) .

[0047] Each portion 36 comprises a respective plurality of burners (or heaters) 38, for example symmetrically placed on both sides of a transversal section of the channel 34 and equally spaced among them along a longitudinal section of the channel 34. The burners 38 can be provided with combustion material (e.g., a mixture of air and methane) and can generate heat through combustion of such material, to heat the molten glass 16 flowing through the channel 34. In detail, the combustion in the burners 38 can be selectively controlled based on heating signals, received by the burners 38 and better described in the following. The heating signals control the amount of combustion material that is received and ignited by the burners 38, thus controlling the amount of heat produced by the burners 38.

[0048] Each portion 36 also comprises one or more respective fans (or generical heat exchangers) 40, for example placed in the middle of the transversal section of the channel 34. The fans 40 can be used to circulate the air in the channel 34, to cool the molten glass 16. The air circulation by the fans 40 can be controlled based on cooling signals, received by the fans 40 and better described in the following. The cooling signals thus control the cooling effect achieved through the fans 40.

[0049] In general, the burners 38 and the fans 40 define temperature controlling means (also called heat exchange means) of each portion 36, that are controllable based on temperature setting signals (i.e., the heating signals and the cooling signals) so as to adjust the temperature of the molten glass 16.

[0050] Each portion 36 also comprises one or more temperature sensor 42, e.g. thermocouples or optical pyrometers. In the example of Figure 2, one temperature sensor 42 is present for each portion 36.

[0051] The temperature sensor 42 is placed in a predefined and fixed position in the respective portion 36. For example, the temperature sensor 42 is placed at an exit end of the respective portion 36, in the longitudinal section of the channel 34, and is placed at the free surface of the molten glass 16 (e.g., few centimetres below the free surface of the molten glass 16) .

[0052] The temperature sensor 42 is used to measure the local temperature of the molten glass 16 passing through each portion 36. Therefore, the temperature sensor 42 generates in use a temperature signal indicative of the local temperature of the molten glass 16, as better described in the following.

[0053] The portions 36 are placed in succession among them so that the molten glass 16 in the forehearth 30 consecutively passes through each one of these portions 36.

[0054] The forehearth 30 also comprises an output temperature sensing device, also called temperature grid (or equalizing grid) , 44, housed in the channel 34. For example, the temperature grid 44 is placed at the exit end 32" of the forehearth 30.

[0055] The temperature grid 44, of per se known type, comprises a plurality of output temperature sensors (not shown) , e . g . thermocouples or optical pyrometers . The output temperature sensors are arranged in a per se known way so as to measure the temperatures of the molten glass 16 at respective locations , in proximity to the exit end 32" of the forehearth 30 , in order to map the temperature distribution of the molten glass 16 exiting the forehearth 30 .

[0056] For example , the temperature grid 44 is arranged transversally to the channel 34 and can comprise nine output temperature sensors , substantially arranged in a 3x3 matrix . For example , three central output temperature sensors can be placed centrally in the channel 34 and arranged so as to be vertically aligned among them in a column ( e . g . , one output temperature sensor at about few centimetres from the bottom surface of the channel 34 , one output temperature sensor at about few centimetres from the free surface of the molten glass 16 and one output temperature sensor which is substantially equally spaced from the other two output temperature sensors ) , three left output temperature sensors can be placed on the left side of the central output temperature sensors , in particular at a distance from the three central output temperature sensors which is about 33% of the transversal width of the channel 34 , and arranged so as to be vertical ly aligned among them in a respective column ( e . g . , analogously to the central output temperature sensors ) , and three right output temperature sensors can be placed on the right side of the central output temperature sensors , in particular at a distance from the three central output temperature sensors which is about 33% of the transversal width of the channel 34 , and arranged so as to be vertically aligned among them in a respective column ( e . g . , analogously to the central output temperature sensors ) .

[0057] Therefore , during use the temperature grid 44 generates grid data indicative of the temperatures measured by the output temperature sensors , for example in the form o f a signal indicative of a matrix of temperatures , each temperature corresponding to the measurement of a respective output temperature sensor . Based on this matrix of temperatures , measured at a same time instant , it is possible to analyse the temperature distribution and homogeneity of the molten glass 16 exiting the forehearth 30 at said time instant .

[0058] In detail , this temperature mapping is indicative of the ef ficiency of the forehearth 30 : the more homogeneous are the temperatures , the higher is the working ef ficiency of the forehearth 30 .

[0059] The IS machine 14 ' can also comprise a control unit , shown in Figure 2 with the reference number 48 .

[0060] The control unit 48 is an electronic control unit , such as a dedicated processor, an FPGA, etc .

[0061] The control unit 48 can be external to the forehearth 30 and operatively coupled it or can be integrated in the forehearth 30 . In the following, the second option is exemplarily considered .

[0062] The control unit 48 is operatively coupled with the burners 38 and the fans 40 , to control them by providing them the temperature setting signals , and is operatively coupled with the temperature sensors 42 and the temperature grid 44 , to acquire from them the temperature measurements .

[0063] Figure 3 shows a control system 50 for controlling the forehearth 30 . In detail, the control system 50 implements in use a control method for controlling the forehearth 30. The description of the control method will be evident in view of the following description of the control system 50.

[0064] The control system 50 is for example comprised in the control unit 48, so that the control method can be implemented by means of the control unit 48. However, other embodiments can be analogously considered, as better detailed in the following.

[0065] The control method, as now described, allows to determine a control setup for the forehearth 30, in particular for a new glass production. In the following, this is referred to as a glass production setup phase of the control method.

[0066] In detail, a new glass production requires that the forehearth 30 is controlled so as to reach a target efficiency (or efficiency setpoint) E AR, which is predefined and specific for such a new production. The choice of the target efficiency E AR is performed in a per se known way based on the desired glass production, for example is manually set by a skilled operator or is automatically determined based on lookup tables (e.g., empirically determined and associating to each glass production line a correspondent target efficiency) .

[0067] The target efficiency ETAR corresponds to a setpoint (i.e., a target) of the temperature map of the temperature grid 44. Thus, the target efficiency ETAR is indicative of a matrix of target temperatures, each one relative to a respective output temperature sensor of the temperature grid 44. Typically, a matrix having all elements equal to a same output temperature (predefined and selected based on per se known criteria) has the maximum efficiency (e.g., 100%) ; in practical applications, a lower efficiency value is generally considered as a target (e.g., greater than 90%, in particular greater than 95%, for example equal to about 97%) to limit the overall time and energy consumption that is required to approach the ideal efficiency.

[0068] According to an embodiment, the target efficiency E AR depends on a target variance of the temperatures of the temperature grid 44, i.e. depends on an allowable upper limit for the temperature variation during the considered glass production. As a consequence, the target efficiency ETAR is indicative of a target homogeneity of the temperatures of the temperature grid 44 (in detail, the lower is the variance, the higher is the homogeneity) .

[0069] Therefore, in the glass production setup phase of the control method, the target efficiency E AR is received as an input by the control system 50.

[0070] As shown in Figure 3, the control system 50 defines an iterative optimization loop and, in the optimization loop, an iterative convergence loop.

[0071] As better described in the following, the optimization loop is iteratively run to find a production control parameter set (also called actual control parameter set) for controlling the forehearth 30, which is optimized in terms of energy consumption required by the forehearth 30. The production control parameter set is the set of control parameters required to operate the temperature controlling means; in other words, the temperature setting signals are indicative of said production control parameter set. In Figure 3, the control parameters are indicated with the references Qi, Q2, ..., QN, considering a plurality N of portions 36; therefore, the control parameter Qi (with i=l, N) is indicative of the heat generated in the channel 34 by the burners 38 of the i-th portion 36, as well as possibly also of the entity of air ventilation caused by the fan 40 of the i-th portion 36.

[0072] In detail, the optimization loop allows to find a production temperature setpoint set (also called actual temperature setpoint set) that corresponds to the optimal set of setpoint temperatures for controlling the portions 36. In Figure 3, the production temperature setpoints are indicated with the references TSE I, TSET2, ..., TSETN, considering a plurality N of portions 36; therefore, the production temperature setpoint TSETI (with i=l, ..., N) is indicative of the optimal setpoint temperature for the molten glass 16 in the i-th portion 36, in particular considered at the position of the temperature sensor 42.

[0073] Then, the convergence loop is iteratively run to determine the control parameter set, starting from the temperature setpoint set.

[0074] Therefore, in the following the iterations of the optimization loop are referred to as optimization iterations and the iterations of the convergence loop are referred to as convergence iterations. As evident from the above, each optimization iteration comprises a plurality of convergence iterations .

[0075] With reference again to Figure 3, the control system 50 comprises an optimizer 54.

[0076] The optimizer 54 is configured to implement known optimization techniques based on the generation of candidate solutions through the different optimization iterations, and their analysis until an optimal solution can be found. The optimizer 54 can be implemented through per se known Al or ML optimization tools, such as gradient-based methods (e.g., Conjugate Gradient Method or Stochastic Gradient Descent method) or direct search methods (e.g., Nelder-Mead method) . In this case, the optimal solution is found when optimization criteria, better discussed in the following, are met.

[0077] In each optimization iteration, the optimizer 54 receives as input the target efficiency (also called target temperature outcome) ETAR and a simulated efficiency (also called predicted temperature outcome) ESIM.

[0078] At the first optimization iteration of the glass production setup phase, the simulated efficiency ESIM can be predefined (e.g., can be null) , whereas in the following optimization iterations the simulated efficiency ESIM can be determined as discussed in the following.

[0079] In detail, the optimizer 54 can receive an error between the target efficiency ETAR and the simulated efficiency ESIM, for example calculated by an efficiency error module 53 preceding the optimizer 54, or can calculate this error on its own.

[0080] Based on the error between the target efficiency ETAR and the simulated efficiency ESIM, the optimizer 54 determines a candidate temperature setpoint set TSETI, TSET2, ..., TSETN for the considered optimization iteration. The candidate temperature setpoint set TSETI, TSET2, ..., TSETN comprises one respective candidate setpoint temperature TSETI for each of the portions 36, in particular considered at the position of the respective temperature sensor 42. For example, the values of the candidate temperature setpoints TSETI, TSET2, ..., TSETN of each set can decrease going from the first portion 36 to the last portion 36 (i.e., going from the entry end 32' to the exit end 32") .

[0081] The control system 50 also comprises a control parameter module 56 that is coupled to the optimizer 54 so as to, at each optimization iteration, receive the respective candidate temperature setpoint set TSE I, TSET2, ..., TSETN and generate a corresponding converging candidate control parameter set Qi, Q2, ..., QN. In particular, this is possible through the execution of a plurality of convergence iterations that end when convergence criteria are met, i.e. when the converging candidate control parameter set is selected among the considered candidate control parameter sets .

[0082] In detail, in the example of Figure 3 the control parameter module 56 comprises a respective control parameter unit 56i for each portion 36, that receives the respective candidate temperature setpoint TSETI . However, other analogous implementations are possible, es evident (e.g., a single module instead of a plurality of units) .

[0083] In each convergence iteration of the glass production setup phase, each control parameter unit 56± also receives as input a respective simulated temperature T± .

[0084] At the first convergence iteration of each optimization iteration of the glass production setup phase the simulated temperature T± can be predefined (e.g., can be null) , whereas in the following convergence iterations the simulated temperature T± can be determined as discussed in the following. Moreover, it is also possible to assume, at the first convergence iteration, that the distribution of the molten glass 16 in the channel 34 is homogeneous and that the temperature controlling means are off but ready to be activated .

[0085] In detail, each control parameter unit 56± can receive an error between the candidate temperature setpoint TSETI and the simulated temperature T±, for example calculated by a respective temperature error module 55± preceding the control parameter unit 56±, or can calculate this error on its own.

[0086] Based on the error between the candidate temperature setpoint TSE I and the simulated temperature T±, the control parameter unit 56± determines a correspondent candidate control parameter Qi .

[0087] For example, the candidate control parameter Q± can be calculated by multiplying for a gain coefficient (e.g., equal to about 20000) the error between the candidate temperature setpoint TSE I and the simulated temperature T±. However, other implementations are also possible (e.g., based on lookup tables associating to each error a respective candidate control parameter Q±, etc.) .

[0088] The control system 50 also comprises a simulation module 60, coupled to the control parameter module 56 so as to, at each convergence iteration, receive the candidate control parameter set Qi, Q2, ..., QN.

[0089] According to an embodiment, the simulation module 60 is configured to implement a model of the forehearth 30, so as to realistically simulate its functioning. In detail, this model is a virtual representation of the forehearth 30 that allows to accurately simulate its functioning. In other words, this model is a digital twin of the forehearth 30.

[0090] In detail, the model can be a mathematical model, in particular a numerical model, defined by one or more set of mathematical formulas that mimics the structure and the functioning of the forehearth 30. Such mathematical formulas can be determined according to per se known empirical methods .

[0091] For example, the model can be based on CFD ("computational fluid dynamics") techniques, in particular thermal CFD techniques.

[0092] CFD simulations are usually performed through Finite Volume Methods (FVM) , which is a family of numerical models that integrate the Partial Differential Equations (DDEs) for Mass and Momentum over small control volumes, to solve them in terms of conserved variables through a discrete formulation. Such equations can be formulated in 1-D, 2-D or 3-D. The solution of the Heat Exchange Equation occurs in the same manner, whereas the temperature is described by a 1-D, 2-D or 3-D partial differential equation incorporating the advection, diffusion and reaction phenomena. FVMs typically solve iteratively the unsteady (i.e., timevariant) version of the DDEs until a statistically steady state condition (i.e., time-invariant) is found, which is defined as the steady solution to the system of DDEs with its initial and boundary conditions.

[0093] In detail, the model can simulate the temperature distribution in the channel 34 of the forehearth 30 starting from the knowledge of the heat generated in the simulation by the burners 38 and the air ventilation caused in the simulation by the fans 40 (i.e., starting from the candidate control parameter set Qi, Q2, ..., QN) .

[0094] Moreover, the model can also simulate the temperature distribution in the channel 34 of the forehearth 30 starting from the actual temperature measurements performed by the temperature sensors 42. In other words, it also allows to homogenously reconstruct the temperature variations in the channel 34 by interpolating the temperature measurements of the temperature sensors 42 .

[0095] In each convergence iteration, the simulation module 60 simulates the functioning of the forehearth 30 starting from the received candidate control parameter set Qi, Q2 , ..., QN and generates a corresponding simulated temperature set Ti, T2, TN.

[0096] The simulated temperature set Ti, T2, TN comprises one respective simulated temperature T± for each candidate control parameter Q±, that corresponds to the simulated value of the temperature at a predefined point in space in the respective portion 36 ( in detail , corresponding to the position of the respective temperature sensor 42 ) .

[0097] The simulated temperature set Ti, T2, ..., TN are then sent back to the temperature error modules 55 so that they can be used as feedbacks for the next convergence iteration . Therefore , the error in input to each control parameter unit 56i is the error between the candidate temperature setpoint TSETI and the simulated temperature T± that has been simulated by the model at the previous convergence iteration .

[0098] The control parameter module 56 stops the convergence iterations when the convergence criteria are met , for example when the errors in input ( alternatively, the mean of the absolute errors in input ) are lower than a convergence threshold ( e . g . , equal to about 1 degree ) .

[0099] When the convergence criteria are met , the last candidate control parameter set Qi, Q2, ..., QN ( i . e . , the one determined in the convergence iteration wherein convergence was found) is stored as the converging control parameter set to be used for the current optimi zation iteration .

[0100] When the convergence criteria are met , the simulation module 60 also outputs the simulated efficiency ESIM associated to the converging candidate control parameter set Q Q2, QN and the respective input candidate temperature setpoint set TSETI, TSET2, ..., TSETN.

[0101] In detail, the simulated efficiency ESIM corresponds to the efficiency of the temperature grid 44 that is simulated (i.e., predicted) by the model based on the last candidate control parameter set Qi, Q2, ..., QN. The definition of efficiency for the simulated efficiency ESIM is analogous to the one that has already been provided for the target efficiency ETAR (e.g., may depend on the predicted variance of the simulated temperatures of the temperature grid 44) .

[0102] The simulated efficiency ESIM is then sent back to the optimizer 54 so that it can be used as feedback for the next optimization iteration. Therefore, the error in input to the optimizer 54 is the error between the target efficiency ETAR and the simulated efficiency ESIM that has been simulated by the model at the previous optimization iteration.

[0103] The optimizer 54 can generate the candidate temperature setpoint sets TSETI, TSET2, ..., TSETN through the optimization iterations according to per se known techniques, for example in a random way, in a pseudo-random way or, advantageously, based on the efficiency error calculated for the last candidate control parameter set Qi, Q2, ..., QN. This last option allows to speed up the time required for the optimization process.

[0104] For example, in the conjugate gradient method the local value of the gradient of the response surface of the model is computed at each iteration step. Once the local value of the gradient is known, the next iteration is computed counter gradient-wise, changing the values of the optimization variables in the direction of locally minimum gradient. If the response surface only admits a global minimum, then the optimization also converges to the global minimum.

[0105] The optimizer 54 stops the optimization iterations when the optimization criteria are met.

[0106] For example, the optimization criteria are met when the simulated efficiency ESIM satisfies a predefined efficiency relationship with the target efficiency E AR (e.g., the simulated variance is lower than the target variance, so that the simulated efficiency ESIM is greater than the target efficiency ETAR) and, preferably, also when a simulated total energy consumption satisfies a predefined energy condition (e.g., is lower than an optimization threshold or has asymptotic behaviour) .

[0107] In detail, at each optimization loop the simulated total energy consumption corresponds to the sum of the simulated energy consumptions of the temperature controlling means of the portions 36, when they are controlled based on the candidate control parameter set Qi, Q2, ..., QN that led to convergence in the convergence loop of the considered optimization loop. In other words, the simulated total energy consumption is correlated to the sum of the candidate control parameters Qi, Q2, ..., QN that led to convergence. For example, the simulated total energy consumption can be outputted by the simulation module 60 when convergence of the convergence loop is reached, or it can be calculated by the optimizer 54 based on the candidate control parameter set Qi, Q2, ..., QN that led to convergence.

[0108] When optimization is reached, the control system 50 outputs the last candidate control parameters Qi, Q2, ..., QN (i.e., the ones that led to convergence in the convergence loop of the last optimization loop in which optimization was achieved) as the production (or actual) control parameter set Qi, Q2, QN for controlling the temperature controlling means of the forehearth 30. Optionally, it may also store and / or output the correspondent simulated efficiency ESIM.

[0109] Therefore, by means of the simulation module 60 it is possible to predict the functioning of the forehearth 30 under different conditions (i.e., according to different candidate solutions of the simulation) and to analyse it in order to find the optimal solution in terms of energy consumption .

[0110] In real-time use of the forehearth 30 (i.e., during the glass production) , the forehearth 30 is controlled based on the temperature setting signals, in particular is controlled in closed-loop to maintain the measurements of the temperature sensors 42 as close as possible to the production temperature setpoints TSETI, TSET2, ..., TSETN.

[0111] In addition to the above, the control method can allow to update in real-time the control setup for the forehearth 30, in particular for the same glass production that is being currently performed. In the following, this is referred to as a glass production update phase of the control method.

[0112] In the glass production update phase, the control system 50 receives as inputs the measured temperatures from the temperature sensors 42, a measured efficiency (also called measured temperature outcome) from the temperature grid 44 and measured energy consumption from the temperature controlling means. In detail, the measured energy consumption corresponds to the measured value of the energy consumed by the temperature controlling means during their functioning. For example, these inputs are received by the control system 50 periodically, e.g. each 5-10 min.

[0113] In the glass production update phase, the control system 50 periodically runs again the previously discussed operations starting from the received inputs, so as to analyse in real-time if there are any updated control setting (i.e., any updated control parameter set Qi, Q2, ..., QN) that may lead to a further reduction in energy consumption. If so, the forehearth 30 starts to be controlled based on this updated control setting, thus reducing the overall energy consumption required for its functioning. Therefore, the received inputs are the starting points for the new simulations and iterations of the control method, thus allowing to explore new conditions for the forehearth 30 as it is being currently operated.

[0114] For example, the control system 50 starts again a new running of the operations based on the updated input each 5- 10 min.

[0115] In the glass production update phase, the measured efficiency from the temperature grid 44 is used as the target efficiency E AR. Moreover, the measured temperatures from the temperature sensors 42 are used as the simulated temperature set Ti, T2, ..., TN in the first convergence iteration of each optimization iteration. Furthermore, the measured energy consumption from the temperature controlling means is used as the optimization threshold when assessing the optimization criteria.

[0116] From what has been described and illustrated previously, the advantages of the present invention are evident .

[0117] The model allows for a series of fluid dynamics and heat exchange simulations aimed at evaluating what happens to the molten glass i f certain temperatures are set . This allows , through optimization loops , to automatically set the desired temperatures on the portions 36 .

[0118] The high speed of execution of the simulations allows to explore an enormous amount of dif ferent temperature configurations . Moreover, by implementing the optimi zation, it is possible to set the configuration which, in addition to having the desired glass features , minimi zes the amount of energy consumed by the temperature controlling means , which is a parameter with a strong impact on the total energy consumption of the glass production .

[0119] This solution therefore allows to avoid choosing the desired temperatures in the preliminary phase of production . Moreover, during production it can be used, thanks to the high calculation speed, as a digital twin of the forehearth 30 that , starting from the real-time measurements , simultaneously performs parallel simulations and allows to predict the changes in the conditions of the glass , before actually implementing this new setting in reality .

[0120] The control system 50 can be easily installed in the glass factory, without the addition of sensors . In fact , it uses the sensors already present in the control of the known forehearths . It only needs compatible hardware on which to run the simulations .

[0121] The simulations can be performed using the C++ libraries of OpenFOAM, which is an open-source software . The simulations replicate the forehearth layout and can also estimate the energy losses in the forehearth during use , as well as other parameters that can be useful when designing a new forehearth .

[0122] Moreover, the possibility for the model to simulate the temperature distribution in the channel 34 starting from the actual temperature measurements performed by the temperature sensors 42 is a useful tool for the designers when designing new projects of the forehearth. In particular, in the design phase it helps to determine the optimal channel layout for a specific production line.

[0123] Furthermore, the possibility for the model to simulate the temperature distribution in the channel 34 starting from the actual temperature measurements performed by the temperature sensors 42 is a useful diagnostic tool to analyse, in particular in real-time, the functioning of the forehearth 30. In fact, if the model periodically simulates the temperature distribution in the channel 34 based on the real-time measurements from the forehearth 30, it is possible to determine any temperature anomaly and / or malfunctioning of the forehearth 30 in real-time, without the need of using advanced sensing techniques (that may be more expensive, more difficult to be analysed and managed and may require structural modifications of the layout of the forehearth 30) . In case of temperature anomaly and / or malfunctioning of the forehearth 30, the control system 50 can generate an alert signal (e.g., a visual or audio alert for the operator using or controlling the forehearth 30) .

[0124] Moreover, the control system 50 can also be coupled to a user interface (not shown) to perform at least one of the following: view the distribution of temperatures inside the forehearth 30 in real-time (i.e., diagnosis of the forehearth 30) ; get automated suggestions for possible energetically cheaper control configurations providing the same temperature and homogeneity at the temperature grid 44 (i.e., Al-based augmented workforce) ; automatically suggest the best configuration for a production change .

[0125] In particular, with respect to the known solution disclosed in the previously cited document BYRSKI WITOLD, the present solution allows for an improved control of the forehearth 30 s ince it exploits a 3D model ling of the forehearth 30 that is based on its ef ficiency . In the present solution, the control setpoints are not modi fied in an arbitrary way but rather are iteratively optimi zed by taking into account the 3D structure of the forehearth 30 and by considering the homogeneity of the temperature map ( thus , the temperature spread) at the temperature grid 44 . Moreover, the present solution allows to optimi ze both the thermal ef ficiency and the energy consumption of the forehearth 30 . Furthermore , the present solution allows to dynamically update the control setpoints during operation of the forehearth 30 , e . g . during the glass production .

[0126] Finally, it is clear that modi fications and variations may be made to what has been described and illustrated herein, without thereby departing from the scope of the present invention, as defined in the annexed claims . For example , the di f ferent embodiments described can be combined with each other to provide further solutions .

[0127] For example , the operations that have been previously described as being performed by control unit 48 can be performed according to other analogous techniques , as evident to the skilled person . For example , these operations could be performed by a server operatively coupled to the forehearth 30 , or by generical computational resources interfaced with the forehearth 30 ( such as in cloud) .

[0128] For example , the simulation module 60 can be configured to implement , instead of the previously described model , Al ("artificial intelligence") and / or ML ("machine learning") techniques that have been trained to perform the same operations that have been previously discussed (e.g., based on measurements acquired from the forehearth 30 during its functioning or based on simulation results from the abovementioned mathematical model) .

[0129] For example, the simulation module 60 can be based on a neural network, such as a Physics-Informed Neural Network (FINN) trained on the Thermal-CFD. Exemplary and nonlimiting NN that are available for this purpose are described in the following links: https : / / asmedi git al col lection . asme . org / heattransf er / article / 143 / 6 / 060801 / 1104439 / Physics-Informed-Neural-Networks-for- Heat-Transf er ; https: / / link. springer . com / article / 10.1007 / sl0409-021-01148- 1 ; https : / / link . springer . com / article / 10.1007 / sl0915-022-

[0130] 01939-z .

[0131] For example, these AI / ML tools can be implemented by using available libraries (e.g., PyTorch or Tensorflow, both available open source for Python) .

[0132] The use of AI / ML techniques allows for faster prediction and, hence, faster optimization.

Claims

CLAIMS1. Control system (50) for controlling heating of a forehearth (30) of a glass production system (10) , the forehearth (30) defining a channel (34) and comprising a plurality of subsequent portions (36) along the channel (34) , said portions (36) comprising respective heat exchange means (38, 40) operable to adjust the temperature of molten glass (16) flowing in the respective portions (36) in response to actual control parameters, the control system (50) comprising:- an optimizer (54) configured to: o receive a target temperature outcome (ETAR) indicative of desired temperatures of molten glass (16) at an outlet of the channel (34) , and o generate at least one candidate temperature setpoint set (TSE I, TSET2, ..., TSETN) comprising a respective candidate temperature setpoint(TSETI) for the molten glass (16) flowing in each portion (36) of the forehearth (30) ;- a simulation module (60) configured to: o receive candidate control parameter sets (Qi, Q2 , ..., QN) , each comprising a plurality of candidate control parameters for controlling respectively the heat exchange means (38, 40) of the portions (36) , and o on the basis of each received candidate control parameter set (Qi, Q2, ..., QN) , determine a respective simulated temperature set (Ti, T2, ..., TN) , indicative of predicted temperatures of the molten glass (16) flowing respectively at said portions (36) ; anda control parameter module (56) operatively coupled to the optimizer (54) and the simulation module (60) and configured to: o receive the candidate temperature setpoint set (TSETI, TSE 2, TSETN) and the simulated temperature sets (Ti, T2, TN) , o for each received simulated temperature set (Ti, T2, TN) , iteratively generate the candidate control parameter set (Qi, Q2, ..., QN) , and o determine a converging candidate control parameter set among the generated candidate control parameter sets (Qi, Q2, ..., QN) , with convergence between the simulated temperature set (Ti, T2, TN) and the candidate temperature setpoint set (TSETI, TSE 2, ..., TSE N) ; the simulation module (60) being also configured to determine a predicted temperature outcome (ESIM) for the molten glass (16) at the outlet of the channel (34) on the basis of the converging candidate control parameter set, the optimizer (54) being also configured to:- check if the predicted temperature outcome (ESIM) meets the target temperature outcome (ETAR) , and- if the result of the check is negative, generate a further candidate temperature setpoint set (TSETI, TSET2, ..., TSETN) so as to determine a further converging candidate control parameter set and a further predicted temperature outcome,- if the result of the check is positive, select the converging candidate control parameter set (Qi, Q2, ..., QN) as the actual control parameters for actually controlling the heat exchange means (38, 40) .

2. Control system according to claim 1, wherein the simulation module (60) is configured to implement a mathematical model, in particular a numerical model, of the forehearth (30) that is configured to simulate the structure and functioning of the forehearth (30) so that each simulated temperature set (Ti, T2, ..., TN) is determined starting from the respective candidate control parameter set (Qi, Q2, ..., QN) •3. Control system according to claim 2, wherein the mathematical model is based on thermal computational fluid dynamics techniques.

4. Control system according to claim 1, wherein the simulation module (60) is configured to implement artificial intelligence and / or machine learning techniques, in particular a neural network, in further detail a Physics- Informed Neural Network, trained to associate to each candidate control parameter set (Qi, Q2, ..., QN) the respective simulated temperature set (Ti, T2, ..., TN) .

5. Control system according to anyone of the preceding claims, wherein the optimizer (54) is configured to be based on a numerical method, in detail a gradient-based method, in further detail Conjugate Gradient Method or Stochastic Gradient Descent method, or a direct search method, in further detail a Nelder-Mead method.

6. Control system according to anyone of the preceding claims, wherein the control system (50) defines an iterative optimization loop and, in the optimization loop, an iterative convergence loop, wherein one respective candidate temperature setpoint set (TSE I, TSE 2, ..., TSETN) is generated at each optimization iteration of the optimization loop,wherein a plurality of convergence iterations are configured to be executed for each optimization iteration, wherein one respective candidate control parameter set (Q Q2, QN) is generated at each convergence iteration of the convergence loop.

7. Control system according to claim 6, wherein the optimizer (54) is configured to stop the optimization iterations when optimization criteria are met, wherein the control parameter module (56) is configured to stop the convergence iterations when convergence criteria are met, wherein the optimization criteria are met when the predicted temperature outcome (ESIM) is greater than the target temperature outcome (ETAR) and when a simulated total energy consumption is lower than an optimization threshold, the simulated total energy consumption depending on a sum of the candidate control parameters (Qi, Q2, ..., QN) for which convergence is found between the respective simulated temperature set (Ti, T2, ..., TN) and the candidate temperature setpoint set (TSETI, TSET2, ..., TSETN) , wherein the convergence criteria are met when a mean of errors between the simulated temperatures (Ti, T2, ..., TN) and the candidate temperature setpoints (TSETI, TSE 2, ..., TSE N) is lower than a convergence threshold.

8. Control system according to anyone of the preceding claims, wherein the predicted temperature outcome (ESIM) corresponds to a predicted efficiency of the forehearth (30) that depends on a predicted variance of temperatures of the molten glass (16) at said outlet of the channel (34) , the target temperature outcome (ETAR) corresponds to a target efficiency of the forehearth (30) that depends on a targetvariance of temperatures of the molten glass (16) at said outlet of the channel (34) , and wherein the predicted temperature outcome (ESIM) is greater than the target temperature outcome (E AR) when the predicted variance is lower than the target variance.

9. Control system according to claim 7 or 8 when dependent upon claim 7, wherein the control system (30) further comprises a temperature grid (44) placed in the channel (34) at the outlet of the channel (34) , the temperature grid (44) comprising output temperature sensors configured to map the temperatures of the molten glass (16) at said outlet, wherein the control system (30) further comprises a temperature sensor (42) for each portion (36) , configured to measure the temperature of the molten glass (16) in said portion (36) , wherein, in a glass production update phase, the control system (50) is also configured to receive a measured temperature outcome from the temperature grid (44) , indicative of temperatures of molten glass (16) measured by the output temperature sensors, receive measured temperatures from the temperature sensors (42) and receive a measured energy consumption indicative of the measured energy consumption used for operating the heat exchange means , wherein, in the glass production update phase, the measured temperature outcome is used as the target temperature outcome (ETAR) , the measured temperatures are used as the simulated temperature set (Ti, T2, ..., TN) in a first convergence iteration of each optimization iteration and the measured energy consumption is used as theoptimization threshold.

10. Control system according to anyone of claims 6-9 when dependent upon claim 6, wherein the optimizer (54) is configured to generate the candidate temperature setpoint set (TSE I, TSE 2, ..., TSETN) for a current optimization iteration based on an error between the target temperature outcome (ETAR) and the simulated temperature outcome (ESIM) generated in the previous optimization iteration.

11. Forehearth (30) for a glass production system (10) , the forehearth (30) defining a channel (34) and comprising a plurality of subsequent portions (36) along the channel (34) , said portions (36) comprising respective heat exchange means (38, 40) operable to adjust the temperature of molten glass (16) flowing in the respective portions (36) in response to actual control parameters, the forehearth (30) further comprising a control system (50) according to anyone of the preceding claims, wherein the heat exchange means (38, 40) are configured to be controlled based on the actual control parameter set (Q Q2, ..., QN) .

12. Glass production system (10) comprising:- a furnace (15) configured to receive batched materials (13) and melt them into molten glass (16) ;- a gob feeder apparatus (17) with a feeder bowl ( 17 ’ ) , configured to receive the molten glass (16) in the feeder bowl (17' ) and to generate streams (18) of molten glass for the glass production; and- a forehearth (30) according to claim 11, reciprocally coupling the furnace (15) and the gob feeder apparatus (17) and configured to transfer the molten glass (16) from the furnace (15) to the gob feeder apparatus (17) .

13. Control method for controlling heating in a forehearth (30) of a glass production system (10) , the forehearth (30) defining a channel (34) and comprising a plurality of subsequent portions (36) along the channel (34) , said portions (36) comprising respective heat exchange means (38, 40) operable to adjust the temperature of molten glass (16) flowing in the respective portions (36) in response to actual control parameters, the control method being executed by means of the control system (50) and comprising the steps of:- receive a target temperature outcome (ETAR) indicative of desired temperatures of molten glass (16) at an outlet of the channel (34) ;- generate at least one candidate temperature setpoint set (TSE I, TSE 2, ..., TSETN) comprising a respective candidate temperature setpoint (TSETI) for the molten glass (16) flowing in each portion (36) of the forehearth (30) ;- receive candidate control parameter sets (Qi, Q2, ..., QN) , each comprising a plurality of candidate control parameters for controlling respectively the heat exchange means (38, 40) of the portions (36) ;- on the basis of each received candidate control parameter set (Qi, Q2, ..., QN) , determine a respective simulated temperature set (Ti, T2, ..., TN) , indicative of predicted temperatures of the molten glass (16) flowing respectively at said portions (36) ;- for each simulated temperature set (Ti, T2, ..., TN) , iteratively generate the candidate control parameter set (Qi, Q2, ..., QN) ;- determine a converging candidate control parameterset among the generated candidate control parameter sets (Qi, Q2, ..., QN) , with convergence between the simulated temperature set (Ti, T2, TN) and the candidate temperature setpoint set (TSETI, TSET2, ..., TSETN) ;- determine a predicted temperature outcome (ESIM) for the molten glass (16) at the outlet of the channel (34) on the basis of the converging candidate control parameter set;- check if the predicted temperature outcome (ESIM) meets the target temperature outcome (ETAR) ; and- if the result of the check is negative, generate a further candidate temperature setpoint set (TSETI, TSE 2, ..., TSE N) so as to determine a further converging candidate control parameter set and a further predicted temperature outcome, or- if the result of the check is positive, select the converging candidate control parameter set (Qi, Q2, ..., QN) as the actual control parameters for actually controlling the heat exchange means (38, 40) .

14. Computer program product storable in a control unit, the computer program being designed so that, when executed, the control unit becomes configured to carry out a control method according to claim 13.

Images