Continuous casting plant and corresponding regulation method

Abstract

Plant (10) for the continuous casting of metal products comprising a mould (12) in which a metal (L) in the molten state is able to be poured with a determinate flow rate (Q), a regulation device (16) capable of regulating said flow rate (Q), a control unit (18) configured to manage at least the movements of said regulation device (16), and at least one detection device (19) capable of detecting every punctual variation of the level (14) of metal (L) in said mould (12) with respect to a nominal value thereof and generating a corresponding variation signal (SV) and sending it said control unit (18), which is capable of generating a command signal (RS) for said regulation device (16) in order to cause a desired variation of the flow rate (Q). The present invention also concerns the regulation method for generating the above command signal (RS).

Claims

1. Plant for continuous casting of metal products comprising a mould in which a metal in a molten state is able to be poured with a determinate flow rate, a regulation device capable of regulating said flow rate, a control unit configured to manage at least movements of said regulation device, and at least one detection device capable of detecting every punctual variation of a level of metal in said mould with respect to a nominal value thereof and generating a corresponding variation signal having a frequency, an amplitude and a phase and sending the corresponding variation signal to said control unit, wherein said control unit comprises a central processing unit and at least one memory unit connected to the central processing unit in which at least one first compensation algorithm is stored, wherein the at least one first compensation algorithm is configured to make said central processing unit, without using any measurement or estimate of a disturbance acting on said plant, selectively generate a command signal for said regulation device starting from said variation signal and taking into account an overall delay between the generation of said variation signal and actual actuation of said regulation device, said command signal having at least a command signal variable amplitude and a command signal variable phase.

2. Regulation method for a plant for continuous casting of metal products, having a mould to contain metal in a molten state, regulation means to regulate a flow rate of said metal and keep a level thereof stable in said mould, and a control unit connected to said regulation means, wherein said method comprises a detection step in which detection means detect a punctual variation of said level with respect to a nominal value, and generate a corresponding variation signal, having a frequency, an amplitude and a phase, which is sent to the control unit, and wherein the method further comprises a control step, in which said control unit, in response to said variation signal and in accordance with at least one first compensation algorithm, selectively generates, without using any measurement or estimate of a disturbance acting on said plant, a command signal for said regulation device starting from said variation signal and taking into account an overall delay between the generation of said variation signal and actual actuation of said regulation device, said command signal having at least a command signal variable amplitude and a command signal variable phase.

3. Method as in claim 2, wherein said control step comprises a calculation sub-step in which said control unit, before calculating said command signal, processes said variation signal and generates, in an iterative manner and instant by instant, a control signal having a control signal amplitude and a control signal phase, which serves as a basis for generation of said command signal, together with a correction value generated on the basis of said at least one first compensation algorithm which takes into account a mathematical system associated with said overall delay.

4. Method as in claim 3, wherein said mathematical system is considered in a frequency domain and associates a transfer function having a gain and a phase which takes into account said overall delay, said control signal being an input signal of said mathematical system and said command signal being an output signal from said mathematical system, expressed by RS=A.sub.rsen(t+.sub.r)A.sub.cA.sub.ssen(t++.sub.s).

5. Method as in claim 4, wherein during said calculation sub-step said control unit uses said at least one first compensation algorithm first to calculate an error function defined by a difference between said variation signal and said command signal and expressed as E=A.sub.cA.sub.ssen(t++.sub.s)+A.sub.dsen(t+.sub.d), and subsequently a cost function proportional to a square of said error function, wherein said cost function is optimized to find optimal values of amplitude and phase of said control signal.

6. Method as in claim 5, wherein during said calculation sub-step said cost function is optimized on the basis of optimization algorithms selected from a gradient descent algorithm, or an algorithm based on an estimation of a set of moments of a gradient.

7. Plant for continuous casting of metal products comprising: a mould in which a metal in a molten state is able to be poured with a determinate flow rate; a regulation device capable of regulating said flow rate; a control unit configured to manage movements of said regulation device; and at least one detection device capable of detecting every punctual variation of a level of metal in said mould with respect to a nominal value thereof and generating a corresponding variation signal having a frequency, an amplitude and a phase and sending the corresponding variation signal to said control unit; wherein said control unit comprises a central processing unit and at least one memory unit connected to the central processing unit in which at least one first compensation algorithm is stored, and wherein the control unit is programmed and configured to execute the regulation method of claim 2.

8. Method as in claim 3, wherein said at least one first compensation algorithm is configured to also compensate the frequency of said control signal in the event that an initial estimate of the frequency of said variation signal is wrong.

9. Method as in claim 3, wherein said at least one first compensation algorithm generates said control signal only if said variation signal has a frequency greater than or around 0.2 Hz.

10. Method as in claim 5, wherein said at least one first compensation algorithm is implemented with a neural network artificial intelligence having an input layer containing information relating to the frequency of said control signal, a hidden layer containing functions of activation of oscillatory phenomenon of said variation signal, that is, sine and cosine, weighted with orthogonal components of said control signal, and an output layer which is a linear combination of said control signal, and wherein a backpropagation algorithm of said error function is applied to said neural network for iterative estimation of the orthogonal components and relating to the control signal.

Description

DESCRIPTION OF THE DRAWINGS

(1) These and other aspects, characteristics and advantages of the present invention will become apparent from the following description of an embodiments, given as a non-restrictive example with reference to the attached drawings wherein:

(2) FIG. 1 is a schematic section view of a part of a continuous casting plant according to the present invention;

(3) FIG. 2 shows four graphs relating to an example of operation of the plant of FIG. 1 which have the same time scale in seconds in the abscissa axis, while in the ordinate axis they have, respectively: (a) a curve that indicates the level of the metal in the molten state in a mould; (b) a curve that indicates a command signal for a device for regulating the flow rate of the metal in the molten state in the mould; (c) a curve that indicates a signal of variation of the level of the metal in the molten state in the mould; (d) the trend of the casting speed expressed in m/min;

(4) FIG. 3 shows four graphs relating to another example of operation of the plant of FIG. 1 which have the same time scale in seconds in the abscissa axis, while in the ordinate axis they have, respectively: (a) a curve that indicates the level of the metal in the molten state in a mould; (b) a curve that indicates a command signal for a device for regulating the flow rate of the metal in the molten state in the mould; (c) a curve that indicates a signal of variation of the level of the metal in the molten state in the mould; (d) the trend of the casting speed expressed in m/min.

(5) We must clarify that in the present description the phraseology and terminology used, as well as the figures in the attached drawings also as described, have the sole function of better illustrating and explaining the present invention, their function being to provide a non-limiting example of the invention itself, since the scope of protection is defined by the claims.

(6) To facilitate comprehension, the same reference numbers have been used, where possible, to identify identical common elements in the drawings. It is understood that elements and characteristics of one embodiment can be conveniently combined or incorporated into other embodiments without further clarifications.

DESCRIPTION OF AN EMBODIMENT OF THE PRESENT INVENTION

(7) With reference to FIG. 1, a plant 10 for the continuous casting of metal products, according to the present invention, comprises in sequence and from top to bottom: a tundish 11 configured to contain a determinate quantity of metal L in the molten state, possibly coming from a ladle, not shown in the drawings; a mould 12 into which the metal L is able to be poured with a determinate flow rate Q by means of a nozzle 13 of a known type, for example a submerged entry nozzle (SEN), so that a level 14, or meniscus, of the metal L is formed at a determinate distance H from the base of the mould 12, for example; and a plurality of pairs of rollers 15 disposed along a determinate path to effect a gradual solidification (soft reduction) of the cast product.

(8) Associated with the tundish 11 is a regulation device 16 of a known type, which can comprise a vertically mobile rod for example, capable of regulating, for example by moving vertically, the flow rate Q of the metal L, so that the level 14 of the latter in the mould 12 is as constant and stable as possible, despite the various disturbances of the plant 10 due to multiple causes, such as, for example, and without limitations to generality: solid agglomerates accumulated on the internal walls of the nozzle 13 which can suddenly detach and pollute the metal L; variations in the casting speed S; generation of dynamic bulges along the pairs of rollers 15 disposed at the foot, or immediately downstream, of the mould 12, caused by the pressures present in the cast product not yet completely solidified.

(9) In particular, the dynamic bulges as above can cause the MLFs mentioned above, that is, periodic fluctuations of the level 14 with respect to a nominal value defined in the design of the plant 10 on the basis of other casting parameters, such as for example the casting speed S expressed in meters per minute, and fluctuations of the level 14, which can be as much as ten millimeters.

(10) The plant 10 also comprises both a control unit 18 configured to manage, according to certain criteria which will be described below, at least the movements of the regulation device 16, and also at least one detection device 19 capable of detecting every punctual variation of the level 14 with respect to the nominal value as above, that is, that which would occur in the absence of said disturbances of the plant 10 as a whole, and generating a corresponding variation signal SV which is sent to the control unit 18. In other words, the variation signal SV is equivalent to the actual variation of the level 14, and it is correlated in some way to an indirect measurement of the disturbances of the plant 10. We must clarify that the variation signal SV is not equivalent to a disturbance signal.

(11) The control unit 18 comprises a central processing unit, or CPU, 20 and at least one memory unit 21 connected thereto, in which at least one compensation algorithm ALG1 is stored which is capable of making the same CPU 20 generate, in response to the variation signal SV, a command signal RS for the regulation device 16.

(12) In particular, before generating the command signal RS, the control unit 18 generates a control signal CS on the basis of the compensation algorithm ALG1, which also takes into account the overall delay RC that can occur between the detection of the level 14, the generation of the variation signal SV and the actual actuation of the regulation device 16, due to various phenomena, including non-measurable ones. This overall delay RC may be due, for example, but not only, to the inertia of the masses of the regulation device 16 when they are set in motion.

(13) The command signal RS, instead of being sent directly to the regulation device 16, can be combined, for example, with a main control signalalso known as set-point signalwhich is capable of controlling the position of the regulation device 16 in a standard condition in which there is no plant 10 disturbance.

(14) The present invention provides that the command signal RS is generated in advance (feedforward), once it has been established that the variation signal SV exceeds a certain critical threshold, as will be explained below.

(15) The regulation device 16, following the command signal RS, is moved according to a movement profile capable of compensating the actual variations of the level 14, both in terms of amplitude and also in terms of phase, also compensating the overall delay RC. The command signal RS can be capable of compensating the variation signal SV also in terms of frequency, for example in the case in which the frequency of the variation signal SV deviates from a real frequency that characterizes the disturbance of the plant 10. In particular, the command signal RS generated by the control unit 18 is added to the main control signal to adjust the frequency, the phase or the amplitude of the movement profile of the regulation device 16 in order to minimize the disturbance of the plant 10. The variation signal SV has a substantially oscillating trend and can be represented as a sinusoid with amplitude A.sub.d, phase .sub.d, frequency f.sub.d. In mathematical terms, the variation signal SV can be written in the following form: A.sub.dsen(t+.sub.d), where =2f.sub.d is the pulsation. Since the variation signal SV can be characterized by a band of frequencies, the control unit 18, or the detection device 19 itself, can determine a main frequency f.sub.d intended as the frequency, or one of the frequencies, which characterize the oscillatory phenomenon of the disturbance of the plant 10 in a prevalent manner.

(16) The control signal CS generated therefore also has an oscillating trend and can be represented as a sinusoid with amplitude A.sub.c, phase .sub.c, frequency f.sub.c, where the frequency fc of the control signal CS is equal to the frequency f.sub.d of the variation signal SV. In mathematical terms, the control signal CS can be written in the form: A.sub.csen(t+.sub.c).

(17) Similarly, the command signal RS can be represented as a sinusoid with amplitude A.sub.r, phase .sub.r, frequency f.sub.r, where the frequency f.sub.r of the command signal RS is equal to the frequency f.sub.d of the variation signal SV.

(18) The operation of the plant 10, which corresponds to the regulation method, according to the present invention, provides to regulate the flow rate Q of the metal L in order to keep the level 14 in the mould 12 substantially stable.

(19) In particular, the method comprises a detection step in which the detection device 19 detects a punctual variation of the level 14 with respect to a nominal value and generates a corresponding variation signal SV, having a frequency f.sub.d, an amplitude A.sub.d and a phase .sub.d, which is sent to the control unit 18.

(20) The method can also comprise a step of observing the disturbance acting on the plant 10. The disturbance observation step has the sole purpose of defining when/if it is necessary to generate the command signal RS to be combined with the main control signal. Contrary to the state of the art, the present invention does not provide any estimate of a disturbance signal to be used in the control loop for its reduction. In other words, observing the disturbance is not required for the purposes of the compensation, but only to understand whether to compensate or not. The method also comprises a control step, in which the control unit 18, in response to the variation signal SV and in accordance with the at least one first algorithm ALG1, selectively generates a command signal RS for the regulation device 16 starting from the variation signal SV and taking into account the overall delay RC between the generation of the variation signal SV and the actual actuation of the regulation device 16. The command signal RS has at least a variable amplitude A.sub.r of its own and a variable phase .sub.r of its own.

(21) In particular, the command signal RS has, instant by instant, its own amplitude A.sub.r equal to the amplitude A.sub.d of the variation signal SV and its own phase .sub.r opposite to the phase .sub.d of the variation signal SV.

(22) The command signal RS thus generated allows to drastically reduce the response delay that occurs between the measurement of the variation signal SV and the actual variation of flow rate Q of metal L obtained with the movement of the regulation device 16.

(23) The control step comprises a calculation sub-step in which the control unit 18, before calculating the command signal RS, processes the variation signal SV and generates, in an iterative manner and instant by instant, the control signal CS having an amplitude A.sub.c and phase c, of its own, which serves as the basis for the generation of the command signal RS, together with a correction value generated on the basis of the first compensation algorithm ALG1 which takes into account a mathematical system associated with the overall delay RC.

(24) The above mathematical system is considered in the frequency domain and associates a transfer function having a gain G.sub.s, which is a function of A.sub.s and takes into account an amplitude distortion, and a phase .sub.s which takes into account the overall delay RC and which, with the present method, we intend to identify in order to achieve high control performances.

(25) In this specific case, the control signal CS is an input signal to the mathematical system while the command signal RS is an output signal of the mathematical system and can be expressed in the form RS=A.sub.rsen(t+.sub.r)=A.sub.cA.sub.ssen(t++.sub.s).

(26) The mathematical system as above can represent, for example, all the functional components/objects comprised between the detection of the variation in level 14 and the actual variation of flow rate Q of metal L in the mould 12 in a limited and well-defined portion of the spectrum.

(27) The command signal RS can be expressed with respect to the variation signal SV as RS=A.sub.cA.sub.ssen(t+.sub.c+.sub.s)=A.sub.dsen(t+.sub.d).

(28) The at least one first compensation algorithm ALG1 is configured to solve the equivalence A.sub.cA.sub.ssen(t+.sub.c+.sub.s)=A.sub.dsen(t+.sub.d) in order to obtain the amplitude A.sub.c and the phase .sub.c of the control signal CS. The control unit 18, by means of the at least one first compensation algorithm ALG1, calculates first an error function E defined by the difference between the variation signal SV and the command signal RS and expressed as E=A.sub.cA.sub.ssen(t+.sub.c+.sub.s)+A.sub.dsen(t+.sub.d), and then, given the complexity of the error function E, it calculates a cost function J proportional to the square of the error function E which is minimized in order to find the optimal values of amplitude A.sub.c and phase .sub.c of the control signal CS.

(29) The optimization of the cost function J can occur through any known optimization algorithm whatsoever, for example a gradient descent algorithm known as batch gradient descent, stochastic gradient descent, mini-batch gradient descent, or algorithms based on the estimation of the moments of the gradient, such as NAG, AdaGrad, RMSProp, AdaDelta, AdaMax, Nadam, AMSGrad, kSGD, ADAM.

(30) According to one possible embodiment of the method, the compensation algorithm ALG1 can also be configured to compensate the frequency f of the control signal CS in the event that the initial estimate/choice of the frequency f.sub.d is incorrect.

(31) According to some embodiments, the first compensation algorithm ALG1 can be implemented with an artificial intelligence, for example a neural network. This neural network comprises an input layer containing information relating to the frequency fc of the control signal CS to be calculated, a hidden layer containing the functions of activation of the oscillatory phenomenon, sine and cosine, weighted with the orthogonal components a.sub.c and b.sub.c relating to the control signal CS A.sub.csen(t+.sub.c), and an output layer which is a combination of the control signal CS.

(32) According to some embodiments, the at least one first algorithm ALG1 generates the control signal CS only if the variation signal SV has a frequency f.sub.d greater than or around 0.2 Hz.

(33) Otherwise, that is, for frequencies lower than about 0.2 Hz, the variation signal SV can be compensated in the traditional way, for example with the use of a second compensation algorithm ALG2, also stored in the memory unit 21 and configured to calculate the amplitude A.sub.c and the phase .sub.c of the control signal CS without iteratively considering the characteristics of the mathematical system of the plant 10. In other words, for frequencies lower than about 0.2 Hz, the second compensation algorithm ALG2 does not consider the delays and non-linearities of the plant 10, which for these frequencies typically do not affect the control of the regulation device 16.

(34) The graphs schematically shown in FIG. 2 show an application of the method described above using the plant 10, when the variation signal SV, which corresponds to a measurement of the disturbances, has a high frequency, for example around or above approximately 1 Hz. When the variation signal SV, graph (c), indicative of variations of the level 14, which in this case is expressed in terms of amplitude, exceeds a determinate threshold value F1, the first compensation algorithm ALG1 begins to generate a control signal CS, processing it into a command signal RS, graph (b), which compensates, that is, decreases, the variation signal SV. When the variation signal SV has returned within a preestablished threshold, for example below the threshold value F1, and the control signal CS is sufficiently weak, the first compensation algorithm ALG1 stops generating the control signal CS. The process is repeated when, subsequently, the variation signal SV once again exceeds the threshold value F1. Considering the amplitude of the control signal CS in the self-shutdown logic allows to compensate the MLFs in an optimal manner and not generate possible anomalous peaks, known to persons of skill in the art as overshoot phenomena, caused by a sudden shutdown. As can be seen from graph (d), which shows the casting speed S, the compensation of the variation signal SV occurs without there being a decrease in the casting speed S and therefore without any impact on the productivity of the plant 10.

(35) The graphs shown schematically in FIG. 3 show another application of the method described above using the plant 10, when the variation signal SV, which corresponds to a measurement of the disturbances, has a frequency around approximately 0.5 Hz. In this case, the disturbances may be indicative of the presence of bulging. When the variation signal SV, graph (c), exceeds a determinate threshold value F2, the first compensation algorithm ALG1 starts to generate a control signal CS, processing it into a command signal RS, graph (b), which compensates, that is, decreases, the variation signal SV. In this case, the command signal RS is continuous and has an amplitude that decreases over time. As can be seen from graph (d), as the casting speed S increases, the variation signal SV is optimally compensated without needing to interrupt the process.

(36) It is clear that modifications and/or additions of parts may be made to the plant and to the regulation method as described heretofore, without departing from the field and scope of the present invention, as defined by the claims.

(37) It is also clear that, although the present invention has been described with reference to some specific examples, a person of skill in the art shall certainly be able to achieve many other equivalent forms of casting plant 10 and corresponding regulation method, having the characteristics as set forth in the claims and hence all coming within the field of protection defined thereby.

(38) In the following claims, the sole purpose of the references in brackets is to facilitate their reading and they must not be considered as restrictive factors with regard to the field of protection defined by the same claims.