A method for reducing driving energy consumption of a continuous casting segment

Abstract

The application discloses a method for reducing driving energy consumption of a continuous casting fan-shaped section. According to historical data for eliminating bulging or reducing bulging, a nonlinear data relationship between slab state parameters and driving force is learned, so that a new driving force corresponding to each driving roller can be determined according to current slab state parameters, thereby the bulging can be reduced or eliminated, and the occurrence of a slab stagnation accident caused by the bulging can be avoided. In addition, due to the bulging, the rotating speeds of adjacent fan-shaped section rollers are different, the bulging continuously occurs or slab quality problems caused by the bulging occur, the continuous casting machine works in an abnormal state for a long time, and thus energy is wasted.

Description

Technical Field

[0001] This invention belongs to the field of metallurgical digital technology, specifically relating to a method for reducing the driving energy consumption of the sector section in continuous casting. Background Technology

[0002] In continuous casting production, problems with the sector section equipment or quality issues such as slab bulging can lead to increased resistance during slab pulling at certain stages. This can result in different rotation speeds of the rollers in adjacent sector sections. Proceeding under these conditions can lead to product quality problems, waste of energy as the slower-speed motor becomes a load on other motors, or even slab stalling due to excessive pulling resistance. Summary of the Invention

[0003] In view of the above-mentioned shortcomings in the prior art, the present invention provides a method for reducing the driving energy consumption of the continuous casting sector, which solves the problems existing in the prior art.

[0004] To achieve the above-mentioned objectives, the technical solution adopted by this invention is as follows:

[0005] A method for reducing the driving energy consumption of the sector section in continuous casting includes:

[0006] The working status parameters of the drive rollers in multiple sector sections of the continuous casting machine and the preset billet drawing line speed are acquired in real time. The working status parameters of the drive rollers include the line speed of the drive rollers and the torque change.

[0007] Based on the working parameters of the drive roller and the preset drawing line speed, the bulging of the slab is monitored to obtain the bulging monitoring results, which include whether there is no bulging or whether there is bulging.

[0008] If the bulging detection result indicates the presence of bulging, then based on the working state parameters of the drive roller and the preset drawing line speed, the drive roller corresponding to the bulging position is obtained, and the first target drive roller is obtained.

[0009] Obtain the slab state parameters and predict the optimal driving force of the first target drive roller based on the slab state parameters;

[0010] Based on the optimal driving force of the first target drive roller, the driving force of the first target drive roller is adjusted to reduce the driving energy consumption of the continuous casting sector.

[0011] In one possible implementation, after adjusting the driving force of the first target drive roller according to the optimal driving force of the first target drive roller, the method further includes:

[0012] Determine at least one drive roller in front of the first target drive roller to obtain the second target drive roller;

[0013] Determine at least one drive roller following the first target drive roller to obtain the third target drive roller;

[0014] Based on the slab state parameters, predict the optimal driving force of the second target drive roller and the optimal driving force of the third target drive roller.

[0015] The driving force of the second target drive roller is adjusted according to the optimal driving force of the second target drive roller;

[0016] The driving force of the third target drive roller is adjusted according to the optimal driving force of the third target drive roller.

[0017] In one possible implementation, based on the operating parameters of the drive roller and a preset drawing speed, bulging monitoring of the slab is performed to obtain bulging monitoring results, including:

[0018] Based on the linear velocity in the working state parameters of the drive roller and the preset drawing linear velocity, obtain the linear velocity difference between the drive roller linear velocity and the preset drawing linear velocity.

[0019] Determine whether the difference in linear velocity exceeds a preset first threshold and whether the change in the drive roller exceeds a preset second threshold. If so, determine that the bulging detection result indicates the presence of bulging; otherwise, determine that the bulging detection result indicates the absence of bulging.

[0020] In one possible implementation, if the bulge monitoring result indicates the presence of bulge, then based on the driving roller's operating state parameters and a preset drawing line speed, the driving roller corresponding to the bulge position is obtained to acquire the first target driving roller, including:

[0021] If the bulging detection result indicates the presence of bulging, then determine the difference in linear velocity between the drive roller linear velocity and the preset drawing linear velocity;

[0022] Remove the drive rollers whose linear velocity difference exceeds the first threshold, and identify the drive roller with the largest torque change among the removed drive rollers.

[0023] The drive roller with the largest torque change is selected as the drive roller corresponding to the bulge position, thus obtaining the first target drive roller.

[0024] In one possible implementation, obtaining slab state parameters and predicting the optimal driving force for the first target drive roller based on the slab state parameters includes:

[0025] Obtain slab state parameters, including the calculation results of the secondary cooling model, slab solidification state, slab shell thickness, slab shell temperature data, and abnormal conditions of the secondary cooling nozzles. The abnormal conditions of the secondary cooling nozzles include malfunction or normal operation of the secondary cooling nozzles. The malfunction of the secondary cooling nozzles includes blockage or breakage of the secondary cooling nozzles.

[0026] Feature data were constructed based on the calculation results of the secondary cooling model, the solidification state of the billet, the billet shell thickness, the billet shell temperature data, and the abnormal conditions of the secondary cooling nozzle.

[0027] The feature data is input into the pre-trained first driving force prediction model to obtain the optimal driving force corresponding to the first target driving roller.

[0028] In one possible implementation, the method for obtaining the pre-trained first driving force prediction model includes:

[0029] A radial basis function neural network model was used as the first driving force prediction model, and the weights of the first driving force prediction model were initialized multiple times to obtain multiple weight sequences corresponding to the first driving force prediction model. Among them, W i Let i represent the i-th weight sequence, i = 1, 2, ..., M, where M represents the total number of weight sequences. W represents the i-th weight sequence. i The j-th weight in the sequence, j = 1, 2, ..., N, where N represents the i-th weight sequence W. i Total number of medium weights;

[0030] Acquire training data, which includes historical slab state parameters and corresponding optimal driving forces;

[0031] Multiple weight sequences are randomly divided into two parts, with the first weight in one part serving as the searcher and the second weight in the other part serving as the follower.

[0032] Based on the training data, the searcher is perturbed and updated to obtain the updated searcher;

[0033] Based on the training data, the followers are adaptively updated to obtain the updated followers;

[0034] If the number of updates has reached the maximum number of iterations, then the weight sequence with the highest fitness among the searchers and followers is used as the weight parameter of the first driving force prediction model to obtain the trained first driving force prediction model. Otherwise, the searchers and followers are re-divided and the next round of training is carried out.

[0035] In one possible implementation, initializing the weights of the first driving force prediction model includes:

[0036] A1. Determine the upper and lower limits of the weights corresponding to the weights of the first driving force prediction model, and set the first counter j = 1 and the second counter i = 1;

[0037] A2. Randomly generate an initial parameter value between the upper and lower weight limits.

[0038] A3. Based on the initial parameter values The j-th weight in the first driving force prediction model is determined as follows:

[0039]

[0040] in, denoted as the j-th weight in the first driving force prediction model, max represents the upper limit of the weight, and min represents the lower limit of the weight;

[0041] A4. Determine the (j+1)th initial parameter value based on the j-th initial parameter value:

[0042]

[0043] Where γ represents the constant calculation factor, γ = 2;

[0044] A5. Increment the count value of the first counter j by one, and determine whether the first counter j is equal to the number of weights N of the first driving force prediction model. If so, obtain the weight sequence. Proceed to step A6; otherwise, return to step A3.

[0045] A6. Determine whether the count value of the second counter i is equal to the total number of weight sequences M. If yes, obtain the M weight sequences and complete the initialization; otherwise, return to step A1.

[0046] In one possible implementation, the searcher is perturbed and updated based on training data to obtain the updated searcher, including:

[0047] Historical slab state parameters are used as input data for the first driving force prediction model, and the optimal driving force corresponding to the historical slab state parameters is used as the expected output of the first driving force prediction model. The l-th searcher W is then used as the input data. l Loaded into the first driving force prediction model, the actual output of the first driving force prediction model is obtained; Indicates the total number of searchers;

[0048] Based on the expected output and actual output of the first driving force prediction model, determine the l-th searcher W. l The corresponding error function value is used, and the negative of the error function value is taken as the l-th searcher W. l Corresponding fitness;

[0049] After obtaining the fitness of each searcher, determine the searcher with the highest fitness and use that searcher as the optimal weight sequence F. best ;

[0050] Based on the optimal weight sequence F bestA perturbation update is performed on the remaining searchers, which is as follows:

[0051]

[0052]

[0053]

[0054] in, This indicates the l-th searcher W l The updated weight F bestj Let κ1 represent the j-th weight in the optimal weight sequence, κ2 represent the first random number in the interval (0,1), κ3 represent the third random number in the interval (0,1), κ4 represent the fourth random number in the interval (0,1), sign(κ4) represents the first intermediate parameter, α(κ4) represents the second intermediate parameter, η represents a constant, η=0.0001, and α' represents the decision probability.

[0055] In one possible implementation, the followers are adaptively updated based on the training data to obtain the updated followers, including:

[0056] Historical slab state parameters are used as input data for the first driving force prediction model, and the optimal driving force corresponding to the historical slab state parameters is used as the expected output of the first driving force prediction model. The h-th follower W is then used as the input data. h Loaded into the first driving force prediction model, the actual output of the first driving force prediction model is obtained; Indicates the total number of followers;

[0057] Based on the expected output and actual output of the first driving force prediction model, determine the h-th follower W. h The corresponding error function value is used, and the negative of the error function value is taken as the h-th follower W. h Corresponding fitness F h ;

[0058] According to the h-th follower W h Corresponding fitness F h For the h-th follower W h An update will be performed, and the update is as follows:

[0059]

[0060] in, W represents the h-th follower. h The j-th weight in the middle, Indicates the updated C represents the fifth random number that follows a Cauchy distribution, F avg This represents the average fitness of all followers. W represents the h-th follower. h The j-th weight; when h = 1, Set as This represents the weight update factor.

[0061] In one possible implementation, the weight update factor for:

[0062]

[0063] in, This represents the initial value of the weight update factor. ε represents the final value of the weight update factor, exp represents the exponential function with the natural constant e as the base, ε represents the nonlinear control parameter, t represents the current iteration number, and T represents the maximum iteration number.

[0064] This invention provides a method for reducing the driving energy consumption of the sector segment in continuous casting. By monitoring the bulging of the slab to determine the occurrence of bulging behavior, and by using existing historical data to learn the relationship between the driving force of the sector segment and bulging, the driving force of the sector segment can be adjusted according to the current working state to eliminate the influence of the driving force on bulging, reduce or eliminate bulging, and reduce energy waste or slab stagnation accidents caused by continuous bulging. Attached Figure Description

[0065] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0066] Figure 1 The flowchart illustrates a method for reducing the driving energy consumption of the continuous casting sector segment, as provided in an embodiment of the present invention.

[0067] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0068] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0069] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0070] like Figure 1 As shown, this embodiment of the invention provides a method for reducing the driving energy consumption of the continuous casting sector, comprising:

[0071] S1. Real-time acquisition of the working status parameters of the drive rollers in multiple sector sections of the continuous casting machine and the preset billet drawing linear speed, wherein the working status parameters of the drive rollers include the linear speed of the drive rollers and the torque change.

[0072] During production, the preset drawing speed should match the slab state parameters. However, when bulging occurs, the linear speed and torque of the drive roller will change. Therefore, this can be used as a monitoring parameter for bulging to make a real-time judgment on bulging.

[0073] S2. Based on the working state parameters of the drive roller and the preset drawing line speed, monitor the bulging of the slab to obtain the bulging monitoring results, which include whether there is no bulging or whether there is bulging.

[0074] When there is a difference between the linear speed of the drive roller and the preset drawing linear speed, it can be considered that bulging may occur. In this case, by combining the judgment with the torque change of the drive roller, it can be determined whether bulging exists. This dual judgment method can effectively avoid misjudgment.

[0075] Optionally, in addition to the methods described above for monitoring bulging, bulging can also be monitored based on the operating parameters during the continuous casting process, for example:

[0076] The amount of bulging deformation δ can be obtained in real time:

[0077]

[0078]

[0079]

[0080] Where η1 represents the correction coefficient, a n The shape factor is represented by p, the static pressure of the molten steel is represented by l, and the roll gap is represented by t. sThe time taken for the roll to pass through the roll gap l is represented by E, the elastic modulus by s, and the blank thickness by T. so T represents the solidification temperature of steel. m The average temperature of the slab shell, T s This indicates the surface temperature of the slab.

[0081] When the bulging deformation δ is greater than a certain value, it can be determined that the slab has bulging.

[0082] S3. If the bulging detection result indicates the presence of bulging, then based on the working state parameters of the drive roller and the preset drawing line speed, obtain the drive roller corresponding to the bulging position to obtain the first target drive roller.

[0083] The torque of the drive roller where the bulge is located will change significantly. Therefore, based on this, we can determine the drive roller where the bulge is located at the current time, and then adjust the driving force based on this drive roller.

[0084] S4. Obtain the slab state parameters and predict the optimal driving force of the first target drive roller based on the slab state parameters.

[0085] S5. Adjust the driving force of the first target drive roller according to the optimal driving force of the first target drive roller to reduce the driving energy consumption of the continuous casting sector.

[0086] This invention provides a method for reducing the driving energy consumption of the continuous casting sector section. Based on historical data on eliminating or reducing bulging, the nonlinear data relationship between slab state parameters and driving force is learned. This allows for the determination of the new driving force for each driving roller based on the current slab state parameters, thereby reducing or eliminating bulging and preventing slab stagnation accidents caused by bulging. Furthermore, bulging causes different rotation speeds among adjacent sector section rollers, leading to persistent bulging or slab quality problems, causing the continuous casting machine to operate in an abnormal state for extended periods, resulting in energy waste.

[0087] In one possible implementation, after adjusting the driving force of the first target drive roller according to the optimal driving force of the first target drive roller, the method further includes:

[0088] Determine at least one drive roller in front of the first target drive roller to obtain the second target drive roller.

[0089] Determine at least one drive roller behind the first target drive roller to obtain the third target drive roller.

[0090] Based on the slab state parameters, predict the optimal driving force of the second target drive roller and the optimal driving force of the third target drive roller.

[0091] The driving force of the second target drive roller is adjusted according to the optimal driving force of the second target drive roller.

[0092] The driving force of the third target drive roller is adjusted according to the optimal driving force of the third target drive roller.

[0093] Once a bulge forms, it affects not only the drive roller where the bulge is located but also the operation of several adjacent drive rollers. Therefore, adjustments to these adjacent drive rollers are necessary. For example, based on the drive roller where the bulge is located, a first model is trained according to the slab state parameters to determine the driving force of the drive roller where the bulge is located. Then, for each of the adjacent drive rollers on both sides, a second model is trained to determine the driving force of the adjacent drive rollers. The first and second models are trained using slab state parameters from a historical database and historical data of manually adjusted driving forces.

[0094] Optionally, the bulge will not remain in a fixed position but will continuously move. Therefore, after detecting the bulge, its position can be monitored in real time. When the bulge moves to the next drive roller, it will obtain driving force again to ensure the normal operation of billet pulling, thereby reducing the overall energy consumption of the continuous casting system.

[0095] In one possible implementation, based on the operating state parameters of the drive roller and a preset drawing line speed, bulging monitoring of the slab is performed to obtain bulging monitoring results. This includes: obtaining the linear velocity difference between the drive roller linear velocity and the preset drawing line speed based on the linear velocity in the drive roller operating state parameters and the preset drawing line speed; determining whether the linear velocity difference exceeds a preset first threshold and whether the change in drive roller speed exceeds a preset second threshold; if so, the bulging monitoring result is determined to indicate the presence of bulging; otherwise, the bulging monitoring result is determined to indicate the absence of bulging.

[0096] In one possible implementation, if the bulging detection result indicates the presence of bulging, then based on the working state parameters of the drive roller and the preset drawing line speed, the drive roller corresponding to the bulging position is obtained to obtain the first target drive roller, including: if the bulging detection result indicates the presence of bulging, determining the linear velocity difference between the drive roller line speed and the preset drawing line speed; removing drive rollers whose linear velocity difference exceeds a first threshold, and determining the drive roller with the largest torque change among the removed drive rollers; and using the drive roller with the largest torque change as the drive roller corresponding to the bulging position to obtain the first target drive roller.

[0097] In one possible implementation, obtaining slab state parameters and predicting the optimal driving force for the first target drive roller based on the slab state parameters includes:

[0098] Obtain slab state parameters, including the calculation results of the secondary cooling model, slab solidification state, slab shell thickness, slab shell temperature data, and abnormal conditions of the secondary cooling nozzles. The abnormal conditions of the secondary cooling nozzles include malfunction or normal operation of the secondary cooling nozzles. The malfunction of the secondary cooling nozzles includes blockage or rupture of the secondary cooling nozzles.

[0099] Optionally, when the slab exhibits bulging, the cooling parameters can be redefined based on the secondary cooling model to suit the current production conditions. Once the secondary cooling model generates calculation results, these can be used as a basis for obtaining the driving force.

[0100] Feature data were constructed based on the calculation results of the secondary cooling model, the solidification state of the billet, the billet shell thickness, the billet shell temperature data, and the abnormal conditions of the secondary cooling nozzle.

[0101] In this embodiment, feature data refers to data that can be input into the first driving force prediction model. The calculation results of the secondary cooling model, the solidification state of the billet, the billet shell thickness, the billet shell temperature data, and the abnormal conditions of the secondary cooling nozzle can be connected and converted into data of fixed length. This ensures that the data length of each input to the first driving force prediction model is the same, making the final driving force prediction result more accurate.

[0102] Optionally, before constructing the feature data, the calculation results of the secondary cooling model, the solidification state of the cast billet, the billet shell thickness, the billet shell temperature data, and the abnormal conditions of the secondary cooling nozzle can be preprocessed. For example, text can be converted into numerical strings and normalized. It is worth noting that the calculation results of the secondary cooling model, the solidification state of the cast billet, and the abnormal conditions of the secondary cooling nozzle may not be numerical data. A mapping table between numbers and data can be pre-defined, allowing direct conversion into numerical strings. For example, an abnormal secondary cooling nozzle corresponds to the number 1; a normal secondary cooling nozzle corresponds to the number 0.

[0103] The feature data is input into the pre-trained first driving force prediction model to obtain the optimal driving force corresponding to the first target driving roller.

[0104] In one possible implementation, the method for obtaining the pre-trained first driving force prediction model includes:

[0105] A radial basis function neural network model was used as the first driving force prediction model, and the weights of the first driving force prediction model were initialized multiple times to obtain multiple weight sequences corresponding to the first driving force prediction model. Among them, W i Let i represent the i-th weight sequence, i = 1, 2, ..., M, where M represents the total number of weight sequences. W represents the i-th weight sequence. iThe j-th weight in the sequence, j = 1, 2, ..., N, where N represents the i-th weight sequence W. i Total weighted average.

[0106] Optionally, the first driving force prediction model of the present invention is mainly used to learn the nonlinear relationship between historical data and driving forces; in essence, it is a classification model. Therefore, in addition to the radial basis function neural network model described in the present invention, other classification neural networks can also be used as the first driving force prediction model and trained using the training method described in the embodiments of the present invention. For example, a back propagation (BP) neural network or a convolutional neural network (CNN) neural network can also be used as the first driving force prediction model.

[0107] Acquire training data, which includes historical slab state parameters and corresponding optimal driving forces; the optimal driving force refers to the manually adjusted driving force corresponding to the historical slab state parameters.

[0108] Multiple weight sequences are randomly divided into two parts. The first weight in one part is used as the searcher, and the second weight in the other part is used as the follower. Based on the training data, the searcher is perturbed and updated to obtain the updated searcher. Based on the training data, the follower is adaptively updated to obtain the updated follower. It is determined whether the number of updates has reached the maximum number of iterations. If so, the weight sequence with the highest fitness among the searcher and the follower is used as the weight parameters of the first driving force prediction model to obtain the trained first driving force prediction model. Otherwise, the searcher and the follower are re-divided, and the next round of training is carried out.

[0109] By dividing the weight sequence into two parts, while performing local optimum shrinkage, a better solution can be sought, which can not only effectively avoid getting trapped in local optima, but also accelerate the convergence speed.

[0110] In one possible implementation, initializing the weights of the first driving force prediction model includes:

[0111] A1. Determine the upper and lower limits of the weights corresponding to the weights of the first driving force prediction model, and set the first counter j = 1 and the second counter i = 1.

[0112] A2. Randomly generate an initial parameter value between the upper and lower weight limits.

[0113] A3. Based on the initial parameter values The j-th weight in the first driving force prediction model is determined as follows:

[0114]

[0115] in, denoted as the j-th weight in the first driving force prediction model, max represents the upper limit of the weight, and min represents the lower limit of the weight.

[0116] A4. Determine the (j+1)th initial parameter value based on the j-th initial parameter value:

[0117]

[0118] Where γ represents the constant calculation factor, γ = 2.

[0119] A5. Increment the count value of the first counter j by one, and determine whether the first counter j is equal to the number of weights N of the first driving force prediction model. If so, obtain the weight sequence. Proceed to step A6; otherwise, return to step A3.

[0120] A6. Determine whether the count value of the second counter i is equal to the total number of weight sequences M. If yes, obtain the M weight sequences and complete the initialization; otherwise, return to step A1.

[0121] Optionally, in addition to the initialization method described in the embodiments of the present invention, other initialization methods can also be used to initialize the weight sequence. For example, an upper or lower limit for the weights can be set, and for the first driving force prediction model, each weight of the first driving force prediction model can be randomly generated between the upper and lower weight limits.

[0122] In one possible implementation, the searcher is perturbed and updated based on training data to obtain the updated searcher, including:

[0123] Historical slab state parameters are used as input data for the first driving force prediction model, and the optimal driving force corresponding to the historical slab state parameters is used as the expected output of the first driving force prediction model. The l-th searcher W is then used as the input data. l Loaded into the first driving force prediction model, the actual output of the first driving force prediction model is obtained; This indicates the total number of searchers.

[0124] Based on the expected output and actual output of the first driving force prediction model, determine the l-th searcher W. l The corresponding error function value is used, and the negative of the error function value is taken as the l-th searcher W. l The corresponding fitness level.

[0125] After obtaining the fitness of each searcher, determine the searcher with the highest fitness and use that searcher as the optimal weight sequence F. best .

[0126] Based on the optimal weight sequence F best A perturbation update is performed on the remaining searchers, which is as follows:

[0127]

[0128]

[0129]

[0130] in, This indicates the l-th searcher W l The updated weight F bestj Let κ1 represent the j-th weight in the optimal weight sequence, κ2 represent the first random number in the interval (0,1), κ3 represent the third random number in the interval (0,1), κ4 represent the fourth random number in the interval (0,1), sign(κ4) represents the first intermediate parameter, α(κ4) represents the second intermediate parameter, η represents a constant, η = 0.0001, and α' represents the decision probability, which can be set to 0.3.

[0131] Alternatively, fitness can be obtained using the following methods:

[0132]

[0133] Where F represents fitness, Y m D represents the actual output of the first driving force prediction model corresponding to the m-th input data; m This represents the expected output of the first driving force prediction model corresponding to the m-th input data, i.e., the optimal driving force corresponding to the historical slab state parameters; m = 1, 2, ..., M, where M represents the total number of input data, i.e., the total number of historical slab state parameters.

[0134] In one possible implementation, the followers are adaptively updated based on the training data to obtain the updated followers, including:

[0135] Historical slab state parameters are used as input data for the first driving force prediction model, and the optimal driving force corresponding to the historical slab state parameters is used as the expected output of the first driving force prediction model. The h-th follower W is then used as the input data. h Loaded into the first driving force prediction model, the actual output of the first driving force prediction model is obtained; Indicates the total number of followers;

[0136] Based on the expected output and actual output of the first driving force prediction model, determine the h-th follower W. hThe corresponding error function value is used, and the negative of the error function value is taken as the h-th follower W. h Corresponding fitness F h ;

[0137] According to the h-th follower W h Corresponding fitness F h For the h-th follower W h An update will be performed, and the update is as follows:

[0138]

[0139] in, W represents the h-th follower. h The j-th weight in the middle, Indicates the updated C represents the fifth random number that follows a Cauchy distribution, F avg This represents the average fitness of all followers. W represents the h-th follower. h The j-th weight; when h = 1, Set as This represents the weight update factor.

[0140] In one possible implementation, the weight update factor for:

[0141]

[0142] in, This represents the initial value of the weight update factor. ε represents the final value of the weight update factor, exp represents the exponential function with the natural constant e as the base, ε represents the nonlinear control parameter, t represents the current iteration number, and T represents the maximum iteration number.

[0143] This invention provides a method for reducing the driving energy consumption of the sector segment in continuous casting. By monitoring the bulging of the slab to determine the occurrence of bulging behavior, and by using existing historical data to learn the relationship between the driving force of the sector segment and bulging, the driving force of the sector segment can be adjusted according to the current working state to eliminate the influence of the driving force on bulging, thereby reducing or eliminating bulging and reducing energy waste or slab stagnation accidents caused by continuous bulging.

[0144] It is worth noting that any method utilizing the inventive concept should fall within the scope of protection of this invention. Other embodiments of the invention will readily conceive of those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein.

[0145] It should be understood that the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A method for reducing the driving energy consumption of the sector section in continuous casting, characterized in that, include: The working status parameters of the drive rollers in multiple sector sections of the continuous casting machine and the preset billet drawing line speed are acquired in real time. The working status parameters of the drive rollers include the line speed of the drive rollers and the torque change. Based on the working parameters of the drive roller and the preset drawing line speed, the bulging of the slab is monitored to obtain the bulging monitoring results, which include whether there is no bulging or whether there is bulging. If the bulging detection result indicates the presence of bulging, then based on the working state parameters of the drive roller and the preset drawing line speed, the drive roller corresponding to the bulging position is obtained, and the first target drive roller is obtained. Obtain the slab state parameters and predict the optimal driving force of the first target drive roller based on the slab state parameters; Based on the optimal driving force of the first target drive roller, the driving force of the first target drive roller is adjusted to reduce the driving energy consumption of the continuous casting sector. After adjusting the driving force of the first target drive roller based on its optimal driving force, the process also includes: Determine at least one drive roller in front of the first target drive roller to obtain the second target drive roller; Determine at least one drive roller following the first target drive roller to obtain the third target drive roller; Based on the slab state parameters, predict the optimal driving force of the second target drive roller and the optimal driving force of the third target drive roller. The driving force of the second target drive roller is adjusted according to the optimal driving force of the second target drive roller; The driving force of the third target drive roller is adjusted according to the optimal driving force of the third target drive roller; Based on the operating parameters of the drive rollers and the preset drawing speed, bulging monitoring of the slab is performed, and the bulging monitoring results are obtained, including: Based on the linear velocity in the working state parameters of the drive roller and the preset drawing linear velocity, obtain the linear velocity difference between the drive roller linear velocity and the preset drawing linear velocity. Determine whether the difference in linear velocity exceeds a preset first threshold and whether the change in the drive roller exceeds a preset second threshold. If so, determine that the bulging detection result indicates the presence of bulging; otherwise, determine that the bulging detection result indicates the absence of bulging. If the bulge monitoring result indicates the presence of bulge, then based on the driving roller's operating state parameters and the preset drawing line speed, the driving roller corresponding to the bulge position is obtained, thus obtaining the first target driving roller, including: If the bulging detection result indicates the presence of bulging, then determine the difference in linear velocity between the drive roller linear velocity and the preset drawing linear velocity; Remove the drive rollers whose linear velocity difference exceeds the first threshold, and identify the drive roller with the largest torque change among the removed drive rollers. The drive roller with the largest torque change is taken as the drive roller corresponding to the bulge position, and the first target drive roller is obtained. Obtain the slab state parameters and, based on these parameters, predict the optimal driving force for the first target drive roller, including: Obtain slab state parameters, including the calculation results of the secondary cooling model, slab solidification state, slab shell thickness, slab shell temperature data, and abnormal conditions of the secondary cooling nozzles. The abnormal conditions of the secondary cooling nozzles include malfunction or normal operation of the secondary cooling nozzles. The malfunction of the secondary cooling nozzles includes blockage or breakage of the secondary cooling nozzles. Feature data were constructed based on the calculation results of the secondary cooling model, the solidification state of the billet, the billet shell thickness, the billet shell temperature data, and the abnormal conditions of the secondary cooling nozzle. The feature data is input into the pre-trained first driving force prediction model to obtain the optimal driving force corresponding to the first target driving roller.

2. The method for reducing the driving energy consumption of the continuous casting sector segment according to claim 1, characterized in that, The method for obtaining the pre-trained first driving force prediction model includes: A radial basis function neural network model was used as the first driving force prediction model, and the weights of the first driving force prediction model were initialized multiple times to obtain multiple weight sequences W corresponding to the first driving force prediction model. i ={ , ,…, ,…, }; where W i Let i represent the i-th weight sequence, i = 1, 2, ..., M, where M represents the total number of weight sequences. W represents the i-th weight sequence. i The j-th weight in the sequence, j=1,2,...,N, where N represents the i-th weight sequence W. i Total number of medium weights; Acquire training data, which includes historical slab state parameters and corresponding optimal driving forces; Multiple weight sequences are randomly divided into two parts, with the first weight in one part serving as the searcher and the second weight in the other part serving as the follower. Based on the training data, the searcher is perturbed and updated to obtain the updated searcher; Based on the training data, the followers are adaptively updated to obtain the updated followers; If the number of updates has reached the maximum number of iterations, then the weight sequence with the highest fitness among the searchers and followers is used as the weight parameter of the first driving force prediction model to obtain the trained first driving force prediction model. Otherwise, the searchers and followers are re-divided and the next round of training is carried out.

3. The method for reducing the driving energy consumption of the continuous casting sector segment according to claim 2, characterized in that, Initialize the weights of the first driving force prediction model, including: A1. Determine the upper and lower limits of the weights corresponding to the weights of the first driving force prediction model, and set the first counter j=1 and the second counter i=1; A2. Randomly generate an initial parameter value between the upper and lower weight limits; A3. Based on the initial parameter values, determine the j-th weight in the first driving force prediction model as follows: =min+(max-min) , in, denoted as the j-th weight in the first driving force prediction model, max represents the upper limit of the weight, and min represents the lower limit of the weight; A4. Determine the (j+1)th initial parameter value based on the j-th initial parameter value: = ， Where γ represents the constant calculation factor, γ=2; A5. Increment the count value of the first counter j by one, and determine whether the first counter j is equal to the number of weights N of the first driving force prediction model. If so, obtain the weight sequence W. i ={ , ,…, ,…, If the condition is met, proceed to step A6; otherwise, return to step A3. A6. Determine whether the count value of the second counter i is equal to the total number of weight sequences M. If yes, obtain the M weight sequences and complete the initialization; otherwise, return to step A1.

4. The method for reducing the driving energy consumption of the continuous casting sector segment according to claim 3, characterized in that, Based on the training data, the searcher is perturbed and updated to obtain the updated searcher, including: Historical slab state parameters are used as input data for the first driving force prediction model, and the optimal driving force corresponding to the historical slab state parameters is used as the expected output of the first driving force prediction model. The l-th searcher W... l Loaded into the first driving force prediction model, the actual output of the first driving force prediction model is obtained; l=1,2,…, , Indicates the total number of searchers; Based on the expected output and actual output of the first driving force prediction model, determine the l-th searcher W. l The corresponding error function value is used, and the negative of the error function value is taken as the l-th searcher W. l Corresponding fitness; After obtaining the fitness of each searcher, determine the searcher with the highest fitness and use that searcher as the optimal weight sequence F. best ; Based on the optimal weight sequence F best A perturbation update is performed on the remaining searchers, which is as follows: = ， a(k4)= , sign(κ4)= , in, This indicates the l-th searcher W l The updated weight F bestj Let κ1 represent the j-th weight in the optimal weight sequence, κ2 represent the first random number in the interval (0,1), κ3 represent the third random number in the interval (0,1), κ4 represent the fourth random number in the interval (0,1), sign(κ4) represents the first intermediate parameter, α(κ4) represents the second intermediate parameter, η represents a constant, η=0.0001, and α' represents the decision probability.

5. The method for reducing the driving energy consumption of the continuous casting sector segment according to claim 4, characterized in that, Based on the training data, the followers are adaptively updated to obtain the updated followers, including: Historical slab state parameters are used as input data for the first driving force prediction model, and the optimal driving force corresponding to the historical slab state parameters is used as the expected output of the first driving force prediction model. The h-th follower W is then used as the input data. h Loaded into the first driving force prediction model, the actual output of the first driving force prediction model is obtained; h=1,2,…, , Indicates the total number of followers; Based on the expected output and actual output of the first driving force prediction model, determine the h-th follower W. h The corresponding error function value is used, and the negative of the error function value is taken as the h-th follower W. h Corresponding fitness F h ; According to the h-th follower W h Corresponding fitness F h For the h-th follower W h An update will be performed, and the update is as follows: = ， in, W represents the h-th follower. h The j-th weight in the middle, Indicates the updated C represents the fifth random number that follows a Cauchy distribution, and F avg This represents the average fitness of all followers. W represents the h-th follower. h The j-th weight; when h=1, Set as φ represents the weight update factor.

6. The method for reducing the driving energy consumption of the continuous casting sector segment according to claim 5, characterized in that, The weight update factor φ is: φ=φ0+(φ E -φ0)×exp(-(εt / T) 2 ) Where φ0 represents the initial value of the weight update factor, φ E ε represents the final value of the weight update factor, exp represents the exponential function with the natural constant e as the base, ε represents the nonlinear control parameter, t represents the current iteration number, and T represents the maximum iteration number.

Images