Data-driven prediction and decision-making method for strip steel continuous backing deviation

Abstract

The present invention discloses a data-driven method for predicting and making decisions on the deviation of continuous annealing of variable specifications of plate and strip steel, including: collecting historical production data of a continuous annealing unit and its preceding cold rolling unit; wherein the historical production data includes deviation amount data and deviation inducing factor data affecting the deviation amount in historical production; training a deviation prediction model based on the historical production data; wherein the input of the deviation prediction model is the deviation inducing factor data, and the output is the deviation amount data; using the trained deviation prediction model to obtain the deviation amount prediction result of the strip steel to be produced; when it is predicted that the strip steel to be produced will deviate, based on the deviation inducing factor data and the deviation amount prediction result corresponding to the strip steel to be produced, selecting the deviation correction decision that meets the requirements of the strip steel to be produced from the historical deviation correction decision data. The present invention can solve the technical problem that the prior art is difficult to cover multi-dimensional influencing factors and cope with the frequent changes in the specifications of incoming materials, thereby ensuring the stable plate passing of the continuous annealing process of plate and strip steel.

Description

Technical Field

[0001] The present invention relates to the field of metallurgical technology, and in particular to a data-driven method for predicting and making decisions on the deviation of continuous backing of plate and strip steel with varying specifications. Background Art

[0002] As an important process for tempering heat treatment after strip rolling, continuous annealing is an essential preparation process for plate and strip steel products. After large plastic deformation during cold rolling, the grains of the plate and strip steel are refined, the dislocation density increases, and the deformation resistance is greatly improved. A series of process treatments such as heating, soaking, slow cooling, fast cooling, aging, water quenching, and leveling are required to reduce its deformation resistance and provide it for use in downstream industries.

[0003] As an important carrier of the above-mentioned process, the continuous annealing unit can complete the above-mentioned heat treatment while the strip is passing through the furnace. However, the long-distance strip transmission is very susceptible to the external stress field, internal residual stress, thermal stress and other forces, and then various deviations occur in the furnace. The deviation type, deviation direction and deviation degree are not only affected by the continuous annealing base material, but also by the intervention of the characteristics of the continuous annealing unit, thus presenting the phenomenon of "different units, different deviation rules".

[0004] In response to this problem, steel companies and scholars in the industry have proposed a series of strip deviation prediction methods. The invention patent with application number 201910934118.4 discloses a strip deviation method at the entrance of a continuous annealing unit based on the lateral distribution function of the internal stress of the strip; the invention patent with application number 201410751949.5 discloses a strip deviation method for a continuous annealing unit based on deviation factor calculation; the invention patent with application number 202210062546.4 discloses a continuous annealing deviation prediction method based on a neural network.

[0005] The above schemes all provide a basis for the prediction of strip deviation in continuous annealing units, but the above methods all adopt the evaluation method of influencing factors with equal weight or similar weight. Although such methods can be used to predict deviation behavior, it is difficult to fully present the comprehensive influence of multi-dimensional parameters of continuous annealing units, and the prediction results are difficult to fully reflect the actual deviation situation. Therefore, it is urgent to propose a prediction and decision-making method for continuous annealing deviation of plate and strip steel that can fully reflect multi-dimensional influencing factors and cope with frequent changes in incoming material specifications. Summary of the invention

[0006] The present invention provides a data-driven method for predicting and making decisions on the deviation of continuous back-off of variable-specification plate and strip steel, so as to solve the problem that the prior art is difficult to cover multi-dimensional influencing factors and cope with frequent changes in the specifications of incoming materials.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] On the one hand, the present invention provides a data-driven method for predicting and making decisions on the deviation of continuous strip steel with variable specifications. The data-driven method for predicting and making decisions on the deviation of continuous strip steel with variable specifications includes:

[0009] Collecting historical production data of the continuous annealing unit and its preceding cold rolling unit; wherein the historical production data includes the deviation amount data and the deviation inducing factor data that affect the deviation amount in the historical production process;

[0010] Based on the historical production data, a preset deviation prediction model is trained; wherein the input of the deviation prediction model is the deviation factor data, and the output is the deviation amount data;

[0011] The deviation prediction result of the strip to be produced is obtained by using the trained deviation prediction model;

[0012] When it is predicted that the strip to be produced will deviate, based on the deviation-inducing factor data and deviation amount prediction results corresponding to the strip to be produced, a correction decision that is suitable for the strip to be produced is screened out from the historical correction decision data.

[0013] Furthermore, the historical production data of the continuous annealing unit and its preceding cold rolling unit are collected, including:

[0014] According to the steel coil number and steel grade, the historical production data and historical deviation data of the continuous annealing unit of the corresponding steel coil are collected; wherein the historical production data of the continuous annealing unit includes the posture data of each furnace roller, the deviation correction data of each deviation correction roller, the continuous annealing speed data and the continuous annealing tension percentage data of the continuous annealing unit in the historical production process; wherein the furnace roller posture data includes the furnace roller axis crossing angle and the furnace roller axis offset; the historical deviation data includes the preheating section deviation data, the heating section deviation data and the final deviation data of the strip steel of the continuous annealing unit in the historical production process;

[0015] Perform mutation detection on the collected historical production data and historical deviation data of the continuous annealing unit, and remove the mutation value; use the historical production data of the continuous annealing unit after removing the mutation value as the unit deviation inducement data;

[0016] According to the steel coil number and steel grade, collect the historical production data of the cold rolling unit corresponding to the steel coil; wherein the historical production data of the cold rolling unit includes the distribution of the strip width, strip thickness and full-length plate shape;

[0017] According to the full-length plate shape distribution, each strip section is fitted according to a homogeneous polynomial without cubic terms, and the distribution function of its linear coefficient along the plate length direction is extracted; the belt width, strip thickness and the distribution function of the extracted linear coefficient along the plate length direction are used as the incoming material deflection induction data;

[0018] According to the steel coil number and steel grade, the historical furnace temperature control parameters of the continuous annealing unit of the corresponding steel coil are collected; wherein the furnace temperature control parameters include the preset temperature value of each section and the axial temperature difference value of each section; the furnace temperature control parameters are used as the environmental decoy data;

[0019] The collected historical deviation data, unit deviation data, incoming material deviation data and environmental deviation data are matched one by one according to the coil number and steel grade to form an integrated historical data set;

[0020] Deduplication is performed on the data in the obtained historical data set to obtain a deduplication data set;

[0021] Based on the deduplicated data set, the weight coefficient of each indicator is calculated; wherein, the weight coefficient is used to characterize the influence of the corresponding indicator on the deviation amount; the indicators include the posture of each furnace roller of the continuous annealing unit, the deviation amount of each deviation correction roller, the continuous annealing speed, the continuous annealing tension percentage, the preset temperature value of each section and the axial temperature difference value of each section, as well as the deviation amount of the preheating section and the heating section of the strip, and the distribution function of the bandwidth, strip thickness and first-order coefficient in the cold rolling unit along the plate length direction;

[0022] Screen each indicator according to the weight coefficient, and use the screened indicators as decoy factors;

[0023] The data that are bias-inducing factors are selected from the deduplicated data set to form a deviation-inducing data set.

[0024] Furthermore, deduplication of the data in the obtained historical data set includes:

[0025] Compare the strip shape distribution functions of each section, preheating section deviation data, heating section deviation data and final deviation data of strips of the same brand, width and thickness. If the difference between the strips is less than the preset difference threshold, the data corresponding to the strip is considered to be repeated, and only one of the strip data is retained.

[0026] Furthermore, the weight coefficient of each indicator is calculated based on the deduplicated data set, including:

[0027] Based on the deduplicated data set, a random forest algorithm model was established with the preheating section deviation, heating section deviation and final deviation as target values, and the weight coefficient of each indicator was calculated.

[0028] Furthermore, the step of filtering out data belonging to bias-inducing factors from the deduplicated data set includes:

[0029] If the weight coefficient of an indicator exceeds the preset weight threshold, it is judged to be a bias factor;

[0030] If the weight coefficient of the indicator does not exceed the preset weight threshold, it is judged not to be a bias factor.

[0031] Furthermore, the deviation prediction model is a particle swarm algorithm model.

[0032] Furthermore, based on the historical production data, a preset deviation prediction model is trained, including:

[0033] The deviation induction data set is summarized and classified according to the incoming material shape defects, and is divided into a middle wave shape data set, an edge wave shape data set, a high-order wave shape data set, and an asymmetric wave shape data set;

[0034] According to the unit capacity and production plan of the continuous annealing unit and its preceding cold rolling unit, the base material parameter fluctuation range of each deviation inducing factor is set respectively to construct a particle swarm algorithm model for deviation trend prediction to prevent some predicted values ​​from exceeding the actual parameter range;

[0035] A particle swarm prediction model is constructed, and the constructed particle swarm prediction model is trained based on a middle wave shape data set, an edge wave shape data set, a high-order wave shape data set, and an asymmetric wave shape data set; wherein the output of the particle swarm prediction model is the preheating section deviation, the heating section deviation, and the final deviation of the strip.

[0036] Furthermore, the method of using the trained deviation prediction model to obtain the deviation amount prediction result of the strip to be produced includes:

[0037] The trained deviation prediction model is used to predict the deviation of the preheating section, heating section and final deviation of the strip to be produced.

[0038] When the predicted deviation of the preheating section or the heating section exceeds the deviation warning value of the corresponding section, it is determined that the current strip cannot pass through the plate;

[0039] If the predicted deviation of the preheating section and the heating section do not exceed the deviation warning value of the corresponding section, and the predicted final deviation exceeds the maximum deviation warning value of the heating section, it is determined that the current strip cannot pass through the plate;

[0040] If the predicted deviation amount of the preheating section and the heating section do not exceed the deviation warning value of the corresponding section, and the predicted final deviation amount does not exceed the maximum deviation warning value of the heating section, it is determined that the current strip can pass through the plate.

[0041] Furthermore, when it is predicted that the strip steel to be produced will deviate, based on the deviating factor data and the deviation amount prediction result corresponding to the strip steel to be produced, a deviation correction decision that meets the requirements of the strip steel to be produced is selected from the historical deviation correction decision data, including:

[0042] Based on the deviation induction data set, the deviation correction principles adopted by the continuous annealing and stable plate passing of each coil of strip steel in historical production are extracted to construct a deviation correction strategy set;

[0043] Based on the deviation amount prediction result, the deviation inducing data of the unit, the deviation inducing data of the incoming material and the deviation inducing data of the environment are compared, the deviation correction strategy set is matched, and the deviation correction action of the strip to be produced is summarized;

[0044] According to the summarized deviation correction actions, continuous deviation correction is performed on the production strip.

[0045] Further, after performing continuous withdrawal correction on the steel strip to be produced, the method further comprises:

[0046] Determine whether the strip to be produced can pass through the plate stably after continuous withdrawal and correction; if the strip can pass through the plate stably, the deviation prediction model is determined to be accurate and no further correction is required; if the strip cannot pass through the plate stably, the unit deviation induction data, incoming material deviation induction data and environmental deviation induction data corresponding to the current strip at this time are recorded and expanded to the deviation induction data set. The expanded deviation induction data set is used for the next deviation prediction and correction analysis.

[0047] On the other hand, the present invention further provides an electronic device, comprising a processor and a memory; wherein the memory stores at least one instruction, and the instruction is loaded and executed by the processor to implement the above method.

[0048] In yet another aspect, the present invention further provides a computer-readable storage medium, wherein the storage medium stores at least one instruction, and the instruction is loaded and executed by a processor to implement the above method.

[0049] The beneficial effects brought about by the technical solution provided by the present invention include at least:

[0050] The data-driven prediction and decision-making method for the continuous deviation of plate and strip steel with variable specifications provided by the present invention can target the problem of continuous deviation of plate and strip steel, start from the factors inducing deviation based on historical production data, and predict the deviation trend of the strip steel to be produced from multiple dimensions and angles. And by using the deviation prediction and comparing the historical deviation correction strategy, the deviation correction decision of the strip steel to be produced is made, so as to solve the technical problems that the existing technical means are difficult to cover the multi-dimensional influencing factors and deal with the frequent changes in the specifications of incoming materials, thereby ensuring the stable plate passing of the continuous annealing process of plate and strip steel. Moreover, compared with the existing technology, the present invention takes more indicators into consideration in the deviation factor analysis stage, so the prediction accuracy is higher. Moreover, the present invention associates the deviation prediction result with the deviation correction strategy, thereby realizing the connection between the deviation prediction and the deviation correction decision, which is more conducive to the management of the deviation problem of continuous annealing of plate and strip steel. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0052] Figure 1 It is a schematic diagram of the execution flow of the data-driven plate and strip steel variable specification continuous retreat deviation prediction and decision-making method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0053] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0054] First embodiment

[0055] This embodiment provides a data-driven method for predicting and making decisions on the deviation of strip steel during continuous backing with variable specifications. The method can be implemented by an electronic device, such as Figure 1 As shown, the method includes a deviation data collection and classification sub-process, a deviation trend prediction sub-process and a deviation correction action decision sub-process; the details are described below.

[0056] S1, sub-process for collecting and classifying deviation data: collecting historical production data of the continuous annealing unit and its preceding cold rolling unit; wherein, the historical production data includes deviation amount data and deviation factor data affecting the deviation amount in the historical production process; the historical production data is matched according to the steel coil number and steel grade, that is, the data of the two units can be matched one by one to each coil of strip steel.

[0057] Specifically, in this embodiment, S1 includes the following steps:

[0058] S11, collection and preliminary analysis of unit deviation data: according to the steel coil number and steel grade, collect the historical production data and historical deviation data of the continuous annealing unit corresponding to the steel coil; wherein, the historical production data of the continuous annealing unit includes the posture data of each furnace roller, the deviation correction data of each deviation correction roller, the continuous annealing speed data and the continuous annealing tension percentage data of the continuous annealing unit in the historical production process; wherein, the furnace roller posture data includes the furnace roller axis cross angle and the furnace roller axis offset; the historical deviation data includes the preheating section deviation data, heating section deviation data and final deviation data of the strip steel of the continuous annealing unit in the historical production process; for the collected data, preliminarily analyze the time domain variation law of each variable to determine whether there is a sudden change value that exceeds the original value by 150% to 200%. If it appears, delete the value, and calculate the true value of the point according to the difference between the two points before and after the deletion point. The difference formula used is The historical production data of the continuous annealing unit after removing the mutation value is used as the unit bias-induced data;

[0059] S12, collection and preliminary analysis of incoming material deviation data: according to the steel coil number and steel grade, collect the historical production data of the cold rolling unit corresponding to the steel coil; wherein the historical production data of the cold rolling unit includes the bandwidth B, the strip thickness H and the full-length plate shape distribution f(n, m, l); according to the full-length plate shape distribution, according to the homogeneous polynomial y=a0+a1x+a2x without cubic terms 2 +a3x 4 , fit each strip section, extract the distribution function g(l) of its linear coefficient a1 along the length direction of the plate; use the bandwidth B, strip thickness H and the extracted linear coefficient distribution function g(l) along the length direction of the plate as the incoming material deflection data;

[0060] S13, collection of preset environmental bias data: according to the steel coil number and steel grade, collect the historical furnace temperature control parameters of the continuous annealing unit corresponding to the steel coil; wherein the furnace temperature control parameters include the preset temperature value T of each section i and the axial temperature difference ΔT of each section; the furnace temperature control parameter is used as the environmental bias data;

[0061] S14, the collected historical deviation data, unit deviation data, incoming material deviation data and environmental deviation data are matched one by one according to the steel coil number and steel grade, so as to form an integrated historical data set;

[0062] S15, comparing the data of strip steels with the same grade, width and thickness, analyzing the strip shape distribution function of each section, the deviation amount of the preheating section, the deviation amount of the heating section and the final deviation amount, if the difference between the strip steels is less than 5% of the basic value, only one data is retained in this series of data to avoid the influence of repeated data on the weight analysis of the deviation induction index, and obtain the deduplicated data set;

[0063] S16, based on the deduplicated data set, with the deviation of the preheating section, the deviation of the heating section and the final deviation as the target value, a random forest algorithm model is established to calculate the weight coefficient of each index. The indexes include: the posture of each furnace roller (axis crossing angle, axis offset), the deviation amount of each deviation correction roller, the deviation of the strip in the preheating section and the heating section, the continuous withdrawal speed and the continuous withdrawal tension percentage, the bandwidth B, the strip thickness H, the distribution function g(l) of the first-order coefficient a1 along the length direction of the plate, the preset value T of each section temperature i and the axial temperature difference ΔT of each section;

[0064] S17, screening each indicator according to the weight coefficient, and using the screened indicators as deviation inducing factors; wherein, if the weight coefficient of the indicator exceeds 5%, the indicator is determined to be a deviation inducing factor; if the weight coefficient of the indicator is less than 5%, the indicator is determined not to be a deviation inducing factor;

[0065] S18, sorting out the deviation inducing factors to form a deviation inducing indicator set, and filtering out the data belonging to the deviation inducing factors from the deduplicated data set to form a deviation inducing data set.

[0066] S2, deviation trend prediction sub-process: based on the historical production data, a preset deviation prediction model is trained; wherein the deviation prediction model is a particle swarm algorithm model, whose input is the deviation factor data and output is the deviation amount data; the deviation amount prediction result of the strip to be produced is obtained by using the trained model;

[0067] Specifically, in this embodiment, S2 includes the following steps:

[0068] S21, classifying the deviation induced data set according to the incoming material shape defects, and dividing it into a middle wave shape data set, an edge wave shape data set, a high-order wave shape data set, and an asymmetric wave shape data set;

[0069] S22, according to the unit capacity and production plan of the continuous annealing unit and its preceding cold rolling unit, set the posture of each furnace roller, the correction amount of each correction roller, the continuous annealing speed and the continuous annealing tension percentage, the bandwidth B, the strip thickness H, and the preset temperature value T of each section. i The base material parameter fluctuation range of the axial temperature difference ΔT of each section is used to build a particle swarm algorithm model for deviation trend prediction to prevent some predicted values ​​from exceeding the actual parameter range;

[0070] S23, constructing a particle swarm prediction model, and training the constructed particle swarm prediction model based on various types of plate shape defect data sets obtained in S22; wherein the output of the particle swarm prediction model is the deviation amount of the preheating section, the deviation amount of the heating section, and the final deviation amount of the strip;

[0071] S24, using the trained deviation prediction model, predicting the deviation amount of the preheating section, the heating section and the final deviation amount of the strip to be produced;

[0072] When the predicted deviation of the preheating section and the heating section exceeds the deviation warning value of the corresponding section, it can be determined that the coil cannot pass the plate, otherwise it can be passed; if the preconditions are met and the final deviation exceeds the maximum deviation warning value of the heating section, it can be determined that the coil cannot pass the plate, otherwise it can be released.

[0073] S3, deviation correction action decision sub-process; when it is predicted that the strip to be produced will deviate, based on the deviation inducement factor data and deviation amount prediction result corresponding to the strip to be produced, a deviation correction decision that meets the requirements of the strip to be produced is selected from the historical deviation correction decision data; specifically, in this embodiment, it includes the following steps:

[0074] S31, based on the various plate shape defect data sets obtained in S22, the deviation correction principles adopted by the continuous annealing and stable plate passing of each coil of strip steel in the historical production are correspondingly extracted to construct a deviation correction strategy set;

[0075] S32, based on the deviation amount prediction result, comparing the unit deviation inducing data, the incoming material deviation inducing data and the environmental deviation inducing data, matching the deviation correction strategy set, and summarizing the deviation correction action of the strip to be produced;

[0076] S33, performing continuous retraction correction on the production strip according to the summarized correction actions;

[0077] S34, determines whether the strip to be produced can pass through the plate stably after continuous withdrawal and deviation correction; if the strip can pass through the plate stably, it is determined that the deviation prediction model is accurate and no further correction is required; if the strip cannot pass through the plate stably, the unit deviation induction data, incoming material deviation induction data and environmental deviation induction data corresponding to the current strip at this time are recorded, and S11 is re-executed to expand the data in this volume to the data set for the next prediction and deviation correction analysis.

[0078] In summary, this embodiment provides a data-driven prediction and decision-making method for the continuous annealing deviation of variable-specification strip steel. This method can be based on the historical production data of the continuous annealing unit and its preceding cold rolling unit, and analyze the influence weights of the unit deviation-inducing data, the incoming material deviation-inducing data and the environmental deviation-inducing data through a random forest model. Combined with the actual historical deviation, a particle swarm prediction model is established to further predict the continuous annealing deviation of the incoming strip steel. At the same time, the deviation correction strategy set is matched to obtain the pre-correction strategy of the strip to be annealed for pre-adjustment. This method can be run in parallel with the control system of the continuous annealing unit to provide a guarantee for the stable production scheduling of the continuous annealing unit.

[0079] Second embodiment

[0080] This embodiment uses a practical application example to illustrate the implementation process of the method of the present invention, including:

[0081] Based on the pickling and annealing production line of a steel plant, the historical data of the production line for two quarters was collated and obtained, totaling 18,639 coils of strip steel data. Based on the above coils, the biased data collection and classification steps were performed, including:

[0082] S11, collect historical production data of the pickling-rolling-annealing production line; wherein, the historical production data is matched according to the steel coil number and steel grade, that is, the data of the two units can be matched one by one to each roll of strip steel. According to the above process from S111 to S113, data mutation value detection is performed, 2597 groups of duplicate data are screened out, and the final data set contains 16042 rolls. Among them, the situation of some collected data is shown in Table 1.

[0083] Table 1 Collected data

[0084]

[0085] (Continued)

[0086]

[0087]

[0088] (Continued)

[0089]

[0090] S12, based on the above data set, takes the deviation in the preheating section, the deviation in the heating section and the final deviation as the target values, establishes a random forest algorithm model, calculates the weight coefficient of each indicator, and the calculation results are shown in Table 2.

[0091] Table 2 Weight coefficients of each indicator

[0092]

[0093]

[0094] After analysis, the axis crossing angle, CPC6.0, deviation in preheating section, CPC7.0, linear distribution function along the length direction, preheating section temperature preset value T1, preheating section axial temperature difference value ΔT1, heating section axial temperature difference value ΔT2, cooling section temperature preset value T4, cooling section axial temperature difference value ΔT4 are the deviation inducing factors;

[0095] The deviation induction data set is further sorted out, as shown in Table 3.

[0096] Table 3. Further sorted data of the deviation induction data set

[0097]

[0098] (Sub-table)

[0099]

[0100] Then, the deviation trend prediction sub-process is executed, including:

[0101] S21, according to the incoming material shape defects, further classify the deviation induction data set obtained by the deviation induction data collection and classification sub-process into a middle wave shape data set, an edge wave shape data set, a high-order wave shape data set and an asymmetric wave shape data set.

[0102] S22, according to the unit capacity and production plan of the continuous annealing unit and its preceding cold rolling unit, the base material parameter fluctuation range of each deviation induction data is set respectively, as shown in Table 4.

[0103] Table 4 Fluctuation range of base material parameters for various deviation induction data

[0104]

[0105] S23, construct a particle swarm prediction model, and combine it with various plate shape defect data sets obtained in S21 to perform model training.

[0106] S24, using the trained particle swarm prediction model, predicting the deviation amount of the strip preheating section, the deviation amount of the strip heating section and the final deviation amount of the strip to be produced by continuous annealing.

[0107] Among them, the deviation warning value of the preheating section is 10mm, the deviation warning value of the heating section is 15mm, and the final deviation warning value is 20mm. The deviation of the strip to be produced can be predicted as shown in Table 5.

[0108] Table 5 Prediction results of deviation of the strip to be produced

[0109]

[0110]

[0111] According to the analysis results, the first two rolls of strip steel did not exceed the warning value and could be released; the deviation of the third roll in the preheating section exceeded the warning value and needed to be corrected.

[0112] Furthermore, for the deviated roll, the decision-making sub-process of corrective action is executed, which specifically includes:

[0113] S31, for various types of plate shape defect data sets obtained in S21, the correction principles adopted for the continuous annealing and stable passing of each coil of steel strip are extracted to construct a correction strategy set.

[0114] S32, combined with the incoming material deviation amount predicted by S24, compares the unit deviation induction data, incoming material deviation induction data, and preset environment deviation induction data, matches the correction strategy set of S31, and further summarizes the correction action of the strip to be annealed.

[0115] According to the correction action, the correction strategy can be matched: increase the CPC6.0 adjustment amount to 19.3mm, and increase the CPC7.0 adjustment amount to 9.08mm.

[0116] S33, execute continuous withdrawal and correction, and make sure that stable board passing is achieved, then the model is accurate and no further correction is required.

[0117] Third embodiment

[0118] This embodiment provides an electronic device, which includes a processor and a memory; wherein the memory stores at least one instruction, and the instruction is loaded and executed by the processor to implement the method of the first embodiment.

[0119] The electronic device may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) and one or more memories, wherein the memory stores at least one instruction, and the instruction is loaded by the processor to execute the above method.

[0120] Fourth embodiment

[0121] This embodiment provides a computer-readable storage medium, which stores at least one instruction, and the instruction is loaded and executed by a processor to implement the method of the first embodiment. The computer-readable storage medium may be a ROM, a random access memory, a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc. The instructions stored therein may be loaded by a processor in a terminal to execute the method.

[0122] In addition, it should be noted that the present invention can be provided as a method, an apparatus or a computer program product. Therefore, the embodiments of the present invention can take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the embodiments of the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code.

[0123] The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0124] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device, so that a series of operation steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable terminal device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0125] It should also be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. The terms "include", "comprises" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements.

[0126] Finally, it should be noted that the above is a preferred embodiment of the present invention. It should be pointed out that although the preferred embodiment of the present invention has been described, for those skilled in the art, once the basic creative concept of the present invention is known, several improvements and modifications can be made without departing from the principles of the present invention, and these improvements and modifications should also be regarded as the protection scope of the present invention. Therefore, the attached claims are intended to be interpreted as including the preferred embodiment and all changes and modifications that fall within the scope of the embodiments of the present invention.

Claims

1. A data-driven method for predicting and making decisions on the deviation of strip steel during continuous withdrawal with variable specifications. It is characterized in that The data-driven continuous backing deviation prediction and decision-making method for plate and strip steel with variable specifications includes: Collecting historical production data of the continuous annealing unit and its preceding cold rolling unit; wherein the historical production data includes the deviation amount data and the deviation inducing factor data that affect the deviation amount in the historical production process; Based on the historical production data, a preset deviation prediction model is trained; wherein the input of the deviation prediction model is the deviation factor data, and the output is the deviation amount data; The deviation prediction result of the strip to be produced is obtained by using the trained deviation prediction model; When it is predicted that the strip to be produced will deviate, based on the deviation-inducing factor data and deviation amount prediction results corresponding to the strip to be produced, a deviation correction decision that meets the requirements of the strip to be produced is selected from the historical deviation correction decision data; Collect historical production data of the continuous annealing unit and its preceding cold rolling unit, including: According to the steel coil number and steel grade, the historical production data and historical deviation data of the continuous annealing unit of the corresponding steel coil are collected; wherein the historical production data of the continuous annealing unit includes the posture data of each furnace roller, the deviation correction data of each deviation correction roller, the continuous annealing speed data and the continuous annealing tension percentage data of the continuous annealing unit in the historical production process; wherein the furnace roller posture data includes the furnace roller axis crossing angle and the furnace roller axis offset; the historical deviation data includes the preheating section deviation data, the heating section deviation data and the final deviation data of the strip steel of the continuous annealing unit in the historical production process; Perform mutation detection on the collected historical production data and historical deviation data of the continuous annealing unit, and remove the mutation value; use the historical production data of the continuous annealing unit after removing the mutation value as the unit deviation inducement data; According to the steel coil number and steel grade, collect the historical production data of the cold rolling unit corresponding to the steel coil; wherein the historical production data of the cold rolling unit includes the distribution of the strip width, strip thickness and full-length plate shape; According to the full-length plate shape distribution, each strip section is fitted according to a homogeneous polynomial without cubic terms, and the distribution function of its linear coefficient along the plate length direction is extracted; the belt width, strip thickness and the distribution function of the extracted linear coefficient along the plate length direction are used as the incoming material deflection induction data; According to the steel coil number and steel grade, the historical furnace temperature control parameters of the continuous annealing unit of the corresponding steel coil are collected; wherein the furnace temperature control parameters include the preset temperature value of each section and the axial temperature difference value of each section; the furnace temperature control parameters are used as the environmental decoy data; The collected historical deviation data, unit deviation data, incoming material deviation data and environmental deviation data are matched one by one according to the coil number and steel grade to form an integrated historical data set; Deduplication is performed on the data in the obtained historical data set to obtain a deduplication data set; Based on the deduplicated data set, the weight coefficient of each indicator is calculated; wherein, the weight coefficient is used to characterize the influence of the corresponding indicator on the deviation amount; the indicators include the posture of each furnace roller of the continuous annealing unit, the deviation amount of each deviation correction roller, the continuous annealing speed, the continuous annealing tension percentage, the preset temperature value of each section and the axial temperature difference value of each section, as well as the deviation amount of the preheating section and the heating section of the strip, and the distribution function of the bandwidth, strip thickness and first-order coefficient in the cold rolling unit along the plate length direction; Screen each indicator according to the weight coefficient, and use the screened indicators as decoy factors; The data that are bias-inducing factors are selected from the deduplicated data set to form a deviation-inducing data set.

2. The data-driven method for predicting and making decisions on the deviation of strip steel during continuous back-off during specification change according to claim 1, It is characterized in that Deduplication of the data in the obtained historical data set includes: Compare the strip shape distribution functions of each section, preheating section deviation data, heating section deviation data and final deviation data of strips of the same brand, width and thickness. If the difference between the strips is less than the preset difference threshold, the data corresponding to the strip is considered to be repeated, and only one of the strip data is retained.

3. The data-driven method for predicting and making decisions on the deviation of strip steel during continuous withdrawal according to claim 1, It is characterized in that The weight coefficient of each indicator is calculated based on the deduplicated data set, including: Based on the deduplicated data set, a random forest algorithm model was established with the preheating section deviation, heating section deviation and final deviation as target values, and the weight coefficient of each indicator was calculated.

4. The data-driven method for predicting and making decisions on the deviation of strip steel during continuous withdrawal of variable specifications according to claim 1, It is characterized in that The method of screening out data belonging to bias factors from the deduplicated data set includes: If the weight coefficient of an indicator exceeds the preset weight threshold, it is judged to be a bias factor; If the weight coefficient of the indicator does not exceed the preset weight threshold, it is judged not to be a bias factor.

5. The data-driven method for predicting and making decisions on the deviation of strip steel during continuous withdrawal with varying specifications as claimed in claim 1, It is characterized in that The deviation prediction model is a particle swarm algorithm model.

6. The data-driven method for predicting and making decisions on the deviation of strip steel during continuous back-off during specification change according to claim 5, It is characterized in that Based on the historical production data, a preset deviation prediction model is trained, including: The deviation induction data set is summarized and classified according to the incoming material shape defects, and is divided into a middle wave shape data set, an edge wave shape data set, a high-order wave shape data set, and an asymmetric wave shape data set; According to the unit capacity and production plan of the continuous annealing unit and its preceding cold rolling unit, the base material parameter fluctuation range of each deviation inducing factor is set respectively to construct a particle swarm algorithm model for deviation trend prediction to prevent some predicted values ​​from exceeding the actual parameter range; A particle swarm prediction model is constructed, and the constructed particle swarm prediction model is trained based on a middle wave shape data set, an edge wave shape data set, a high-order wave shape data set, and an asymmetric wave shape data set; wherein the output of the particle swarm prediction model is the preheating section deviation, the heating section deviation, and the final deviation of the strip.

7. The data-driven method for predicting and making decisions on the deviation of strip steel during continuous withdrawal of variable specifications according to claim 1, It is characterized in that The method of using the trained deviation prediction model to obtain the deviation amount prediction result of the strip to be produced includes: The trained deviation prediction model is used to predict the deviation of the preheating section, heating section and final deviation of the strip to be produced. When the predicted deviation of the preheating section or the heating section exceeds the deviation warning value of the corresponding section, it is determined that the current strip cannot pass through the plate; If the predicted deviation of the preheating section and the heating section do not exceed the deviation warning value of the corresponding section, and the predicted final deviation exceeds the maximum deviation warning value of the heating section, it is determined that the current strip cannot pass through the plate; If the predicted deviation amount of the preheating section and the heating section do not exceed the deviation warning value of the corresponding section, and the predicted final deviation amount does not exceed the maximum deviation warning value of the heating section, it is determined that the current strip can pass through the plate.

8. The data-driven method for predicting and making decisions on the deviation of strip steel during continuous back-off during specification change according to claim 1, It is characterized in that When it is predicted that the strip steel to be produced will deviate, based on the deviating factor data and the deviation amount prediction result corresponding to the strip steel to be produced, a deviation correction decision that meets the requirements of the strip steel to be produced is selected from the historical deviation correction decision data, including: Based on the deviation induction data set, the deviation correction principles adopted by the continuous annealing and stable plate passing of each coil of strip steel in historical production are extracted to construct a deviation correction strategy set; Based on the deviation amount prediction result, the deviation inducing data of the unit, the deviation inducing data of the incoming material and the deviation inducing data of the environment are compared, the deviation correction strategy set is matched, and the deviation correction action of the strip to be produced is summarized; According to the summarized deviation correction actions, continuous deviation correction is performed on the production strip.

9. The data-driven method for predicting and making decisions on the deviation of strip steel during continuous withdrawal of variable specifications according to claim 8, It is characterized in that After the strip to be produced is corrected by continuous withdrawal, the method further comprises: Determine whether the strip to be produced can pass through the plate stably after continuous withdrawal and correction; if the strip can pass through the plate stably, the deviation prediction model is determined to be accurate and no further correction is required; if the strip cannot pass through the plate stably, the unit deviation induction data, incoming material deviation induction data and environmental deviation induction data corresponding to the current strip at this time are recorded and expanded to the deviation induction data set. The expanded deviation induction data set is used for the next deviation prediction and correction analysis.

Images