Deformation resistance prediction system, deformation resistance prediction method and deformation resistance prediction device

Abstract

A deformation resistance prediction system, a deformation resistance prediction method, and a deformation resistance prediction device are capable of predicting the deformation resistance with high accuracy even for steel types with a small amount of rolling performance data. The deformation resistance prediction device comprises: a rolling data acquisition unit, which acquires the operation data and rolling conditions of the rolling device and accumulates them as rolling performance data, and acquires rolling condition candidates from a rolling condition determination device; a prediction model, which predicts the deformation resistance of the rolled material; a deformation resistance prediction unit, which uses the prediction model to calculate the predicted value of the deformation resistance of the rolled material, the prediction model including: a global model, which is pre-set as a prediction model for estimating the deformation resistance of each steel type of the rolled material based on the overall determination of the rolling performance data; and a category model, which is pre-set as a prediction model for estimating the deformation resistance of a specific steel type based on the rolling performance data of the deformation resistance of the rolled material having a similar or common alloy composition in the rolling performance data.

Description

Technical Field

[0001] The invention relates to a deformation resistance prediction system, a deformation resistance prediction method and a deformation resistance prediction device in a hot rolling production line. Background Art

[0002] In hot rolling, in order to turn a steel slab (steel sheet) having a thickness of several tens of centimeters into a rolled material having a thickness of 1 to 10 mm, the steel slab is heated to about 1200° C. and then subjected to multiple rolling processes. One rolling process is generally referred to as a “pass”.

[0003] In order to produce a rolled material of the desired size and shape from a steel billet, the rolling amount needs to be determined according to the rolling pass. Therefore, in the hot rolling line control system, the mean flow stress (Mean Flow Stress) generated by rolling, that is, the deformation resistance (Deformation Resistance), is calculated in advance, and the rolling amount of each rolling pass is determined using this calculated amount. If the error in the calculation result of the deformation resistance is large, the setting of the rolling amount becomes inappropriate, resulting in poor rolling.

[0004] As a part of the hot rolling line control system, the device or subsystem for predicting deformation resistance is a deformation resistance prediction device, and the method in which the deformation resistance prediction device calculates deformation resistance based on information input from the upper hot rolling line control system is a deformation resistance prediction method.

[0005] As a technique for predicting deformation resistance, there is a technique disclosed in Non-Patent Document 1. According to this technique, the deformation resistance K is calculated as shown in Formula (1) using the following mathematical model based on the carbon concentration [C] of the rolled material, the absolute temperature T, the strain (Strain) ε during rolling, and the strain rate (Strain Rate) εdot. p This mathematical model is derived by summarizing the results of experiments conducted with multiple combinations of [C], T, ε, and εdot.

[0006]

Mathematical formula 1

[0007] K p =9.8exp(0.126-1.75[C]+0.94[C] 2 )×Exp((2851+2968[C]-1120[C] 2 ) / T)×ε 0.21 ×εdot 0.12 …(1)

[0008] As another technique for predicting deformation resistance, the technique of non-patent document 2 is known. In this technique, as shown in formula (2), the deformation resistance K is calculated using the following mathematical model based on the rotation speed V of the rolling roll, the reduction rate r, the thickness h of the rolled plate, the temperature T of the rolled material, the radius R of the work roll, and the reduction amount Δ. p In this mathematical model, a0 to a8 are coefficients determined for each steel type using rolling performance data.

[0009]

Mathematical formula 2

[0010] K p =(a0+a1V+a2V 2 +a3r+a4r 2 +a5h+a6h 2 +a7T+a8T 2 ) / (RΔ) 1 / 2 …(2)

[0011] As another technique for predicting deformation resistance, the technique of Patent Document 1 is known. In this technique, for a plurality of specific steel types A1, A2, A3 for which deformation resistance is calculated in advance, membership functions F1(C), F2(C), and F3(C) for each component are set, each of which uses the content of each chemical component as a parameter. For the deformation resistance of steel types B1, B2, B3, and B4 other than the specific steel types, the function value corresponding to the content of the chemical component is calculated based on the membership function, and the fitness for each specific steel type is calculated. The calculation is performed based on the fitness and the deformation resistance of each specific steel type. According to Patent Document 1, by this technique, the deformation resistance of steel types other than the specific steel types can be calculated with high accuracy in a relatively simple process.

[0012] As another technique for predicting deformation resistance, the technique of Patent Document 2 is known. In this technique, the following steps are included: a database preparation step, in which factors affecting deformation resistance in hot rolling and the deformation resistance are respectively set as explanatory variables and target variables, and past performance data are accumulated as a database; a required point data input step, in which data of the explanatory variables corresponding to the deformation resistance to be predicted later are input as required point data; a nearby data selection step, in which the distance between the data accumulated in the database and the required point data is calculated, and data with a short calculated distance is selected as nearby data; and a local model preparation step, in which a local model for local fitting near the required point is prepared based on the selected nearby data, and deformation resistance is predicted based on the prepared local model and the required point data.

[0013] According to Patent Document 2, this technology enables prediction of deformation resistance with higher accuracy than before, and since there is no need to create a global approximate equation, it is possible to save trouble such as maintenance during application.

[0014] However, in the technology disclosed in non-patent document 1, it is very difficult, if not impossible, to produce a mathematical model that encompasses a variety of rolling materials and rolling conditions based on experimental results. Therefore, in rolling materials whose alloy composition is different from that of the experimental materials, the predicted error of deformation resistance becomes larger, and there is a possibility of causing poor rolling.

[0015] In the technology disclosed in non-patent document 2, the coefficients of the mathematical model are determined by steel type using rolling performance data. Therefore, it is difficult to determine the coefficients of the mathematical model with high accuracy for new steel types for which there is no performance data. As a result, the prediction error of the deformation resistance of the new steel type becomes larger, which may cause poor rolling.

[0016] In the technology disclosed in Patent Document 1, the deformation resistance is predicted using the fitness calculated based on the membership functions F1(C), F2(C), and F3(C) differentiated by alloy composition for each steel type A1, A2, and A3 for which the deformation resistance has been pre-calculated. Therefore, when the membership function is changed, or when the steel type for which the deformation resistance has been pre-calculated is changed, the predicted results of the deformation resistance for all rolled materials change.

[0017] Therefore, even if there is a steel type with low prediction accuracy of deformation resistance among a variety of steel types, it is difficult to improve the prediction accuracy by adjusting the mathematical model limited to this steel type. In such a steel type, the prediction error of deformation resistance becomes large, which may cause rolling defects.

[0018] In the technology disclosed in Patent Document 2, a local model is used that is obtained by fitting only data close to the required point data. Therefore, it is difficult to fully apply the trends represented by a large amount of data to the model. In the case of steel types with few nearby data, if the nearby data contains errors for some reason, the prediction error of the deformation resistance becomes larger due to the influence of the error, and there is a possibility of causing poor rolling.

[0019] Patent Document 1: Japanese Patent Application Laid-Open No. 10-109106

[0020] Patent Document 2: Japanese Patent Application Publication No. 2010-207900

[0021] Non-patent document 1: The Iron and Steel Institute of Japan, Theory and Practice of Plate Rolling, p.161 (1992)

[0022] Non-patent document 2: VB Ginzburg, Metallurgical Design of Flat Rolled Steels, p. 295 (2005) Summary of the invention

[0023] In view of the above situation, the problem to be solved by the present invention is to provide a deformation resistance prediction system, a deformation resistance prediction method and a deformation resistance prediction device, which can predict the deformation resistance with high accuracy even for steel types with relatively less rolling performance data, and make adjustment for each steel type easy.

[0024] In order to solve the above-mentioned problems, the present invention is a deformation resistance prediction system, comprising: a rolling condition determination device for determining the rolling conditions set for a rolling device, a storage device for collecting the operation data of the rolling performed by the rolling device and the rolling conditions, a deformation resistance prediction device having a processor and a memory, which predicts the deformation resistance of the rolled material based on the operation data and the rolling conditions, and the deformation resistance prediction device having: a rolling data acquisition unit, which acquires the operation data and the rolling conditions and accumulates them as rolling performance data, and acquires rolling condition candidates from the rolling condition determination device; a prediction model, which predicts the deformation resistance of the rolled material; and a deformation resistance prediction unit. A unit that uses the prediction model to calculate a predicted value of the deformation resistance of the rolled material, the prediction model comprising: a global model, which is determined based on the overall rolling performance data and is pre-set as a prediction model for estimating the deformation resistance of each steel grade of the rolled material; and a category model, which is determined based on the rolling performance data of the deformation resistance of a specific steel grade having a similar or common alloy composition of the rolled material in the rolling performance data and is pre-set as a prediction model for estimating the deformation resistance of the specific steel grade, the deformation resistance prediction unit calculates the predicted value of the deformation resistance of the rolled material based on the rolling condition candidates using the global model and the category model.

[0025] If the technology of the present invention is used, the deformation resistance is predicted by overlapping a global model of deformation resistance determined by using the rolling performance data of multiple steel types with a category model of deformation resistance determined by using the rolling performance data of one or a few steel types. As a result, even for steel types with relatively few rolling performance data, the deformation resistance can be predicted with an accuracy at least higher than that of the global model of deformation resistance.

[0026] In addition, since the category model is selected for adjustment, it is easy to adjust only the category model that needs adjustment. In addition, there is a process of evaluating the prediction accuracy of each category model using rolling performance data and a process of displaying the change history of the prediction accuracy of each category model, so it is easy for the user to select the category model that needs adjustment. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is a configuration diagram showing an example of a schematic configuration of a hot rolling line control system according to one embodiment of the present invention, showing the configuration of the hot rolling line control system of the present invention.

[0028] Figure 2 This is a flowchart showing an example of the process performed by the deformation resistance prediction device of the present invention.

[0029] Figure 3 This is a flowchart showing an example of a deformation resistance prediction process performed by the deformation resistance prediction device of the present invention.

[0030] Figure 4 This is a diagram showing an example of rolling conditions in the present invention.

[0031] Figure 5 This is a diagram showing an example of a global model stored in the deformation resistance prediction device of the present invention.

[0032] Figure 6 This is a flowchart showing an example of a prediction accuracy evaluation process performed by the deformation resistance prediction device of the present invention.

[0033] Figure 7 This is a diagram showing an example of a screen displayed by the accuracy display step performed by the deformation resistance prediction device of the present invention.

[0034] Figure 8 This is a diagram showing another example of a screen displayed by the accuracy display step performed by the deformation resistance prediction device of the present invention.

[0035] Fig. 9 This is a diagram showing an example of a screen on which the deformation resistance prediction device of the present invention determines a threshold value to judge whether model adjustment is necessary.

[0036] Fig.10 This is a diagram showing an example of a screen displaying data determined as not requiring model adjustment by the deformation resistance prediction device of the present invention.

[0037] Fig.11 This is a flowchart showing an example of a class model adjustment process performed by the deformation resistance prediction device of the present invention.

[0038] Fig.12 This is a diagram showing an example of a class model stored in the deformation resistance prediction device of the present invention.

[0039] Fig.13 This is a distribution diagram showing an example of comparison between the predicted value of deformation resistance and the actual value by the deformation resistance prediction method of the present invention.

[0040] Fig.14 This is a distribution diagram showing an example of comparison between the predicted value and the actual value of the deformation resistance in the conventional technology.

[0041] Fig.15 This is a structural diagram showing an example of a schematic structure of the deformation resistance prediction device of the present invention.

[0042] Explanation of symbols

[0043] 1 Hot rolling production line control system,

[0044] 2 Host system,

[0045] 3 Hot rolling production line,

[0046] 11 Rolling condition determination device,

[0047] 12 Deformation resistance prediction device,

[0048] 13 Rolling performance data storage device,

[0049] 14 Rolling condition setting device,

[0050] 15 Operate data collection devices,

[0051] 110 Rolling conditions,

[0052] 113 Rolling conditions candidate,

[0053] 121 Deformation resistance prediction department,

[0054] 123 Prediction Model,

[0055] 124 precision records,

[0056] 125 Input device,

[0057] 126 output device,

[0058] 127 Global Model,

[0059] 128 Category Models,

[0060] 130 Rolling performance data,

[0061] 150 Operational data,

[0062] 12101 Rolling data acquisition department,

[0063] 12102 Abnormal Data Exclusion Department,

[0064] 12103 Global Model Application Department,

[0065] 12104 Category Model Application Department,

[0066] 12105 Deformation resistance prediction value output unit,

[0067] 12106 Accuracy Evaluation Department,

[0068] 12107 Precision display unit,

[0069] 12108 Category Model Adjustment Department,

[0070] 12109 Category Model Preservation Department. DETAILED DESCRIPTION

[0071] Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. In addition, in each of the drawings, the same reference numerals are given to common components, and repeated descriptions are omitted.

[0072] Figure 1 The hot rolling line control system 1 (or deformation resistance prediction system) of the present invention is shown. The hot rolling line control system 1 determines rolling conditions 110 based on the upper system information input from the upper system 2, and sets the rolling conditions 110 to the hot rolling line 3 (rolling device). The hot rolling line control system 1 includes: a rolling condition determination device 11, a deformation resistance prediction device 12, a rolling performance data storage device 13, a rolling condition setting device 14, and an operation data collection device 15.

[0073] The rolling condition determination device 11 determines the rolling condition 110 based on the information of the host system 2. In order to determine the rolling condition 110, the rolling condition determination device 11 uses the deformation resistance prediction device 12 to calculate the predicted value of the deformation resistance under various rolling conditions 110, and determines the optimal rolling condition 110 within the range of satisfying the constraints caused by the equipment conditions of the hot rolling line 3, etc.

[0074] The rolling condition determination device 11 stores the determined rolling condition 110 in the rolling performance data storage device 13 and then outputs the determined rolling condition 110 to the rolling condition setting device 14. The rolling condition setting device 14 sets the rolling condition 110 determined by the rolling condition determination device 11 to the hot rolling line 3.

[0075] Various sensors (eg, temperature, position, thickness, etc.) are arranged in the rolling line 3 to measure operation data 150 during the operation of the rolling line 3. The operation data 150 is collected by the operation data collection device 15 and stored in the rolling performance data storage device 13.

[0076] The deformation resistance prediction device 12 uses the rolling performance data 130 generated by merging the rolling conditions 110 and the operation data 150 stored in the rolling performance data storage device 13, and the rolling condition candidates 113 from the rolling condition determination device 11 to evaluate the prediction accuracy of the prediction model 123 (mathematical model) used for predicting the deformation resistance and adjust the coefficients of the mathematical model.

[0077] In addition, the deformation resistance prediction device 12 performs deformation resistance prediction based on the receipt of the rolling condition candidate 113 from the rolling condition determination device 11. In addition, the rolling condition candidate 113 is composed of the same data as the rolling condition 110, including the steel type of the rolling material to be rolled thereafter and the set value (predetermined value) of the rolling.

[0078] The prediction model 123 is composed of a preset global model 127 and a class model 128. The class model 128 is appropriately adjusted in a class model adjustment process as described later.

[0079] The global model 127 is a prediction model predetermined using a large number (e.g., dozens of) of rolling performance data 130 of steel types, and predicts the overall deformation resistance of the steel types. In other words, the model that predicts the trend of the deformation resistance corresponding to the steel type based on the collected rolling performance data 130 is the global model 127.

[0080] On the other hand, the category model 128 is a prediction model predetermined using the rolling performance data of one or a predetermined number (e.g., five) of steel types (categories) having similar or common alloy compositions among the collected rolling performance data 130. In other words, the global model 127 may be a collection of a plurality or a plurality of category models 128.

[0081] As described later, the deformation resistance prediction device 12 uses the global model 127 to grasp the overall (global) deformation resistance trend of the rolled material, and uses the category model to grasp the deformation resistance trend within the category different from the overall trend, and calculates the deformation resistance prediction value K p , provided to the rolling condition determination device 11.

[0082] After describing a series of steps performed by the deformation resistance prediction device 12 , the configuration of the deformation resistance prediction device 12 of the present invention will be described.

[0083] Figure 2 An example of the processing performed by the deformation resistance prediction device 12 of the present invention is shown. The deformation resistance prediction device 12 performs the deformation resistance prediction step S1 (deformation resistance prediction unit) when the rolling conditions are set. This processing can be performed when the manager of the hot rolling line control system 1 sets new rolling conditions 110 based on the information of the host system 2, and the rolling condition decision device 11 outputs the rolling condition candidate 113 to the deformation resistance prediction device 12.

[0084] When the rolling performance data 130 is stored in the deformation resistance prediction device 12, a prediction accuracy evaluation step S2 (prediction accuracy evaluation unit) is performed. For the evaluated prediction accuracy, a judgment step S3 is performed on whether the prediction model 123 needs to be adjusted. When it is determined that the prediction model 123 needs to be adjusted, a category model adjustment step S4 is performed. The following discloses each step of this embodiment in detail.

[0085] Figure 3 The following is a detailed description of the deformation resistance prediction step S1 performed by the deformation resistance prediction device 12 of the present invention. When the deformation resistance prediction step S1 is started, the deformation resistance prediction device 12 performs a rolling condition acquisition step S101 to acquire the rolling condition candidate 113 from the rolling condition determination device 11 .

[0086] Next, the deformation resistance prediction device 12 performs a global model application step S102 to apply the global model 127 of deformation resistance stored in the deformation resistance prediction device 12 to the rolling condition candidate 113 to calculate the global prediction value K of deformation resistance. p，G . Global prediction value K p，G It is a predicted value indicating the trend of the overall deformation resistance of the steel type.

[0087] Next, the deformation resistance prediction device 12 performs a class model application step S103 to apply the deformation resistance class model 128 to the rolling condition candidate 113 to calculate the class correction value A of the deformation resistance. p，C , according to the category correction value A p，C and the global prediction value K of the deformation resistance p，G Calculate the predicted value of deformation resistance K p In addition, the category correction value A p，C This is a correction value for obtaining a predicted value of deformation resistance specific to the alloy composition of the steel.

[0088] In the present invention, the category of the rolled material is a unit for classifying the rolled material based on the similarity of the chemical composition (alloy component), and is typically consistent with the steel type, but multiple steel types may be set as one category, or conversely, one steel type may be divided into multiple categories. The setting of the category may also be defined by the administrator or user of the deformation resistance prediction device 12.

[0089] In the present invention, the global model 127 of deformation resistance and the category model 128 of deformation resistance are used simultaneously, thereby grasping the global trend represented by multiple categories of data through the global model 127 of deformation resistance, and grasping the trend of data within a category different from the global trend through the category model 128.

[0090] Next, the deformation resistance prediction device 12 performs the deformation resistance prediction value output step S104 , outputs the deformation resistance prediction value to the rolling condition determination device 11 , and then ends the deformation resistance prediction step S1 .

[0091] Hereinafter, a simple embodiment of the global model 127 and the class model 128 of the deformation resistance in the present invention will be described.

[0092] One embodiment of the method for predicting deformation resistance using the overlap of the global model 127 and the class model 128 is to use the global prediction value K of the deformation resistance p，G and category correction value A p，C The predicted value of deformation resistance K is expressed by the product of p Example.

[0093]

Mathematical formula 3

[0094] K p =K p,G ×A p,C …(3)

[0095] Another embodiment of the method of predicting deformation resistance using the overlap of the global model 127 and the class model 128 is to use the global prediction value K of the deformation resistance. p，G and category correction value A p，C The sum of the deformation resistance K is used to express the predicted value of deformation resistance p Example.

[0096]

Mathematical formula 4

[0097] K p =K p,G +A p,C …(4)

[0098] The global model 127 of deformation resistance is in the form of multiplication of two factors, one of which is composed of the addition of a constant term, a logarithmic term of strain velocity, and a power term of temperature, and the other is composed of a power term of strain. The following formula (5) represents an example of such a global model 127.

[0099]

Mathematical formula 5

[0100] K p，G =[G0+G1ln(εdot)+G2 / T]×ε G3

[0101] K p，G =[G0+G1ln(εdot)+G2 / T+G3×T]×ε G4

[0102] K p，G=[G0+G1ln(εdot)+G2 / T+G3 / T 2 ]×ε G4 …(5)

[0103] In each example of the global model 127 described above, G0, G1, G2, G3, G4, etc. are functions of the alloy composition of the rolled material, and as a simple embodiment, they can be represented by a mathematical model shown in the following formula (6).

[0104]

Mathematical formula 6

[0105] G0=G 0,const +G 0,C W C +G 0,Si W Si +G 0,Mn W Mn +…

[0106] G1=G 1,const +G 1,C W C +G 1,Si W Si +G 1,Mn W Mn +…

[0107] G2=G 2,const +G 2,C W C +G 2,Si W Si +G 2,Mn W Mn +…

[0108] G3=G 4,const +G 3,C W C +G 3,Si W Si +G 3,Mn W Mn +…

[0109] G4=G 4,const +G 4,C W C +G 4,Si W Si +G 4,Mn W Mn +……(6)

[0110] Here, W C , W Si , W Mn These are mass ratios of carbon (C), silicon (Si), and manganese (Mn), respectively, and are examples of a part of alloy component terms included in a mathematical model representing a global coefficient.

[0111] In addition to these terms, the mathematical model of the global coefficient can also include nickel (Ni), chromium (Cr), molybdenum (Mo), niobium (Nb), titanium (Ti), vanadium (V), nitrogen (N), aluminum (Al), copper (Cu), tungsten (W), boron (B) and other terms.

[0112] The coefficients G0, G1, G2, G3, G4, etc. of the global model 127 of the deformation resistance in the present invention are functions of the alloy composition including multiple alloy composition terms as shown in the above formula, and the contribution parameter G of each alloy composition term is 0，const , G 0，C The parameters are determined using the rolling performance data 130 of a plurality of steel types that can be collected.

[0113] The present invention uses the rolling performance data 130 of a plurality of steel types to determine the parameters of the global model 127 of deformation resistance, thereby being able to fully apply the trend represented by a large amount of data to the prediction model 123. Therefore, even in the case of a steel type with little nearby data, the influence of errors that may be included in the nearby data can be reduced, thereby reducing the prediction error of the deformation resistance.

[0114] An example of a method for determining parameters of a mathematical model of a global coefficient will be described below using formula (7) as an example.

[0115]

Mathematical formula 7

[0116] K p,G =[G0+G1ln(εdot)+G2 / T]×ε G3,const …(7)

[0117] Let the strain of the mth data among the M actual data be ε m , set the strain rate to εdot m , set the temperature to T m , the actual value of deformation resistance is set as K p，act，m In addition, the alloy composition from carbon (C) to the Nth element (Elm_N) is set to {W C，m , W Si，m , W Mn，m , …, W Elm_N，m For simplicity, G0, G1, and G2 are all functions of N alloy components, so the following formula (8) is obtained based on the global model 127 of deformation resistance.

[0118]

Mathematical formula 8

[0119] K p,act,m / ε m G3,const =G 0,const +G 0,C W C,m +G0,Si W Si,m +…+G 0,Elm_N W Elm_N,m +(G 1,const +G 1,C W C,m +G 1,Si W Si,m +…+G 1,Elm_N W Elm_N,m )ln(εdot m )+(G 2,const +G 2，C W C，m +G 2,Si W Si,m +…+G 2,Elm_N W Elm_N，m ) / T m

[0120] …(8)

[0121] The M performance data can be expressed by the same formula. When these M formulas are expressed in the form of a matrix,

[0122]

Mathematical formula 9

[0123] Y=L×X…(9)

[0124] Here, Y on the left side of the above formula (9) is a vector of M rows, and its mth row is

[0125]

Mathematical formula 10

[0126] Y m =K p，act，m / ε m G3，const …(10)

[0127] In the above formula (9), L on the right is a matrix of M rows and 3N+3 columns, and its m-th row is given by the following formula.

[0128]

Mathematical formula 11

[0129] L m，{1，…，N+1} =1,W C,m , W Si，m , …, W Elm_N，m

[0130] L m,{N+2,…,2N+2} =ln(εdot m ),ln(εdot m )W C,m ,ln(εdot m )W Si,m ,…,ln(εdot m )W Elm_N,m

[0131] L m,{2N+3,…,3N+3} =1 / T m ,W C,m / T m ,W Si,m / T m ,…,W Elm_N,m / T m

[0132] …(11)

[0133] In the above formula (9), X on the right side is a vector of 3N+3 rows consisting of parameters of the mathematical model of the global coefficient, and its elements are calculated by the following formula (12).

[0134]

Mathematical formula 12

[0135] X {1,…,N+1} =G 0,const ,G 0,C , G 0,Si ,…,G 0,Elm_N

[0136] X {N+2,…,2N+2} =G 1,const ,G 1,C ,G 1,Si ,…,G 1,Elm_N

[0137] X {2N+3,…,3N+3} =G 2,const ,G 2,C ,G 2,Si ,…,G 2,Elm_N …(12)

[0138] Therefore, by changing G 3，const At the same time, for each G 3，const Calculate the square error |YL×X| 2 The vector X is the smallest, from which we further select G that minimizes the square error 3，const and X, thus obtaining the mathematical formula of the global coefficients and the parameters of the model. Calculate the square error |YL×X| 2 The method of finding the minimum vector X is well known in the field of scientific and technical computing, and can be easily calculated using, for example, an open source numerical analysis library such as LAPCK (Linear Algebra PACKage).

[0139] In the above, an example is used in which G0, G1, and G2 all include terms of N types of alloy components, and G3 only includes a constant term. However, even if the terms of alloy components included in G0, G1, G2, G3, and G4 are different from each other, the parameters can be determined using actual performance data in the same way as above.

[0140] In addition, the mathematical model of the global coefficient is explained above as an example of a linear function of the alloy composition, but the mathematical model of the global coefficient can also be a nonlinear function. In the case of a nonlinear function, a nonlinear optimization method of a continuous function can be used to determine the parameters. As such a nonlinear optimization method, a variety of methods such as Newton's method, Simplex method, Simulated annealing method, Levenberg-Marquardt method, etc. are well known.

[0141] As shown in the above example, the parameters of the global model 127 of the deformation resistance of the present invention are determined using the rolling performance data 130 of a plurality of steel types.

[0142] The class model 128 of deformation resistance is obtained by multiplying two factors, one of which is composed of an exponential function having an exponential part including the sum of an exponential term, a logarithmic term of strain velocity, and a power term of temperature, and the other factor is composed in a form including the power of strain. The following formula (13) represents an example of such a class model 128.

[0143]

Mathematical formula 13

[0144] A p,C =exp(C0+C1ln(εdot)+C2 / T)×ε C3

[0145] A p,C =exp(C0+C1ln(εdot)+C2 / T+C3×T)×ε C4

[0146] A p,C =exp(C0+C1ln(εdot)+C2 / T+C3 / T 2 )×ε C4 …(13)

[0147] In the above-mentioned category model 128, C0, C1, C2, C3, and C4 are coefficients determined based on data of rolled materials belonging to the same category. As described above, in the present invention, the category is a unit for classifying rolled materials based on the similarity of chemical composition, and is typically consistent with the steel grade.

[0148] However, the present invention is not limited thereto, and multiple steel types may be set as one category, or conversely, one steel type may be divided into multiple categories. Therefore, the coefficients C0, C1, C2, C3, and C4 of the category model 128 of the deformation resistance in the present invention are determined using actual performance data of rolled materials having similar chemical compositions.

[0149] The coefficients C0, C1, C2, C3, and C4 of the category model 128 are functions of the alloy composition of the rolled material. As a simple implementation, they can be expressed using a mathematical model shown in the following formula (14) in the same form as the coefficients of the global model 127.

[0150]

Mathematical formula 14

[0151] C0=C 0,const +C 0,C W C +C 0,Si W Si +C 0,Mn W Mn +…

[0152] C1=C 1,const +C 1，C W C +C 1,Si W Si +C 1，Mn W Mn +…

[0153] C2=C 2,const +C 2,C W C +C 2,Si W Si +C 2,Mn W Mn +…

[0154] C3=C 3,const +C 3,C W C +C 3,Si W Si +C 3,Mn W Mn +…

[0155] C4=C 4,const +C 4,C W C +C 4,Si W Si +C 4,Mn W Mn +……(14)

[0156] In the mathematical formula model of the coefficient of the category model 128, only the terms of C, Si, and Mn are shown, but the present invention is not limited thereto. In addition to these terms, the mathematical formula model of the coefficient of the category model 128 may also include terms such as nickel (Ni), chromium (Cr), molybdenum (Mo), niobium (Nb), titanium (Ti), vanadium (V), nitrogen (N), aluminum (Al), copper (Cu), tungsten (W), and boron (B).

[0157] The coefficients C0, C1, C2, C3, and C4 of the category model 128 are functions of the alloy composition including multiple alloy composition terms as shown in the above formula, and the contribution parameter C of each alloy composition term is 0，const , C 0，C The parameters are determined using the rolling performance data 130 of the rolled material belonging to the category. The method of determining the parameters of the mathematical model of the coefficients of the category model 128 will be described later together with the explanation of the adjustment method.

[0158] Figure 4 An example of rolling conditions 110 is shown. For each rolling material number, there is a header 111 including alloy components common to one rolling material, and a condition section 112 including temperature, strain, and strain rate that differ for each rolling pass or each position of the rolling material.

[0159] The header 111 includes a material number 1101 storing an identifier of the rolled material, a category 1102 storing an identifier of the category set for the rolled material, a C 1103 storing a carbon content (or content) of an alloy component, a Si 1104 storing a value of silicon of an alloy component, and a Mn 1150 storing a value of manganese. In addition, the alloy components are not limited to the above, and fields can be set according to the alloy components contained in the rolled material.

[0160] The condition part 112 includes a pass 1106, a position 1107, T 1108 indicating temperature, ε 1109 indicating strain, and ε dot 1110 indicating strain velocity in one entry. The position 1107 stores the position at which the state of the rolled material is measured in the hot rolling line 3. The position to be measured can be expressed as a ratio of the position from the front end in the entire length of the rolled material.

[0161] exist Figure 4 In the example, the temperature (T1108), strain (ε1109), and strain rate (εdot1110) at the 5% (front end), 50% (center), and 95% (tail end) positions of the length of the rolled material in each pass 1106 are input as rolling conditions.

[0162] exist Figure 42 shows an example in which the type 1102 of the rolled material is determined by the rolling condition determination device 11 and input to the deformation resistance prediction device 12 as a rolling condition, but the deformation resistance prediction device 12 may determine the type based on the alloy composition of the rolled material.

[0163] In addition, Figure 4 An example is shown in which the strain (ε1109) and strain velocity (εdot1110) are calculated by the rolling condition determination device 11 and input into the deformation resistance prediction device 12 as rolling conditions, but the rolling conditions may also include the reduction amount of each pass, the rolling speed or the roller rotation speed, etc. instead of the strain and the strain velocity. The deformation resistance prediction device 12 calculates the strain and the strain velocity based on the reduction amount (load), the rolling speed, etc.

[0164] In addition, although not shown in the figure, the operation data 150 can include sensor data measured by sensors installed in the hot rolling production line 3, for example, it can be composed of load, temperature, speed of rolled material, thickness of rolled material, gap between rollers (rolled material), etc.

[0165] The rolling performance data 130 is Figure 4 The data obtained by merging the rolling condition 110 and the operation data 150 is not shown in the figure, but for example, according to the material number 1101 of the rolling condition 110, it is composed of the condition part 112 of the rolling condition 110 and the above-mentioned sensor data of the operation data 150.

[0166] In addition, the rolling performance data 130 can be accumulated according to the rolled material that has been rolled, but can also be set as statistical data obtained by statistically processing the maximum value, average value (central value), minimum value, etc. of the sensor data in a predetermined number (or time) unit. In addition, the rolling performance data 130 can be generated by the deformation resistance prediction device 12 based on the rolling conditions 110 and the operation data 150, and can also be generated by the rolling performance data storage device 13 based on the rolling conditions 110 and the operation data 150.

[0167] Figure 5 An example of a global model 127 stored in the deformation resistance prediction device 12 is shown. The global model 127 includes: a coefficient 1271, a Const 1272 storing constants, and fields (1273 to 1275) storing values ​​corresponding to alloy components. In the example shown in the figure, examples of C, Si, and Mn are shown as alloy components, but are not limited to this. In addition, as shown in the above formula (6), the coefficient 1271 can store an identifier of a mathematical model. In addition, the category model 128 stored in the deformation resistance prediction device 12 is shown later together with a description of the adjustment method.

[0168] Figure 6This represents the prediction accuracy evaluation step S2 of the present invention. Figure 2 In the prediction accuracy evaluation step S2 shown, the deformation resistance prediction device 12 performs a rolling performance acquisition step S201 to acquire the rolling performance data 130 to be evaluated for accuracy from the rolling performance data storage device 13 .

[0169] In addition, which rolling performance data 130 in the rolling performance data 130 is set as the object of the accuracy evaluation is determined according to the execution frequency of the prediction accuracy evaluation process S2 and the evaluation frequency setting performed by the user. For example, when the execution frequency is set to each rolled material, regardless of the setting of the evaluation frequency, each time the rolling is completed for one rolled material and the operation data 150 is saved, the deformation resistance prediction device 12 performs the prediction accuracy evaluation process S2 and obtains the rolling performance data 130 of the rolled material just rolled as the object of the accuracy evaluation.

[0170] On the other hand, when the execution frequency is set to every 10 rolled materials and the evaluation frequency is set to every 1 rolled material, the deformation resistance prediction device 12 performs the prediction accuracy evaluation process S2 every time the rolling of 10 rolled materials is completed and the operation data is saved, and obtains the rolling performance data 130 of the 10 rolled materials just rolled as the accuracy evaluation object.

[0171] Alternatively, when the execution frequency is set to every 10 rolled materials and the evaluation frequency is set to every 2 rolled materials, the deformation resistance prediction device 12 performs the prediction accuracy evaluation step S2 every time the rolling of 10 rolled materials is completed and the operation data 150 is saved, and data for 2 rolled materials out of the rolling performance data 130 of the 10 rolled materials just rolled are obtained as accuracy evaluation objects. In addition, various combinations of settings other than those exemplified can of course be performed.

[0172] Next, the deformation resistance prediction device 12 performs an abnormal data elimination step S202 to determine whether there is an abnormality in the temperature, strain, strain rate, etc. of the operation data 150, and excludes the data from the accuracy evaluation object if there is an abnormality. In order to determine whether there is an abnormality, the deformation resistance prediction device 12 uses the difference between the operation data 150 and the rolling condition 110.

[0173] For example, when the difference between the temperature of the operation data 150 and the temperature of the rolling condition 110 exceeds a predetermined threshold and deviates from the normal range, the operation data 150 is judged to be abnormal. In this case, the normal range may be set by the user, or may be automatically calculated by applying statistical analysis to the rolling performance data 130, or may be determined by combining these two methods.

[0174] Next, the deformation resistance prediction device 12 performs a global model application step S203, applying the global model 127 of deformation resistance stored in the deformation resistance prediction device 12 to the alloy composition of the rolling conditions and the temperature, strain, and strain rate of the operation data 150, thereby calculating the global prediction value K of the deformation resistance. p， G.

[0175] Next, the deformation resistance prediction device 12 performs a class model application step S204, applies the deformation resistance class model 128 stored in the deformation resistance prediction device 12 to the alloy composition of the rolling conditions and the temperature, strain, and strain rate of the operation data 150, and calculates the class correction value A of the deformation resistance. p，C , according to the correction value A p，C and the global prediction value K p，G Calculate the predicted value of deformation resistance K p In addition, the predicted value K p As described above, the correction value A can be p，C With the global prediction value K p，G The sum or product is used for calculation.

[0176] Next, the deformation resistance prediction device 12 performs an accuracy evaluation step S205 to evaluate the accuracy of the deformation resistance prediction model 123. Therefore, the deformation resistance prediction device 12 calculates the predicted value K of the deformation resistance using the prediction model 123. p The actual value K of the deformation resistance included in the operation data 150 (or the rolling performance data 130) p，act The comparison is performed to calculate the accuracy of the prediction model 123 of the deformation resistance.

[0177] In addition, the actual value K of the deformation resistance included in the operation data 150 is p，act The value calculated by the rolling performance data storage device 13 based on the operation data 150 can be used.

[0178] The index of this accuracy is typically the predicted value K calculated at multiple positions and multiple passes in one rolled material. p -K p，act The average of the squares of 1 / 2, that is, RMSE (Root Mean Squared Error), but you can also use (K p -K p，act ) / K p，act Alternatively, multiple accuracy indicators may be used. The deformation resistance prediction device 12 calculates the predicted value K evaluated in the accuracy evaluation step S205. p , K used for the above comparison p，act, one or more accuracy indicators are stored in the accuracy history 124 in the deformation resistance prediction device 12 together with the number of the rolled material (material number 1101), the category (1102), and the rolling date and time.

[0179] Next, the deformation resistance prediction device 12 performs an accuracy display step S206 to present the accuracy of the deformation resistance prediction model 123 to the user of the deformation resistance prediction device 12. In the accuracy display step S206, the accuracy index of the selected rolling material is read from the deformation resistance prediction device 12 and presented to the user via an output device (described later).

[0180] The accuracy display step S206 changes the displayed accuracy index, the rolled material for displaying the accuracy index, and the display method of the accuracy index according to the user's settings, and displays the prediction accuracy of the deformation resistance prediction model 123 by category and then ends.

[0181] Figure 7 An example of a screen 700 displayed on the output device in the accuracy display step S206 is shown. Figure 7 Screen 700 shows the RMSE of the prediction value of the prediction model 123 of the deformation resistance of 100 rolled materials in each category that have been recently evaluated for 6 categories from category 1001 to category 1023. The horizontal axis of screen 700 represents the history of the 100 rolled materials from the previous (-100) to the most recent (0), and the vertical axis represents the RMSE of the prediction model 123 of the deformation resistance evaluated in each rolled material.

[0182] Figure 7 This is an example of accuracy display, and other various display methods are also possible. For example, by setting the horizontal axis to the rolling date and time, the time change of the accuracy index can also be displayed. In addition, by setting the vertical axis to the RMSE of the predicted value of a specific pass, or to the RMSE of a specific position, instead of the RMSE of the predicted values ​​at multiple passes and multiple positions in one rolled material, the accuracy index can also be displayed more analytically. In addition, instead of displaying the accuracy index of each rolled material, the moving average for multiple materials is displayed, thereby making it possible to set a smoother and easier to understand display of the trend of change.

[0183] Figure 8 Indicates that 10 rolled materials are Figure 7 The RMSE is shown as an example of a moving average. Figure 7 If we observe carefully in , we can see that the RMSE of category 1002 increases, but Figure 8 It is easier to find the increase in RMSE for class 1002.

[0184] Figure 2The step S3 of determining whether the model needs to be adjusted is performed when the accuracy index is greater than a predetermined threshold value. The threshold value may be set by the user or determined by the deformation resistance prediction device 12 through statistical analysis of the accuracy index data of each category.

[0185] Fig. 9 This figure shows an example of a screen 900 in which the deformation resistance prediction device 12 determines a threshold value through statistical analysis and judges whether model adjustment is required.

[0186] exist Fig. 9 In the example of , the deformation resistance prediction device 12 defines the threshold value as Avg+2×std_dev using the moving average Avg and the moving standard deviation std_dev of the RMSE in the rolled material evaluated most recently.

[0187] The RMSE of each rolled material shown by the solid line in the figure rises from the horizontal axis = "-20", that is, the rolled material 20 pieces ago exceeds the threshold line (Avg + 2 × std_dev). In this case, the deformation resistance prediction device 12 determines that the model needs to be adjusted and sends a signal to the user to urge the adjustment. The signal is, for example, a screen display, sound, vibration, or email.

[0188] When the deformation resistance prediction device 12 determines that the model needs to be adjusted, the deformation resistance prediction device 12 may issue a signal urging the user to adjust the model instead of issuing the signal, or may execute the category model adjustment step S4 at the same time as issuing the signal.

[0189] Fig.10 An example of a screen 950 is shown in which it is determined that the model adjustment is not necessary. In the screen 950, the RMSE of each rolled material of the category 1001 stops in a region lower than the threshold value defined as Avg+2×std_dev, so the deformation resistance prediction device 12 determines that the model adjustment is not necessary. Fig. 9 and Fig.10 2 shows an example in which Avg+2×std_dev is used as a threshold for determining whether the model needs to be adjusted, but other thresholds based on statistical analysis may also be used.

[0190] Fig.11 The class model adjustment step S4 of the present invention is shown. When the deformation resistance prediction device 12 starts the class model adjustment step S4, it executes the adjustment class selection step S401 to select the class to be adjusted. The class to be adjusted can be selected by the deformation resistance prediction device 12 based on the accuracy index and the threshold for determining whether the model needs to be adjusted, or by the user. The deformation resistance prediction device 12 can also prompt the recommended class to the output device and the user makes the final selection.

[0191] The deformation resistance prediction device 12 performs the steps of the class data extraction step S402 to the class model storage step S406 for each class of the adjustment target selected in the class data extraction step S402 , and adjusts the class model 128 for each class of the deformation resistance.

[0192] The adjustment completion determination S407 related to the selected category is a process for adjusting the categories of all the selected adjustment objects. Therefore, the adjustment completion determination S407 determines whether the processing for the categories of all the selected adjustment objects is completed. If there are still unadjusted categories, the process is returned to the category data extraction step S402 to continue the adjustment for the unadjusted categories.

[0193] Hereinafter, the above-mentioned steps ( S402 to S406 ) executed by the deformation resistance prediction device 12 for each type of the adjustment target will be described.

[0194] First, the deformation resistance prediction device 12 performs the category data extraction step S402 to obtain the rolling performance data 130 consisting of the rolling conditions 110 and the operation data 150 belonging to the category of the adjustment object from the rolling performance data storage device 13. Next, the deformation resistance prediction device 12 performs the abnormal data elimination step S403 to eliminate abnormal data from the obtained rolling performance data 130. The processing in this step may be the same as the abnormal data elimination step S202 performed in the prediction accuracy evaluation step S2.

[0195] Next, the deformation resistance prediction device 12 executes the global model application step S404, applies the global model 127 of deformation resistance stored in the deformation resistance prediction device 12 to the alloy composition (1103-1105) of the rolling condition 110 and the temperature, strain, and strain rate of the operation data 150, and calculates the global prediction value K of the deformation resistance as described above. p，G .

[0196] Next, the deformation resistance prediction device 12 executes the class model adjustment step S405 to adjust the parameters included in the mathematical model of the coefficient of the class model 128.

[0197]

Mathematical formula 15

[0198] A p，C =exp(C0+C1ln(εdot)+C2 / T)×ε C3 …(15)

[0199] The processing in the category model adjustment step S405 will be described.

[0200] The number of data of the rolling performance data 130 of the category provided to the global model application step S404 through the abnormal data elimination step S403 is M. If the performance value of the deformation resistance in the m-th data among the M data is K p，act，m , set the global prediction value of deformation resistance to K p ,G, m , then the expected category correction value A of the deformation resistance in the data p，C，m The following formula (16) is satisfied.

[0201]

Mathematical formula 16

[0202] K p,act,m =K p,G,m ×A p,C,m …(16)

[0203] When the natural logarithms of both sides of the above formula (16) are calculated and the mathematical formula is arranged using the example of the category model 128 of the deformation resistance, the following formula (17) is obtained.

[0204]

Mathematical formula 17

[0205] ln(K p,act,m / K p,G,m )=C0+C1ln(εdot m )+C2 / T m +C3ε m …(17)

[0206] Here, the coefficients C0, C1, C2, and C3 of the category model 128 are functions of the alloy composition as described above. A simple example is shown in the following formula (18), which is a linear function of the alloy composition as described above.

[0207]

Mathematical formula 18

[0208] C0=C 0,const +C 0,C W C +C 0,Si W Si +C 0,Mn W Mn +…

[0209] C1=C 1,const +C 1,C W C +C 1,Si W Si +C 1,Mn W Mn +…

[0210] C2=C 2,const +C 2,C W C+C 2,Si W Si +C 2,Mn W Mn +…

[0211] C3=C 3,const +C 3,C W C +C 3,Si W Si +C 3,Mn W Mn +……(18)

[0212] When the above-mentioned linear function is used for sorting, the relationship between the mth performance data and the parameters included in the mathematical model of the coefficient of the category model 128 is expressed by the following formula (19).

[0213]

Mathematical formula 19

[0214] ln(K p,act,m / K p,G,m )=C 0,const +C 0，C W C,m +C 0,Si W Si,m +…+C 0,Elm_N W Elm_N,m +(C 1,const +C 1,C W C,m +C 1,Si W Si,m +…+C 1,Elm_N W Elm_N,m )ln(εdot m )+(C 2,const +C 2,C W C,m +C 2,Si W Si,m +…+C 2,Elm_N W Elm_N,m ) / T m +(C 3,const +C 3,C W C,m +C 3,Si W Si,m +…+C 3,Elm_N W Elm_N,m )ε m …(19)

[0215] The M rolling performance data 130 provided to the classification model adjustment step S405 can all be expressed by the same formula. When these M formulas are expressed in the form of a matrix,

[0216]

Mathematical formula 20

[0217] Z=U×S…(20)

[0218] Here, Z on the left side of the above formula (20) is a vector of M rows, and its mth row is

[0219]

Mathematical formula 21

[0220] ln(K p,act,m / K p,G,m )…(twenty one)

[0221] In addition, U on the right side of the above formula (20) is a matrix of M rows and 4N+4 columns, and its mth row is given by the following formula (22).

[0222]

Mathematical formula 22

[0223] U m,{1,…,N+1} =1,W C,m ,W Si,m ,…,W Elm_N,m

[0224] U m,{N+2,…,2N+2} =ln(εdot m ),ln(εdot m )W C,m ,ln(εdot m )W Si,m ,…,ln(εdot m )W Elm_N,m

[0225] U m,{2N+3,…,3N+3} =1 / T m ,W C,m / T m ,W Si,m / T m ,…,W Elm_N,m / T m

[0226] U m,{3N+4,…,4N+4} =ε m ,W C,m ε m ,W Si,m ε m ,…,W Elm_N,m ε m …(twenty two)

[0227] In addition, S on the right side of the above formula (20) is a vector of 4N+4 rows consisting of parameters of the mathematical model of the category coefficient, and its elements are calculated by the following formula (23).

[0228]

Mathematical formula 23

[0229] S {1,…,N+1} =C 0,const ,C0,C ,C 0,Si ,…,C 0,Elm_N

[0230] S {N+2,…,2N+2} =C 1,const ,C 1,C ,C 1,Si ,…,C 1,Elm_N

[0231] S {2N+3,…,3N+3} =C 2,const ,C 2,C ,C 2,Si ,…,C 2,Elm_N …(twenty three)

[0232] Therefore, by calculating the square error |ZU×S| 2 The parameters of the mathematical model of the category coefficients are obtained by taking the smallest vector S. Calculate the square error |ZU×S| 2 The method of finding the smallest vector S is well known in the field of scientific and technical computing, and can be easily calculated using, for example, an open source numerical analysis library such as LAPCK (Linear Algebra PACKage).

[0233] In the above, an example in which C0, C1, C2, and C3 all include items of N alloy components is used for explanation. However, even if the items of alloy components included in C0, C1, C2, and C3 are different from each other, the parameters can be determined using actual performance data in the same manner as above.

[0234] In addition, the mathematical model of the above-mentioned class coefficient is an example of a linear function of the alloy composition, but the mathematical model of the class coefficient can also be a nonlinear function. In the case of a nonlinear function, a nonlinear optimization method of a continuous function can be used to determine the parameters. As such a nonlinear optimization method, a variety of methods such as Newton's method, Simplex method, Simulated annealing method, Levenberg-Marquardt method, etc. are well known.

[0235] As shown in the above example, the parameters of the deformation resistance category model 128 of the present invention are determined using the performance data belonging to the category.

[0236] In addition, although Fig.11 Although not shown in FIG. 4 , a step of evaluating the accuracy of the adjusted class model 128 may be inserted after executing the class model adjustment step S405 . Figure 6The process is simply performed by combining the class model application process S204 and the accuracy evaluation process S205 in the prediction accuracy evaluation process S2 shown in the figure. In addition, when the accuracy of the adjusted class model 128 is still insufficient, it is also possible to perform adjustment again.

[0237] Next, the deformation resistance prediction device 12 executes the class model storage step S406 to store the parameters of the class model 128 determined in the class model adjustment step S405 in the deformation resistance prediction device 12 .

[0238] Fig.12 1 shows an example of the stored category model 128. The category model 128 includes a category 1281, a storage date and time 1282, a model number 1283, a coefficient 1285, Const 1286, C 1287, Si 1288, and Mn 1289 in one record.

[0239] The category 1281 stores the identifier of the category of the rolled material, the storage date and time 1282 stores the date and time when the category model 128 is stored, the model number 1283 stores the number of the category model 128, and the coefficient 1285 stores the parameter of each coefficient of C0, C1, C2, and C3 as shown.

[0240] In addition, Const1286 is a constant term, C1287 is a term for carbon in the alloy component, and the terms Si1288 and Mn1289 are terms for silicon and manganese, respectively. In addition, the terms related to the alloy components are not limited to the above content, and the fields can be set according to the alloy components contained in the rolled material.

[0241] The deformation resistance prediction device 12 performs an adjustment completion determination S407 to determine whether the adjustment of all the categories selected as the adjustment target categories has been completed, and ends the category model adjustment step S4 if the adjustment has been completed.

[0242] In the present invention, the class model 128 of a class determined to have low prediction accuracy of deformation resistance is adjusted to improve prediction accuracy. Therefore, the class model 128 for one class increases every time adjustment is performed.

[0243] For example, if there are changes in the conditions of upstream processes such as billet manufacturing and reheating, or changes in the equipment of upstream processes, or changes in rolling equipment, or changes in sensors installed in the hot rolling line 3, the predicted accuracy of deformation resistance may change.

[0244] In addition, the magnitude of the change in the prediction accuracy of the deformation resistance and the direction of increase or decrease vary according to the category. In the present invention, there is a process of evaluating the prediction accuracy of the deformation resistance by category at any time, and a process of selectively adjusting the prediction model 123 of the deformation resistance of the category with insufficient prediction accuracy, thereby achieving an improvement in the prediction accuracy of the deformation resistance and a reduction in rolling failure.

[0245] Fig.13 and Fig.14 A distribution diagram showing predicted values ​​and actual values ​​of deformation resistance of the technology of the present invention and the prior art. Fig.13 The prediction results of the technology of the present invention show that the RMSE (Root Mean Squared Error) is 14 MPa. On the other hand, Fig.14 The RMSE of the prediction result of the prior art shown is 34 MPa. The technology of the present invention reduces the RMSE by about 60% from the RMSE of the prior art, and can predict the deformation resistance with high accuracy compared with the prior art.

[0246] Fig.15 The schematic structure of the deformation resistance prediction device 12 of the present invention is shown. In one embodiment of the present invention, the deformation resistance prediction device 12 is composed of an information processing device, including: a processor 21, a memory 22, a storage device 23, a network interface (NE / IF in the figure) 24, an input device 125 and an output device 126.

[0247] The network interface (NE / IF in the figure) 24 is connected to each device (rolling condition determination device 11, rolling performance data storage device 13, etc.) of the hot rolling line control system 1 via the network 30. The input device 125 is composed of a keyboard, a mouse or a touch panel. The output device 126 is composed of a display.

[0248] The storage device 23 constituted by a nonvolatile storage device stores a prediction model 123 including a global model 127 and a class model 128 , an accuracy history 124 , a deformation resistance prediction value 129 , and rolling performance data 130 .

[0249] The deformation resistance prediction unit 121 is loaded into the memory 22 as a program and executed by the processor 21 .

[0250] The processor 21 executes processing according to the programs of each functional unit, thereby operating as a functional unit that provides a predetermined function. For example, the processor 21 executes processing according to the deformation resistance prediction program, thereby functioning as the deformation resistance prediction unit 121. The same is true for other programs. In addition, the processor 21 also operates as a functional unit that provides the functions of each of the multiple processes executed by each program. Computers and computer systems are devices and systems that include these functional units.

[0251] In addition, the input device 125 may be shared with the input unit of the hot rolling line control system 1. In addition, the output device 126 may be shared with the output unit of the hot rolling line control system 1.

[0252] When the deformation resistance prediction device 12 executes the deformation resistance prediction process S1, the prediction accuracy evaluation process S2, the determination process S3 of whether model adjustment is required, and the category model adjustment process S4, the processor 21 reads the deformation resistance prediction program as computer software stored in the storage device 23 into the memory 22 and executes it.

[0253] The deformation resistance prediction program includes various element software including a rolling data acquisition unit 12101, an abnormal data exclusion unit 12102, a global model application unit 12103, a category model application unit 12104, a deformation resistance prediction value output unit 12105, an accuracy evaluation unit 12106, an accuracy display unit 12107, a category model adjustment unit 12108, and a category model storage unit 12109.

[0254] In the above-mentioned element software, the rolling data acquisition unit 12101 has a function of acquiring the rolling conditions 110 or the operation data 150 from the rolling condition determination device 11 or the rolling performance data storage device 13. The rolling data acquisition unit 12101 can be a functional unit different from the deformation resistance prediction unit 121.

[0255] Abnormal data elimination unit 12102 executes Figure 6 The global model application unit 12103 executes the abnormal data elimination step S202. Figure 6 The global model application step S102. The category model application unit 12104 executes Figure 6 The category model application step S204.

[0256] The deformation resistance prediction value output unit 12105 executes Figure 3 The deformation resistance prediction value output step S104. The accuracy evaluation unit 12106 executes Figure 6 The accuracy evaluation step S205. The accuracy display unit 12107 performs Figure 6 The accuracy of the display step S206.

[0257] Category model adjustment unit 12108 performs Fig.11 The category model storage unit 12109 executes the category model adjustment step S405. Fig.11 The category model saving step S406.

[0258] Summary

[0259] In the above embodiment, the deformation resistance prediction device 12 is applied to the hot rolling line control system 1, but the present invention is not limited thereto. The deformation resistance prediction device 12 can be applied to a warm rolling line or a cold rolling line instead of the hot rolling line 3 of the above embodiment.

[0260] As described above, the hot rolling line control system 1 (or the deformation resistance prediction system) of the above-mentioned embodiment can be configured as follows.

[0261] (1) A deformation resistance prediction system, comprising: a rolling condition determination device 11 for determining a rolling condition 110 to be set for a rolling device (hot rolling line 3); a storage device (rolling performance data storage device 13) for collecting operation data 150 of rolling performed by the rolling device 3 and the rolling condition 110; a deformation resistance prediction device having a processor 21 and a memory 22, which predicts the deformation resistance of a rolled material based on the operation data 150 and the rolling condition 110, wherein the deformation resistance prediction device 12 comprises: a rolling data acquisition unit 12101, which acquires the operation data 150 and the rolling condition 110 and accumulates them as rolling performance data 130, and acquires a candidate for the rolling condition 110 from the rolling condition determination device 11; and a prediction model 123, which predicts the deformation resistance of the rolled material. The deformation resistance of the rolled material; and a deformation resistance prediction unit 121, which uses the prediction model 123 to calculate the predicted value of the deformation resistance of the rolled material, the prediction model 123 includes: a global model 127, which is determined based on the overall rolling performance data 130 and is pre-set as a prediction model for estimating the deformation resistance of each steel type of the rolled material; and a category model 128, which is determined based on the rolling performance data 130 of the deformation resistance of a specific steel type with a similar or common alloy composition of the rolled material in the rolling performance data 130, and is pre-set as a prediction model for estimating the deformation resistance of the specific steel type, the deformation resistance prediction unit 121 uses the prediction model 123 to calculate the predicted value K of the deformation resistance of the rolled material based on the rolling condition 110 candidate. p .

[0262] According to the above structure, the deformation resistance is predicted by overlapping the global model of deformation resistance determined by the rolling performance data of multiple steel types and the category model of deformation resistance determined by the rolling performance data of one or a few (specific) steel types. Thus, even for steel types with relatively few rolling performance data, the deformation resistance can be predicted with an accuracy at least higher than that of the global model of deformation resistance.

[0263] (2) A deformation resistance prediction system according to the above (1), wherein the deformation resistance prediction unit 121 comprises: a global model application unit 12103, which applies the global model 127 to the rolling condition candidate 113, and calculates a global prediction value of the deformation resistance of the rolled material of the rolling condition candidate 113; and a category model application unit 12104, which applies the category model 128 to the rolling condition candidate 113, and calculates a category correction value of the deformation resistance of the rolled material of the rolling condition candidate 113, and calculates a prediction value of the deformation resistance for the rolling condition candidate 110 based on the category correction value and the global prediction value.

[0264] According to the above structure, the deformation resistance prediction device 12 grasps the overall (global) deformation resistance trend of the rolled material through the global model 127, grasps the trend of the deformation resistance within the category (specific steel grade) different from the overall trend through the category model, calculates the predicted value of the deformation resistance, and can provide the predicted value of the deformation resistance with high accuracy to the rolling condition determination device 11.

[0265] (3) According to the deformation resistance prediction system described in (1) above, the deformation resistance prediction unit 121 also has: an accuracy evaluation unit 12106, which calculates the prediction accuracy of the prediction model based on the predicted value of the deformation resistance and the actual value of the deformation resistance contained in the rolling performance data 130.

[0266] According to the above configuration, by evaluating the prediction accuracy of the prediction model, it is possible to determine whether the classification model needs to be adjusted.

[0267] (4) According to the deformation resistance prediction system described in (3) above, the deformation resistance prediction system further has: an accuracy display unit 12107, which generates a change history of the prediction accuracy of the category model 128 based on the prediction accuracy of the prediction model 123.

[0268] According to the above configuration, by understanding changes in the prediction accuracy of the prediction model in a time series manner, it is possible to determine whether the classification model needs to be adjusted.

[0269] (5) According to the deformation resistance prediction system described in (1) above, the deformation resistance prediction unit 121 also has: a category model adjustment unit 12108, which adjusts the category model 128 according to the performance value of the deformation resistance of a specific steel type included in the rolling performance data 130.

[0270] According to the above configuration, the classification model can be adjusted based on the rolling performance data 130 of the steel type corresponding to the classification model 128 .

[0271] (6) The deformation resistance prediction system according to (1) above, wherein the classification model adjustment unit 12108 adjusts the classification model 128 for the selected classification model 128 using the rolling performance data 130 of a specific steel grade.

[0272] (7) According to the deformation resistance prediction system described in (3) above, the deformation resistance prediction unit 121 also has: an abnormal data exclusion unit, which excludes data whose values ​​of the rolling performance data 130 exceed a predetermined range as abnormal data, and the accuracy evaluation unit 12106 calculates the prediction accuracy of the prediction model based on the rolling performance data 130 excluding the abnormal data.

[0273] According to the above configuration, the accuracy evaluation unit 12106 calculates the prediction accuracy of the prediction model based on the rolling performance data 130 from which abnormal data is excluded, thereby being able to accurately calculate the prediction accuracy.

[0274] (8) A deformation resistance prediction system according to the above (1), wherein the global model 127 is formed by multiplying two factors, the first factor of the two factors is formed by adding a constant term, a logarithmic term of the rolling strain rate, and a power term of the temperature, and the second factor of the two factors is formed by the power of the rolling strain.

[0275] According to the above configuration, a global model of deformation resistance can be expressed based on the rolling performance data 130 of a plurality of steel types.

[0276] (9) A deformation resistance prediction system according to the above (1), wherein the category model 128 is formed by multiplying two factors, the first factor of the two factors is composed of an exponential function having an exponential part, the exponential part including a constant term, a logarithmic term of the rolling strain rate, and a power term of the temperature, and the second factor of the two factors is composed of the power of the rolling strain.

[0277] According to the above configuration, it is possible to express a classification model of deformation resistance based on the rolling performance data 130 of a specific steel type.

[0278] In addition, the present invention is not limited to the above-mentioned embodiments, and includes various modified examples. For example, the above-mentioned embodiments are described in detail in order to explain the present invention in an easy-to-understand manner, and are not limited to all the structures described. In addition, a part of the structure of a certain embodiment can be replaced with the structure of other embodiments, and in addition, the structure of other embodiments can be added to the structure of a certain embodiment. In addition, for a part of the structure of each embodiment, the addition, deletion or replacement of other structures can be applied individually or in combination.

[0279] In addition, part or all of the above-mentioned structures, functions, processing parts and processing units, etc. may also be implemented by hardware, for example, by integrated circuit design, etc. In addition, the above-mentioned structures and functions, etc. may also be implemented by software by a processor interpreting and executing programs that implement each function. Information such as programs, tables, files, etc. that implement each function can be placed in a recording device such as a memory, a hard disk, an SSD (Solid State Drive), or a recording medium such as an IC card, an SD card, or a DVD.

[0280] In addition, the control lines and information lines are shown as those considered necessary for explanation, and not necessarily all the control lines and information lines on the product are shown. In fact, it can be considered that almost all the structures are connected to each other.

Claims

1. A deformation resistance prediction system, comprising: a rolling condition determination device that determines rolling conditions to be set for the rolling device; a storage device that collects operation data of rolling performed by the rolling device and the rolling conditions; as well as a deformation resistance prediction device having a processor and a memory, which predicts the deformation resistance of the rolling material based on the operation data and the rolling conditions, It is characterized in that The deformation resistance prediction device comprises: a rolling data acquisition unit that acquires the operation data and the rolling conditions and accumulates them as rolling performance data, and acquires rolling condition candidates from the rolling condition determination device; a prediction model that predicts the deformation resistance of the rolled material; and a deformation resistance prediction unit that calculates a predicted value of the deformation resistance of the rolled material using the prediction model, The prediction model comprises: a global model determined based on the entire rolling performance data and pre-set as a prediction model for estimating the deformation resistance of each steel grade of the rolled material; and a classification model determined based on rolling performance data of deformation resistance of a specific steel type having a similar or common alloy composition of the rolled material in the rolling performance data, and pre-set as a prediction model for estimating the deformation resistance of the specific steel type; The deformation resistance prediction unit calculates a predicted value of the deformation resistance of the rolling material using the global model and the class model based on the rolling condition candidate. The deformation resistance prediction unit comprises: a global model application unit that applies the global model to the candidate rolling conditions to calculate a global prediction value of the deformation resistance of the rolling material of the candidate rolling conditions; as well as A category model application unit applies the category model to the rolling condition candidate, calculates a category correction value of the deformation resistance of the rolling material of the rolling condition candidate, and calculates a predicted value of the deformation resistance for the rolling condition candidate based on the category correction value and the global prediction value.

2. The deformation resistance prediction system according to claim 1, characterized in that: The deformation resistance prediction unit further includes an accuracy evaluation unit that calculates prediction accuracy of a prediction model based on the predicted value of the deformation resistance and an actual value of the deformation resistance included in the rolling performance data.

3. The deformation resistance prediction system according to claim 2, characterized in that: The deformation resistance prediction system further includes an accuracy display unit that generates a change history of the prediction accuracy of the class model based on the prediction accuracy of the prediction model.

4. The deformation resistance prediction system according to claim 1, characterized in that: The deformation resistance prediction unit further includes a classification model adjustment unit that adjusts the classification model based on actual values ​​of deformation resistance of a specific steel type included in the rolling performance data.

5. The deformation resistance prediction system according to claim 4, characterized in that: The classification model adjustment unit adjusts the selected classification model using the rolling performance data of a specific steel grade.

6. The deformation resistance prediction system according to claim 2, characterized in that: The deformation resistance prediction unit further includes an abnormal data exclusion unit that excludes data whose values ​​of the rolling performance data exceed a predetermined range as abnormal data. The accuracy evaluation unit calculates the prediction accuracy of the prediction model based on the rolling performance data from which the abnormal data is excluded.

7. The deformation resistance prediction system according to claim 1, characterized in that: The global model is constructed by multiplying two factors, wherein the first factor of the two factors includes a constant term, a logarithmic term of rolling strain rate, and a power term of temperature, and the second factor of the two factors includes the power of rolling strain.

8. The deformation resistance prediction system according to claim 1, characterized in that: The category model is constructed by multiplying two factors, wherein the first factor of the two factors is composed of an exponential function having an exponential part, the exponential part including the addition of a constant term, a logarithmic term of the rolling strain rate, and a power term of the temperature, and the second factor of the two factors includes the power of the rolling strain.

9. A deformation resistance prediction method for predicting the deformation resistance of a rolling material by a deformation resistance prediction device having a processor and a memory, characterized in that: Include: In a first step, the deformation resistance prediction device obtains a rolling condition candidate from a rolling condition determination device that determines a rolling condition to be set for the rolling device; In a second step, the deformation resistance prediction device acquires the operation data and the rolling conditions from a storage device that collects the operation data and the rolling conditions of the rolling performed by the rolling device and accumulates them as rolling performance data; as well as In a third step, the deformation resistance prediction device calculates a predicted value of the deformation resistance of the rolled material using a prediction model for predicting the deformation resistance of the rolled material. In the third step, the prediction model composed of a global model and a category model is used to calculate a predicted value of the deformation resistance of the rolled material based on the rolling condition candidate, wherein the global model is determined based on the overall rolling performance data and is pre-set as a model for estimating the deformation resistance of each steel grade of the rolled material, and the category model is determined based on the rolling performance data of the deformation resistance of a specific steel grade having a similar or common alloy composition of the rolled material in the rolling performance data and is pre-set as a model for estimating the deformation resistance of the specific steel grade. The third step comprises: A step of applying the global model to the candidate rolling conditions to calculate a global prediction value of the deformation resistance of the rolling material of the candidate rolling conditions; as well as The step of applying the category model to the rolling condition candidate, calculating a category correction value of the deformation resistance of the rolling material of the rolling condition candidate, and calculating a predicted value of the deformation resistance for the rolling condition candidate based on the category correction value and the global predicted value.

10. A deformation resistance prediction device, comprising a processor and a memory, for predicting the deformation resistance of a rolled material, characterized in that: The deformation resistance prediction device comprises: A rolling data acquisition unit that acquires rolling operation data and rolling conditions and accumulates them as rolling performance data, and receives rolling condition candidates; a prediction model that predicts the deformation resistance of the rolled material; as well as a deformation resistance prediction unit that calculates a predicted value of the deformation resistance of the rolled material using the prediction model, The prediction model has: a global model determined based on the entire rolling performance data and pre-set as a prediction model for estimating the deformation resistance of each steel grade of the rolled material; as well as a classification model determined based on rolling performance data of deformation resistance of a specific steel type having a similar or common alloy composition of the rolled material in the rolling performance data, and pre-set as a prediction model for estimating the deformation resistance of the specific steel type; The deformation resistance prediction unit calculates a predicted value of the deformation resistance of the rolling material using the global model and the class model based on the rolling condition candidate. The deformation resistance prediction unit comprises: a global model application unit that applies the global model to the candidate rolling conditions to calculate a global prediction value of the deformation resistance of the rolling material of the candidate rolling conditions; as well as A category model application unit applies the category model to the rolling condition candidate, calculates a category correction value of the deformation resistance of the rolling material of the rolling condition candidate, and calculates a predicted value of the deformation resistance for the rolling condition candidate based on the category correction value and the global prediction value.

Images