Method and system for optimizing mill model predictions of rolling forces

Abstract

The application provides a method and system for optimizing rolling force prediction of a rolling mill model, the method comprising: constructing an initial database, the initial database comprising collected data; the types of the collected data comprising measured temperature at pass inlet, pass inlet thickness, pass outlet thickness, slab speed, slab width, target thickness and target width; saving measured data of the current slab according to the rolling mill model when the steel plate rolling is completed; constructing a measured database according to the measured data, each type of the measured data in the measured database being divided into multiple data intervals; calculating update coefficients by using the least square method on the measured data in the multiple data intervals; comparing three coefficients to update the rolling mill model; predicting new slab rolling force according to the updated rolling mill model; and finally comparing the slab rolling force and the predicted rolling force in the same measured data interval, and marking the correct rolling force through the difference, so as to solve the problem of inaccurate rolling force.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of rolling force prediction, in particular to a method and system for optimizing rolling force prediction of rolling mill model. BACKGROUND

[0002] In the process of high temperature deformation of metal, the dynamic behavior of the material (including work hardening, dynamic softening) has a significant impact on the performance of the material. As the main mechanism of dynamic softening of the material, dynamic recrystallization is the main means to refine the grain of the material. Dynamic recrystallization refers to the recrystallization phenomenon of metal in the process of thermal deformation. Compared with static recrystallization which occurs between passes of thermal deformation and after deformation is completed and heating and cooling, dynamic recrystallization has the following characteristics: dynamic recrystallization requires a critical deformation amount and occurs at a relatively high deformation temperature; similar to static recrystallization, dynamic recrystallization is prone to nucleation at grain boundaries and subgrain boundaries; dynamic recrystallization does not require an incubation period when it is converted to static recrystallization; the time required for dynamic recrystallization is shortened as the temperature increases.

[0003] The specific process of dynamic recrystallization depends on the thermal deformation parameters such as deformation temperature and deformation rate. Therefore, determining reasonable deformation parameters is of great significance for obtaining fine grains and improving the organization and performance of the product. Therefore, a rolling mill model is proposed. The rolling mill model refers to establishing a finite element three-dimensional dynamic rolling mill model based on the process parameters collected in the production field during the rolling process of the rolling mill. The material properties of the billet, the contact friction properties, and the relative motion properties between the rolling mill pass and the billet are considered. The influence of the division rules of the billet unit grid on the rolling result is discussed, and a three-dimensional dynamic simulation model that is more suitable for the rolling condition is obtained.

[0004] Most rolling mill models only consider thickness, but in fact, different pass entry temperatures, pass entry thicknesses, pass exit thicknesses, slab speeds, and slab widths have a significant impact on the deformation resistance and the dynamic recrystallization behavior of the hot deformed austenite (the peak stress can be used as a characteristic parameter of dynamic recrystallization), making the rolling force applied in the process inaccurate. SUMMARY

[0005] The present application provides a method and system for optimizing rolling force prediction of rolling mill model to solve the problem of inaccurate rolling force.

[0006] To solve the above problems, the application provides a method for optimizing rolling mill model to predict rolling force, which comprises the following steps: constructing an initial database, wherein the initial database comprises collected data of a primary controller, and the collected data is divided into training set data and verification set data; the collected data comprises pass entry measured temperature, pass entry thickness, pass exit thickness, slab speed, slab width, target thickness and target width; the training set data of each type is divided into multiple training set data intervals; the verification set data of each type is divided into multiple verification set data intervals, and the historical deformation resistance of each verification set data interval is calculated according to the type of the verification set data; the training set coefficients of the multiple training set data intervals are calculated according to the initial database and the rolling mill model; when the steel plate rolling is completed, the measured data of the current slab is saved according to the rolling mill model; a measured database is constructed according to the measured data, wherein the measured database comprises measured data, the measured data comprises pass entry measured temperature, pass entry thickness, pass exit thickness, slab speed, slab width, target thickness and target width, and the measured data of each type is divided into multiple data intervals; the update coefficients are calculated by using the least square method on the measured data in the multiple data intervals in the measured database; the training set coefficients, the verification set coefficients and the update coefficients are compared, and the rolling mill model is updated according to the comparison result; the new slab rolling force is predicted according to the updated rolling mill model; the same slab rolling force as the measured data interval is obtained to obtain the measured rolling force; the difference between the measured rolling force and the new slab rolling force is compared; if the difference is less than or equal to 200 tons, the new slab rolling force is marked as the correct rolling force; if the difference is greater than 200 tons, the measured rolling force is marked as the correct rolling force.

[0007] Optionally, the pass entry measured temperature data interval is 50 DEG C as an interval, the pass entry thickness data interval is 2 mm as an interval, the pass exit thickness data interval is 2 mm as an interval, the slab width data interval is 10 cm as an interval, the slab thickness data interval is 20 mm as an interval, the target width data interval is 10 cm as an interval, and the target thickness data interval is 2 mm as an interval.

[0008] Optionally, the method further comprises the following steps:

[0009] The strain rate and the strain speed are calculated according to the measured database.

[0010] The deformation resistance of the current slab is calculated according to the update coefficients, the strain rate and the strain speed.

[0011] Optionally, in the step of calculating the strain rate and the strain rate according to the measured database, the strain rate is calculated according to the following formula:

[0012] strain = inThk / outThk

[0013] Wherein, strain is the strain rate, inThk is the thickness at the exit of the previous pass, and outThk is the thickness at the exit of the current pass.

[0014] The strain rate is calculated according to the following formula:

[0015] strainrate = strain * pce_Spd / arc

[0016] Wherein, strainrate is the strain rate, strain is the strain rate, pce_Spd is the linear speed of the steel plate passing through the roller of the rolling mill, and arc is the length of the contact arc between the working roller of the rolling mill and the steel plate.

[0017] Optionally, in the step of calculating the deformation resistance of the current billet according to the updated coefficient, the strain rate and the strain rate, the deformation resistance is calculated according to the following formula:

[0018] fs = C1e C2 / RT strain C3 strainrate C4

[0019] Wherein, fs is the deformation resistance, e is the natural base, R is a physical constant relating various thermodynamic functions in the equation of state, T is the measured temperature at the pass inlet, strain is the strain rate, strainrate is the strain rate, C1 is the hardness of the steel plate, C2 is the reaction activation energy, C3 is the strain index, and C4 is the strain rate index.

[0020] Optionally, the method further comprises:

[0021] putting one of the training set data intervals of the measured temperature at the pass inlet, the thickness at the pass inlet, the thickness at the pass outlet, the speed of the slab, the width of the slab, the target thickness and the target width into a solution matrix, and calculating the C1 hardness of the steel plate, the C2 reaction activation energy, the C3 strain index and the C4 strain rate index by using the least square method and the solution matrix;

[0022] Wherein, the solution matrix is:

[0023] lnfs = lnC1 + C2(1 / RT) + C3ln(strain) + C4ln(strainrate)

[0024]

[0025] wherein, lnC1 is the whole as the coefficient to be solved.

[0026] Optionally, according to the comparison result of the training set coefficient, the verification set coefficient and the updated coefficient, the step of updating the rolling mill model comprises:

[0027] If the training set coefficient is the same as the updated coefficient or the error of the training set coefficient and the updated coefficient is within the error interval, the training set coefficient is replaced by the updated coefficient;

[0028] If the error of the training set coefficient and the updated coefficient is not within the error interval, the training set coefficient is compared with the verification set coefficient;

[0029] If the training set coefficient is the same as the verification set coefficient or the error of the training set coefficient and the verification set coefficient is within the error interval, the training set coefficient is replaced by the verification set coefficient;

[0030] If the error of the training set coefficient and the verification set coefficient is not within the error interval, the training set coefficient in the initial database is maintained, and the verification set coefficient is discarded.

[0031] Optionally, according to the updated rolling mill model, the step of predicting the rolling force of the new billet comprises: calculating the rolling force of the new billet according to the following formula:

[0032] Rolling force = fs*arc*Factr*with

[0033] Wherein, arc is the contact arc length between the work roll and the workpiece, Factr is the geometric term friction condition of the viscous under-rolling bite equation, and with is the width of the current billet.

[0034] Optionally, according to the updated rolling mill model, the step of predicting the rolling force of the new billet further comprises: calculating the reduction table of each pass in the rolling process, the reduction table comprising the thickness of the steel plate, the length of the steel plate, the width of the steel plate, the temperature of the steel plate, the rolling force of the rolling mill, the linear speed of the steel plate passing through the rolling mill roll, and the reduction amount.

[0035] In another aspect, the application provides a system for optimizing the rolling mill model to predict the rolling force, comprising: a collection module, a storage module, a prediction training module, a model module and a comparison module.

[0036] The collection module is used to collect the measured data in the rolling process of the collection data of the primary controller.

[0037] The storage module is used to save the measured data of the current billet according to the rolling mill model when the steel plate rolling is completed, and save the measured data of each pass when the billet is rolled.

[0038] The prediction training module is configured to calculate updating coefficients by using the least square method on the measured data in the data intervals in the measured database, compare the training set coefficients, the validation set coefficients and the updating coefficients, and update the rolling mill model according to the comparison result; calculate updating coefficients by using the least square method on the data in the three intervals in the measured database; the prediction training module is further configured to compare the original coefficients and the updating coefficients, and update the rolling mill model according to the comparison result;

[0039] The model module is configured to predict new billet rolling forces according to the updated rolling mill model;

[0040] The comparison module is configured to obtain the same billet rolling forces as the measured data intervals to obtain measured rolling forces, compare the difference between the measured rolling forces and the new billet rolling forces, mark the new billet rolling forces as correct rolling forces if the difference is less than or equal to 200 tons, and mark the measured rolling forces as correct rolling forces if the difference is greater than 200 tons.

[0041] According to the technical scheme, the application provides a method and system for optimizing a rolling mill model to predict rolling force. The method comprises: constructing an initial database, wherein the initial database comprises collected data of a primary controller, and the collected data is divided into training set data and verification set data; the types of the collected data include measured temperature at pass entrance, pass entrance thickness, pass exit thickness, slab speed, slab width, target thickness, and target width; the training set data of each type is divided into multiple training set data intervals; the verification set data of each type is divided into multiple verification set data intervals, and historical deformation resistance of each verification set data interval is calculated according to the types of the verification set data; the historical deformation resistance is used to solve verification set coefficients of the multiple verification set data intervals; training set coefficients of the multiple training set data intervals are calculated according to the initial database and a rolling mill model; when steel plate rolling is completed, measured data of the current slab is saved according to the rolling mill model; a measured database is constructed according to the measured data, wherein the measured database comprises measured data, the types of the measured data include measured temperature at pass entrance, pass entrance thickness, pass exit thickness, slab speed, slab width, target thickness, and target width, and the measured data of each type is divided into multiple data intervals; least square method is used to calculate updated coefficients according to the measured data in the multiple data intervals in the measured database; the training set coefficients, the verification set coefficients, and the updated coefficients are compared, and the rolling mill model is updated according to the comparison result; new slab rolling force is predicted according to the updated rolling mill model; the same slab rolling force as the measured data interval is obtained to obtain measured rolling force; the difference between the measured rolling force and the new slab rolling force is compared; if the difference is less than or equal to 200 tons, the new slab rolling force is marked as correct rolling force; if the difference is greater than 200 tons, the measured rolling force is marked as correct rolling force, so as to solve the problem of inaccurate rolling force.

[0042] The application divides the collected data into training set and verification set data during the billet rolling process, and calculates the pass entry measured temperature, pass entry thickness, pass exit thickness, slab speed, slab width, target thickness and target width of the training set and verification set data in different zones to obtain the training set and verification set coefficients in different zones, so as to calculate the corresponding deformation resistance, improve the coefficients of the rolling mill model through long-term self-learning, and more accurately predict the rolling force of the next billet under the same pass entry measured temperature, pass entry thickness, pass exit thickness, slab speed, slab width, target thickness and target width interval; the long-term self-learning is adopted, the coefficients can be continuously learned according to the current field equipment working condition, so as to reduce the error between the predicted rolling force and the field measured rolling force, and also reduce the error between the target thickness of the completed rolling and the field measured target thickness, and finally compare the billet rolling force and the predicted rolling force in the same measured data interval, and mark the correct rolling force through the difference, so as to make the prediction of the rolling force more accurate. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. Obviously, for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0044] Figure 1 A method flowchart for optimizing the rolling mill model to predict the rolling force is provided for the embodiments of the present application.

[0045] Figure 2 A process schematic diagram after the billet arrives is provided for the embodiments of the present application.

[0046] Figure 3 A system structure schematic diagram for optimizing the rolling mill model to predict the rolling force is provided for the embodiments of the present application.

[0047] Figure 4 A system data flow direction schematic diagram for optimizing the rolling mill model to predict the rolling force is provided for the embodiments of the present application. DETAILED DESCRIPTION

[0048] The embodiments will be described in detail below, and the examples are shown in the drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementation described in the following embodiments does not represent all the implementations consistent with the present application. It is only an example of the system and method consistent with some aspects of the present application as described in detail in the claims.

[0049] Rolling force is an important equipment parameter and process parameter of rolling mill, and is also an important basis for plastic processing technology, equipment optimization design and process control. The precision of rolling force not only directly affects the setting precision of rolling schedule, but also directly affects the plate thickness precision and plate shape quality, and is the key to fully exert the plate thickness and plate shape system control ability and improve the head hitting rate of strip steel. The precision of rolling force directly determines the yield and quality of rolling products. Moreover, different pass entry temperature, pass entry thickness, pass exit thickness, slab speed and slab width have obvious influence on deformation resistance and dynamic recrystallization behavior of hot deformed austenite, so most rolling mill models do not consider the influence of dynamic recrystallization on steel plate softening, resulting in low prediction accuracy of rolling force.

[0050] To solve the above problems, the method for optimizing the rolling mill model to predict the rolling force provided by some embodiments of the present application is described with reference to Figure 1 The method for optimizing the rolling mill model to predict the rolling force is a flow chart. The method comprises the following steps:

[0051] S100: Construct an initial database.

[0052] The initial database includes the collected data of the primary controller, and the collected data is divided into training set data and validation set data. The types of collected data include pass entry measured temperature, pass entry thickness, pass exit thickness, slab speed, slab width, target thickness and target width. The collected data of each type is divided into multiple training set intervals. The historical deformation resistance of each validation set data interval is calculated by the type of validation set data.

[0053] When the steel is rolled, the primary controller collects data during rolling, such as temperature, strain rate, strain rate, pass entry measured temperature, pass entry thickness, pass exit thickness, slab speed, slab width, target thickness, target width, steel plate thickness, steel plate length, steel plate width, steel plate linear speed through the rolling mill, etc. The data can be adjusted according to the needs. Specifically, the primary controller is a programmable logic controller (PLC). The PLC uses a programmable memory to store instructions for performing logic operations, sequential control, timing, counting and arithmetic operations, etc. inside it. That is, the data to be collected can be completed by programming, which can be flexibly operated according to the production needs, and is efficient and convenient.

[0054] The entry measured temperature of different deformation passes, the entry thickness of different deformation passes, the exit thickness of different deformation passes, the slab speed, the slab width, the target thickness and the target width have obvious influences on the deformation resistance and the dynamic recrystallization behavior of hot deformed austenite, and the dynamic recrystallization can cause the softening of the steel plate, so the embodiment proposes that the data collected by the PLC are divided into different types of entry measured temperature of different deformation passes, entry thickness of different deformation passes, exit thickness of different deformation passes, slab speed, slab width, target thickness and target width, and the data of different types are divided into multiple training set and validation set data intervals, so as to reflect the influences of the entry measured temperature of different deformation passes, the entry thickness of different deformation passes, the exit thickness of different deformation passes, the slab speed, the slab width, the target thickness and the target width on the dynamic recrystallization, and the rolling mill model can calculate the softening coefficient or the softening degree of the steel plate, and the rolling mill model can calculate the rolling force based on this to predict the rolling force of the next steel slab of the same material, and the rolling mill model can further predict the data required in other rolling processes through the rolling force.

[0055] S200: The validation set coefficients of the multiple validation set data intervals are solved through the historical deformation resistance; and the training set coefficients of the multiple training set data intervals are calculated through the rolling mill model according to the initial database.

[0056] The rolling mill model can calculate the training set parameters through the entry measured temperature of different deformation passes, the entry thickness of different deformation passes, the exit thickness of different deformation passes, the slab speed, the slab width, the target thickness and the target width of the rolling mill in the production field collected by the primary controller, establish a finite element three-dimensional dynamic rolling mill model through the parameters, obtain a three-dimensional dynamic simulation model that is more consistent with the rolling working condition, and thus use the rolling mill model to predict the rolling force, and realize that the predicted rolling force is closer and closer to the actual production rolling force through the updating and covering of the coefficients, so as to improve the work efficiency. The validation set coefficients of different types of validation set data intervals are calculated through the historical deformation resistance, and used as the comparison data.

[0057] S300: When the steel plate rolling is completed, the measured data of the current steel slab is saved according to the rolling mill model.

[0058] In some embodiments, the rolling mill model is used to save the measured data of the current steel slab, and other storage modules can also be used for saving, the rolling mill model can save the data that is adopted as the updating coefficient, and the storage module can save all the required data.

[0059] S400: An measured database is constructed according to the measured data.

[0060] The measured database includes the measured data, and the types of the measured data include the entry measured temperature of different deformation passes, the entry thickness of different deformation passes, the exit thickness of different deformation passes, the slab speed, the slab width, the target thickness and the target width, and each type of measured data is divided into multiple data intervals.

[0061] Specifically, when the steel plate of the same material is rolled, the measured data of the billet is collected by the parameters required by the rolling mill model, and the measured data also includes the measured temperature at the pass inlet, the thickness at the pass inlet, the thickness at the pass outlet, the speed of the slab, the width of the slab, the target thickness and the target width of different types. Similarly, the measured data of each type is divided into multiple intervals according to the interval, and it can be understood that in order to facilitate the calculation of the rolling mill model, the data interval of the measured database is the same as the endpoint of the data interval of the initial database.

[0062] S500: The measured data in the multiple data intervals in the measured database is calculated by the least square method to obtain the update coefficient.

[0063] According to the difference between the parameters calculated in the initial database, the data interval in the measured database is calculated by the least square method to obtain the update coefficient; specifically, one of the measured temperature at the pass inlet, the thickness at the pass inlet, the thickness at the pass outlet, the speed of the slab, the width of the slab, the target thickness and the target width of the training set data interval is put into the solution matrix, and the least square method and the solution matrix are used to calculate C1 steel plate hardness, C2 reaction activation energy, C3 strain index and C4 strain rate index; wherein the solution matrix is:

[0064] lnfs=lnC1+C2(1 / RT)+C3ln(strain)+C4ln(strainrate)

[0065]

[0066] Specifically, lnC1 is taken as a coefficient to be solved, that is, lnC1 is the hardness of the steel plate, C2 is the reaction activation energy, the reaction activation energy refers to the energy required for molecules to change from normal state to active state prone to chemical reaction, C3 is the strain index, and C4 is the strain rate index. Cover the above coefficients with the original coefficients in the rolling mill model, that is, update the data of the rolling mill model.

[0067] S600: Compare the training set coefficient, the verification set coefficient and the update coefficient, and update the rolling mill model according to the comparison result.

[0068] Specifically, if the original training set coefficient is the same as the updated coefficient or the error between the original training set coefficient and the updated coefficient is within the error interval, the original training set coefficient is replaced by the updated coefficient; if the error between the training set coefficient and the updated coefficient is not within the error interval, the training set coefficient is compared with the verification set coefficient; if the training set coefficient is the same as the verification set coefficient or the error between the training set coefficient and the verification set coefficient is within the error interval, the training set coefficient is replaced by the verification set coefficient; if the error between the training set coefficient and the verification set coefficient is not within the error interval, the training set coefficient in the initial database is maintained, and the verification set coefficient is discarded. It can be understood that the error interval mentioned above can be set according to different situations of each data interval. In a data interval, when detailed data is needed, the error interval can be set to a small range, and when no special detailed data is needed, the error interval can be set to a large range.

[0069] It should be understood that the discarded coefficient can also be saved by setting a storage module, saving the coefficient and the data in the measured database or the initial database corresponding to the coefficient. If analysis is needed, the analysis can be provided as a reference.

[0070] S700: predicting a new billet rolling force according to the updated rolling mill model.

[0071] The billet rolling force is calculated according to the following formula:

[0072] Rolling force = fs*arc*Factr*with

[0073] wherein arc is the contact arc length between the work roll and the workpiece, Factr is the geometric term friction condition of the viscous down rolling bite equation, and with is the width of the current billet. The contact arc length between the work roll and the workpiece, the geometric term friction condition of the viscous down rolling bite equation, and the width of the current billet can all be collected by PLC for use in calculating the rolling force, saving time.

[0074] In some embodiments, the pass entry measured temperature data interval is 50 °C as an interval, the pass entry thickness data interval is 2 mm as an interval, the pass exit thickness data interval is 2 mm as an interval, the slab width data interval is 10 cm as an interval, the slab thickness data interval is 20 mm as an interval, the target width data interval is 10 cm as an interval, and the target thickness data interval is 2 mm as an interval. For example, the pass entry measured temperature data type data is divided into three intervals, the temperatures are 0 °C-50 °C, 50 °C-100 °C, and 100 °C-150 °C, respectively. The data quantity in each interval can be the same or different. When the data quantity is different, the interval can be further divided according to the data quantity in the interval, for example, 100 °C-150 °C is further divided into multiple intervals.

[0075] S800: Obtain the measured rolling force of the same steel billet as the measured data interval to obtain the measured rolling force.

[0076] S900: Compare the difference between the measured rolling force and the new steel billet rolling force.

[0077] S910: If the difference is less than or equal to 200 tons, mark the new steel billet rolling force as the correct rolling force.

[0078] S920: If the difference is greater than 200 tons, mark the measured rolling force as the correct rolling force.

[0079] For example, when a new slab arrives, the same data as the measured data interval is found, the measured rolling force is calculated, the new steel billet rolling force is calculated through the updated rolling mill model, the measured rolling force and the new steel billet rolling force are compared, and the difference is less than or equal to 200 tons. If the difference is greater than 200 tons, the measured rolling force is used. It can be understood that the reduction table is generated according to the correct rolling force, that is, when the new steel billet rolling force is marked, the new steel billet rolling force is used to generate the reduction table, and when the measured rolling force is marked, the measured rolling force is used to generate the reduction table.

[0080] In some embodiments, the method further comprises:

[0081] Calculating the strain rate and the strain rate according to the measured database;

[0082] The strain rate is calculated according to the following formula:

[0083] strain=inThk / outThk

[0084] Wherein, strain is the strain rate, inThk is the exit thickness of the previous pass, and outThk is the exit thickness of the current pass.

[0085] The strain rate is calculated according to the following formula:

[0086] strainrate = strain * pce_Spd / arc

[0087] Wherein, the strain rate is strain rate, the strain rate is strain, the pce_Spd is the linear speed of the steel plate passing through the rolling mill roller, and the arc is the length of the contact arc between the working roller of the rolling mill and the steel plate.

[0088] In some embodiments, the deformation resistance of the current billet is calculated according to the updated coefficient, the strain rate and the strain rate, and the deformation resistance is calculated according to the following formula:

[0089] fs = C1e c2 / RT strain C3 strainrate C4

[0090] Wherein, the deformation resistance is fs, e is the natural base, R is the physical constant in the equation of state relating various thermodynamic functions, T is the measured temperature at the pass inlet, strain is the strain rate, strain rate is the strain rate, C1 is the hardness of the steel plate, C2 is the activation energy of the reaction, C3 is the strain index, and C4 is the strain rate index.

[0091] In some embodiments, referring to Figure 2 , the step of predicting the rolling force of the new billet according to the updated rolling mill model further comprises: calculating the reduction table of each pass in the rolling process, the reduction table comprising the thickness of the steel plate, the length of the steel plate, the width of the steel plate, the temperature of the steel plate, the rolling force of the rolling mill, the linear speed of the steel plate passing through the rolling mill roller, and the reduction amount.

[0092] Referring to Figure 3 , Figure 4 , another part of the embodiments of the present application provides a system for optimizing the rolling mill model to predict the rolling force, comprising: an acquisition module, a storage module, a prediction training module, a model module and a comparison module; wherein the acquisition module is used to acquire the acquisition data of the primary controller; the storage module is used to save the measured data of the current billet according to the rolling mill model when the steel plate rolling is completed; the prediction training module is used to calculate the updated coefficient by using the least square method on the measured data in multiple data intervals in the measured database; the rolling mill model is updated according to the comparison result of the training set coefficient, the validation set coefficient and the updated coefficient; the prediction model module is used to predict the rolling force of the new billet according to the updated rolling mill model; the comparison module is used to obtain the rolling force of the billet which is the same as the measured data interval to obtain the measured rolling force; the difference between the measured rolling force and the rolling force of the new billet is compared; if the difference is less than or equal to 200 tons, the rolling force of the new billet is marked as the correct rolling force; if the difference is greater than 200 tons, the measured rolling force is marked as the correct rolling force.

[0093] For example, when a new slab arrives, find the same data interval of the measured data, calculate the measured rolling force, and then calculate the new slab rolling force through the updated rolling mill model. Compare the measured rolling force and the new slab rolling force. If the difference is less than or equal to 200 tons, use the new slab rolling force. If the difference is greater than 200 tons, use the measured rolling force. It can be understood that when generating the reduction table, the correct rolling force is generated. That is, when the new slab rolling force is marked, the new slab rolling force is used to generate the reduction table. When the measured rolling force is marked, the measured rolling force is used to generate the reduction table.

[0094] According to the above technical solution, the embodiment provides a method and system for optimizing rolling mill model to predict rolling force. The method comprises: constructing an initial database, the initial database including the collected data of the primary controller, the collected data being divided into training set data and validation set data; the types of the collected data including pass entry measured temperature, pass entry thickness, pass exit thickness, slab speed, slab width, target thickness, and target width; the training set data of each type being divided into multiple training set data intervals; the validation set data of each type being divided into multiple validation set data intervals, and the historical deformation resistance of each validation set data interval being calculated according to the type of the validation set data; the validation set coefficients of the multiple validation set data intervals being solved according to the historical deformation resistance; the training set coefficients of the multiple training set data intervals being calculated according to the initial database and the rolling mill model; the measured data of the current slab being saved according to the rolling mill model when the steel plate rolling is completed; a measured database being constructed according to the measured data, the measured database including the measured data, the types of the measured data including pass entry measured temperature, pass entry thickness, pass exit thickness, slab speed, slab width, target thickness, and target width, and the measured data of each type being divided into multiple data intervals; the update coefficients being calculated by using the least square method on the measured data in the multiple data intervals in the measured database; the rolling mill model being updated according to the comparison result of the training set coefficients, the validation set coefficients, and the update coefficients; the new slab rolling force being predicted according to the updated rolling mill model; the same slab rolling force as the measured data interval being obtained to obtain the measured rolling force; the difference between the measured rolling force and the new slab rolling force being compared; if the difference is less than or equal to 200 tons, the new slab rolling force is marked as the correct rolling force; if the difference is greater than 200 tons, the measured rolling force is marked as the correct rolling force. The present application can continuously learn the coefficients according to the current equipment working condition on site through long-term self-learning, so as to reduce the error between the predicted rolling force and the measured rolling force on site, and also reduce the error between the target thickness after rolling and the target thickness measured on site. Finally, the same slab rolling force and the predicted rolling force in the same measured data interval are compared, and the correct rolling force is marked by the difference, so that the prediction of the rolling force is more accurate.

[0095] The similar parts among the embodiments provided in the application can be referred to each other, the specific embodiments provided above are only several examples under the general concept of the application, and do not constitute the limitation of the protection scope of the application. Any other embodiments extended according to the application scheme without creative labor for those skilled in the art shall fall within the protection scope of the application.

Claims

1. A method for optimizing a rolling mill model to predict rolling force, characterized in that, The method includes: An initial database is constructed, which includes the acquisition data of the primary controller. The acquisition data is divided into training set data and validation set data. The types of acquisition data include measured temperature at the inlet of each pass, thickness at the inlet of each pass, thickness at the outlet of each pass, slab speed, slab width, target thickness, and target width. The training set data for each type is divided into multiple training set data intervals; Each type of validation set data is divided into multiple validation set data intervals, and the historical deformation resistance of each validation set data interval is calculated based on the type of validation set data. The validation set coefficients for multiple validation set data intervals are obtained by using the historical deformation resistance; the training set coefficients for multiple training set data intervals are calculated using the rolling mill model based on the initial database. When the steel plate rolling is completed, the measured data of the current steel billet is saved according to the mill model; a measured database is constructed based on the measured data, the measured database including measured data, the types of measured data including measured temperature at the inlet of the pass, measured thickness at the inlet of the pass, measured thickness at the outlet of the pass, slab speed, slab width, target thickness, and target width, and each type of measured data is divided into multiple data intervals; the update coefficient is calculated using the least squares method for the measured data in the multiple data intervals in the measured database; The training set coefficients, validation set coefficients, and update coefficients are compared, and the rolling mill model is updated based on the comparison results. Based on the updated rolling mill model, predict the new billet rolling force; Obtain the billet rolling force that is the same as the measured data range to obtain the measured rolling force; Compare the difference between the measured rolling force and the new billet rolling force; If the difference is less than or equal to 200 tons, the new billet rolling force is marked as the correct rolling force; If the difference is greater than 200 tons, the measured rolling force is marked as the correct rolling force.

2. The method for predicting rolling force using an optimized rolling mill model according to claim 1, characterized in that, The measured temperature data range for the pass inlet is 50℃, the pass inlet thickness data range is 2mm, the pass outlet thickness data range is 2mm, the slab width data range is 10cm, the slab thickness data range is 20mm, the target width data range is 10cm, and the target thickness data range is 2mm.

3. The method for predicting rolling force using an optimized rolling mill model according to claim 1, characterized in that, The method further includes: Calculate the strain rate and strain rate based on the measured database; The deformation resistance of the current steel billet is calculated based on the update coefficient, strain rate, and strain velocity.

4. The method for predicting rolling force using an optimized rolling mill model according to claim 3, characterized in that, In the steps of calculating strain rate and strain ratio based on the measured database, the strain rate is calculated according to the following formula: strain = inThk / outThk Where strain is the strain rate, inThk is the exit thickness of the previous pass, and outThk is the exit thickness of the current pass; The strain rate is calculated using the following formula: strainrate=strain*pce_Spd / arc Where strainrate is the strain rate, strain is the strain ratio, pce_Spd is the linear velocity of the steel plate passing through the mill rolls, and arc is the length of the contact arc between the mill work rolls and the steel plate.

5. The method for predicting rolling force using an optimized rolling mill model according to claim 3, characterized in that, In the step of calculating the deformation resistance of the current billet based on the update coefficient, strain rate, and strain velocity, the deformation resistance is calculated according to the following formula: fs=C1e C2 / RT strain C3 strainrate C4 Where fs is the deformation resistance, e is the natural base, R is the physical constant relating the thermodynamic functions in the equation of state, T is the measured inlet temperature of the pass, strain is the strain rate, strain rate is the strain speed, C1 is the hardness of the steel plate, C2 is the activation energy of the reaction, C3 is the strain index, and C4 is the strain rate index.

6. The method for predicting rolling force using an optimized rolling mill model according to claim 5, characterized in that, The method further includes: The measured inlet temperature, inlet thickness, outlet thickness, slab speed, slab width, target thickness, and target width of one of the training set data intervals are put into the solution matrix, and the C1 steel plate hardness, C2 reaction activation energy, C3 strain index, and C4 strain rate index are calculated using the least squares method and the solution matrix. The solution matrix is: lnfs=lnC1+C2(1 / RT)+C3ln(strain)+C4ln(strainrate) Here, lnC1 is treated as a coefficient to be determined.

7. The method for predicting rolling force using an optimized rolling mill model according to claim 1, characterized in that, The step of comparing the training set coefficients, validation set coefficients, and update coefficients, and updating the rolling mill model based on the comparison results, includes: If the training set coefficients are the same as the update coefficients, or if the error between the training set coefficients and the update coefficients is within the error range, then the update coefficients are used to replace the training set coefficients. If the error between the training set coefficients and the updated coefficients is not within the error range, then the training set coefficients are compared with the validation set coefficients. If the training set coefficients are the same as the validation set coefficients, or if the error between the training set coefficients and the validation set coefficients is within the error range, then the validation set coefficients are used to replace the training set coefficients. If the error between the training set coefficients and the validation set coefficients is not within the error range, the training set coefficients in the initial database are maintained, and the validation set coefficients are discarded.

8. The method for predicting rolling force using an optimized rolling mill model according to claim 1, characterized in that, In the step of predicting the new billet rolling force based on the updated rolling mill model, the billet rolling force is calculated according to the following formula: Rolling force = fs * arc * Factr * with Where arc is the contact arc length between the work roll and the workpiece, Factr is the geometric friction condition of the viscous rolling bite equation, and with is the current width of the billet.

9. The method for predicting rolling force using an optimized rolling mill model according to claim 1, characterized in that, The step of predicting the new billet rolling force according to the updated rolling mill model further includes: calculating the reduction table for each pass during the rolling process, the reduction table including the steel plate thickness, steel plate length, steel plate width, steel plate temperature, rolling mill force, linear velocity of the steel plate passing through the rolling mill rolls, and reduction amount.

10. A system for optimizing a rolling mill model to predict rolling force, characterized in that, include: The module includes a data acquisition module, a storage module, a prediction and training module, a model module, and a comparison module. The acquisition module is used to acquire data from the primary controller; The storage module is used to save the measured data of the current steel billet according to the rolling mill model when the steel plate rolling is completed; The prediction training module is used to calculate the update coefficients using the least squares method on the measured data in multiple data intervals in the measured database; compare the training set coefficients, validation set coefficients and update coefficients, and update the rolling mill model according to the comparison results; The model module is used to predict new billet rolling forces based on the updated rolling mill model; The comparison module is used to obtain the billet rolling force that is the same as the measured data range to obtain the measured rolling force; compare the difference between the measured rolling force and the new billet rolling force; if the difference is less than or equal to 200 tons, mark the new billet rolling force as the correct rolling force; if the difference is greater than 200 tons, mark the measured rolling force as the correct rolling force.

Images