Material property prediction device for rolled product
Through the approximate model, the three-dimensional grid-like material characteristics of rolled products are predicted, and the problems of high-precision model calculation time and insufficient empirical rule model accuracy in the existing technology are solved, and high-speed and high-precision material characteristics prediction and yield improvement are achieved.
Patent Information
- Application Number
- CN202380028593.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-03-14
- Publication Date
- 2024-11-15
AI Technical Summary
In the prediction of material characteristics of rolled products, the high-precision model has a large calculation load and takes a long time, and the approximate accuracy of the empirical rule model is limited, which cannot meet the needs of high-speed and high-precision online prediction.
The approximate model production department is used to create an approximate model offline, and the data set and model parameters are generated through the data set production department and the model parameter determination department. The material characteristic prediction department uses the approximate model to calculate the three-dimensional grid-like material characteristics of the rolled product online, and passes the material The characteristic correction unit corrects the calculation result.
The material characteristics of rolled products with high speed and high precision are predicted online, which reduces the calculation load and reflects the results in the next batch of rolled products, improving yield and forming quality.
Smart Images

Figure CN118973725A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to a device for predicting material properties of a rolled product, and more particularly, to predicting material properties of a rolled product manufactured in a hot rolling process. Background Art
[0002] The material properties of rolled products (hereinafter also referred to as "products") made of metal materials such as steel vary according to their alloy composition, heating conditions of the hot rolling process, processing conditions and cooling conditions. Material properties are, for example, mechanical properties (strength, formability, toughness, etc.), electromagnetic properties (magnetic permeability, etc.), etc. The alloy composition is adjusted by controlling the addition amount of component elements. When adjusting the composition, for example, a composition adjustment furnace capable of holding about 100 tons of molten steel is used, and a batch unit is relatively large. Therefore, it is impossible to change the addition amount for each rolled product of about 15 tons. Therefore, in order to manufacture hot-rolled product coils of the desired material, it is important to properly control the heating conditions, processing conditions and cooling conditions.
[0003] In the hot rolling process, the target values of various process parameters, which are process conditions related to product quality and operating conditions, are changed to produce each rolled product. Examples of the process parameters include the target temperature of each point on the rolling line, such as the entry and exit temperatures of the finishing mill, the coiling temperature, the drive rod thickness at the exit of the roughing mill, the plate thickness such as the reduction rate at each pass, the schedule related to the reduction of the plate width, the use of the descaling machine at each pass of the finishing mill and the roughing mill, the use of the inter-stand cooling arranged between the stands of the roughing mill and the finishing mill, and the initial flow rate of the use, the amount of lubricating oil used in the finishing mill, and the cooling mode used in the run-out roller table.
[0004] In the past, the following method was generally adopted: for process parameters related to heating, processing and cooling, the target value of heating temperature, the target value of size after processing, the target value of cooling rate, etc. were set for each specification of rolled products, and temperature control and size control were performed to achieve these target values. In addition, the target value of product size was specified in advance, while the target value of plate thickness, target value of temperature and target value of cooling rate at the exit of each stand was determined based on years of experience. However, in recent years, the requirements for product specifications have become more sophisticated and diversified, and sometimes these target values may not be properly determined by empirical methods.
[0005] Patent Document 1 below discloses an apparatus for simulating the manufacturing process offline using a process model that models each manufacturing process of heating, processing, and cooling in order to study in advance whether a product manufactured under a certain alloy composition and process parameters will have the desired product quality.
[0006] Furthermore, there is a need to manage material properties more strictly than the management within the previous guarantee range. In the past, as stipulated in JIS (Japanese Industrial Standards), the material properties exceeding the reference value were used as the condition (permissible range). For example, a tensile test was performed using a sample taken out of the product to determine whether the measured value exceeded the reference value. However, in recent years, high precision has also been required in the processes after the product leaves the factory. In the above-mentioned previous permissible range, for example, it is sometimes not sufficient in the forming process (drawing, bending, stamping, etc.) as the next process. Sometimes the rolled material is too hard to be formed, the rebound amount (elastic recovery amount) after stamping is too large and the shape is lacking in freezing, and the edge is broken during forming. Therefore, in the experience-based setting method and the material property management method, there is a problem that the above-mentioned target values may not be properly controlled.
[0007] As a conventional method for managing the process of the rolling process, there is a method of using the output value of a thermometer disposed in the rolling production line to manage the temperature of the entire rolled coil, and further to manage the material properties that are closely related to the rolling temperature. Specifically, thermometers are disposed on the exit side of the heating furnace of the rolling production line, the entry and exit sides of the roughing mill, the entry and exit sides of the finishing mill, and the entry side of the coiler. The thermometer measures the temperature of the central portion of the rolled material in the plate width direction (hereinafter also referred to as the "width direction"). Then, control is performed so that the output value from the thermometer is consistent with the target temperature determined based on experience by the host computer. In this way, in the past, when managing the process of the rolling process, the material properties of the rolled material in the width direction were not considered.
[0008] The width end (edge) of the rolled material is easily cold cut, resulting in a temperature difference with the center. Sometimes, a rolling line is equipped with a device for raising the temperature of the width end of the rolled material and a device for preventing the temperature of the width end of the rolled material from decreasing. For example, during the cooling process after rolling, an edge mask is used to prevent cooling water from splashing onto the edge. In addition, an induction heating device such as an edge heater is used to heat the edge before finishing rolling.
[0009] Furthermore, in recent years, in order to verify the effect of the above-mentioned device, scanning pyrometers are sometimes installed before and after the above-mentioned device. If a scanning pyrometer is used, the temperature distribution in the width direction of the rolled material can be measured. In addition, in recent years, in the multi-purpose detector used in the rolling production line, a scanning pyrometer is also used to use the temperature distribution in the width direction of the rolled material for correction of the measured value. The multi-purpose detector is a composite measuring instrument that measures the plate thickness, convexity, plate width, etc. with one device, and its measurement accuracy has been significantly improved in recent years.
[0010] Thus, a device for measuring the temperature distribution in the width direction of the rolled material is introduced into the rolling production line. In addition, some attempts have been made to calculate the temperature distribution in the width direction of the rolled material and to flexibly use it in control. The following patent document 2 discloses a method for calculating the temperature distribution in the thickness direction and the width direction of the rolled material. In addition, the following patent document 3 discloses a method for uniformizing the temperature in the width direction by controlling the edge heater based on the calculation of the temperature distribution in the width direction.
[0011] In addition, similar to the plate thickness, plate width, shape, etc., for the material, when an unreachable part is generated, it is generally required to cut off the unreachable part while comparing it with the actual data in the dividing line of the next process, taking into account the margin (margin). Although the amount of cutting off is directly related to the yield rate, in the past, the amount of cutting off was only roughly determined. In view of such a situation, even if a material unreachable part is generated, it is expected to optimize the amount of cutting off from the perspective of improving the yield rate, and further expect to minimize the material unreachable part through process improvement.
[0012] Therefore, conventionally, attempts have been made to control properties by using a property prediction model to predict the properties of a rolled product using manufacturing conditions such as heating, processing, and cooling in the rolling process as input.
[0013] Patent document 4 discloses a method for digitizing a model for predicting changes in the microstructure of a rolled material and the mechanical properties of the final product using actual values of mechanical properties obtained from mechanical property measurement test results such as tensile tests and structural observations on a portion of a product coil. Patent document 5 discloses a method for outputting a material distribution that associates positions in two-dimensional directions in the length and width directions with material property values. Patent document 6 discloses a method for pre-storing operating conditions and actual materials, searching for similar operating conditions, and estimating the materials of all positions of a product coil online in a grid shape. In addition, Patent document 7 discloses a method for predicting materials using a neural network.
[0014] Prior art literature
[0015] Patent Literature
[0016] Patent Document 1: Japanese Patent No. 6292309
[0017] Patent Document 2: Japanese Patent No. 6197676
[0018] Patent Document 3: Japanese Patent No. 6447710
[0019] Patent Document 4: Japanese Patent No. 5396889
[0020] Patent Document 5: Japanese Patent Application Publication No. 2022-48037
[0021] Patent Document 6: Japanese Patent No. 6086155
[0022] Patent Document 7: Japanese Patent Application Publication No. 2005-315703 Summary of the invention
[0023] Problems to be solved by the invention
[0024] However, for the prediction of material properties based on the model that digitizes metallurgical phenomena shown in Patent Documents 4 and 5, the more faithfully the microscopic phenomena are modeled, the higher the accuracy of the model, the greater the calculation load. For example, even if the calculation points of the product coil are a few points, it sometimes takes several seconds of calculation time. For example, if you want to predict a 1km product coil with a grid width of 2m spacing in the rolling direction, 3 points in the plate thickness direction, and 5 points in the plate width direction, the calculation points are 7500 points. Assuming that the calculation amount of each calculation point is set to 1 second, it takes more than two hours of calculation time. Rolling operations are mostly continuous rolling of the same or similar product partitions. In the case of wanting to reflect the results of the material to the next material or the operation change of the material in the same batch, sometimes rolling is performed at intervals of about 2 to 5 minutes per coil, which takes too much time.
[0025] In addition, in Patent Documents 6 and 7, the relationship between past operating conditions and actual material is searched and modeled, and the material of the newly rolled product coil is predicted based on empirical rules. It is known that such empirical rule models are generally high-speed and help solve the problem of calculation load. However, in the model based on past operating conditions and actual material, the approximate accuracy when modeling complex material behavior is limited, and sufficient prediction accuracy cannot be expected.
[0026] The present disclosure is made to solve the above-mentioned problems. An object of the present disclosure is to provide a material property prediction device for rolled products that can predict the material properties of an entire group of rolled products online at high speed and with high accuracy.
[0027] Means for solving problems
[0028] The first aspect relates to a material property prediction device for rolled products, which predicts the material properties of rolled products manufactured in a rolling line. The material property prediction device for rolled products comprises: an approximate model making unit, which makes an approximate model offline, and the approximate model comprehensively predicts the material properties of the group of rolled products manufactured in the rolling line; and a material property prediction unit, which uses the approximate model made by the approximate model making unit to predict the material properties of each region of the three-dimensional grid of the rolled products manufactured in the rolling line online. The approximate model making unit comprises: a data set making unit, which has a condition setting unit for setting the rolling conditions of the group of rolled products, and a material calculation unit for calculating the metallurgical phenomena and material properties under the rolling conditions, and the data set making unit makes a data set for making the approximate model; and a model parameter determination unit, which uses the data set to determine the parameters that express the approximate model.
[0029] The second aspect is characterized in that, in addition to the first aspect, the data set creating unit creates the data set using the rolling conditions set by the condition setting unit as explanatory variables and the material properties calculated by the material calculation unit as target variables.
[0030] The third viewpoint has the following features on the basis of the first viewpoint. The material property prediction unit comprises: a rolling data collection unit, which collects online rolling data obtained when the rolled product is manufactured in the rolling production line; a model input preparation unit, which prepares online input data for the approximate model based on the rolling data collected by the rolling data collection unit; an approximate model calculation unit, which calculates online the material properties of each region of the three-dimensional grid of the product coil by inputting the input data prepared by the model input preparation unit into the approximate model; and a material property output unit, which outputs the material properties of each region calculated by the approximate model calculation unit, information indicating the position of each region in the rolled product, and information related to the material properties.
[0031] The fourth aspect has the following characteristics in addition to the first aspect: The approximate model is a machine learning model.
[0032] The fifth aspect is characterized in that, in addition to any one of the first to fourth aspects, the material property prediction unit includes a material property correction unit that corrects the material property calculated by the approximate model calculation unit using the approximate model, using a material property result calculated using a metallurgical phenomenon model that digitizes the metallurgical phenomenon.
[0033] Effects of the Invention
[0034] According to the present disclosure, a data set corresponding to rolling conditions (temperature conditions, processing conditions, time and speed conditions) that are not implemented in the actual rolling process and rolling conditions with little implementation experience is also pre-made, and an approximate model is made offline using the made data set, thereby making an approximate model that can correspond to the entire group of rolled products and has high approximate accuracy. By using the approximate model made in this way, the calculation load for online prediction of the material properties of each area of the three-dimensional grid of the rolled product can be reduced. Moreover, since the material properties of all areas (parts) of the rolled product can be calculated at high speed, the result of the rolled product can also be reflected in the operation changes of the next rolled product or the rolled products in the same batch. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] Figure 1 This is a diagram showing an example of a thin plate hot rolling line to which the material property prediction device of the rolled product according to the first embodiment is applied.
[0036] Figure 2 This is a block diagram showing the rolling system according to the first embodiment.
[0037] Figure 3 This is a block diagram showing the functions of the material property prediction device for a rolled product according to the first embodiment.
[0038] Figure 4 This is a diagram showing an example of the hardware configuration of a device for predicting material properties of rolled products.
[0039] Figure 5 This is a block diagram showing the structure of an approximate model creation unit.
[0040] Figure 6 This is a diagram for explaining a data set created by the data set creation unit.
[0041] Figure 7 This is a schematic diagram for explaining an example of a cooling pattern of explanation variables.
[0042] Figure 8 This is a diagram for explaining the configuration of a data set creation unit and creation of a data set.
[0043] Fig. 9 This is a diagram for explaining the setting of the material calculation unit and the explanation variables by the condition setting unit.
[0044] Fig.10 This is a diagram for explaining the calculation of metallurgical phenomena and the setting of target variables performed by the material calculation unit.
[0045] Fig.11 This is a schematic diagram showing an example of an approximate model.
[0046] Fig.12 This is a block diagram showing the structure of a material property prediction unit.
[0047] Fig.13 This is a schematic diagram showing an example of a mesh region for which the approximate model calculation unit calculates material properties.
[0048] Fig.14 This is a schematic diagram showing an example of the output of the material property output unit.
[0049] Fig.15 This is a block diagram showing the structure of a material property prediction unit according to the second embodiment.
[0050] Fig.16 This is a schematic diagram showing a correction method of the material property correction unit according to the second embodiment. DETAILED DESCRIPTION
[0051] Hereinafter, with reference to the drawings, an embodiment of the present invention will be described in detail by taking the case of predicting the material properties of a rolled product produced in a hot rolling line as an example. In addition, the same reference numerals are given to common elements in each figure, and repeated descriptions are omitted.
[0052] Implementation method 1.
[0053] [Rolling production line]
[0054] Figure 1 1 is a diagram showing an example of a hot-rolling line for thin plates (hereinafter also referred to as a "rolling line") to which the material property prediction device for rolled products of Embodiment 1 (hereinafter also referred to as a "prediction device") is applied. Figure 1 Although the prediction device for material properties of rolled products manufactured on the rolling line shown in the figure is described, the prediction device disclosed in the present invention can also be applied to other rolling lines.
[0055] The rolling production line is equipped with a heating device, a rolling mill, a cooling device, a down coiler, and a conveyor roller table connecting them. These devices are driven by actuators such as electric motors or hydraulic devices. Specifically, Figure 1 The rolling production line 1 shown includes, from the upstream side of the conveyor roller, a heating furnace 2, a high-pressure descaling device 3, a rough rolling inlet thermometer 4, a rough rolling edge mill 5, a rough horizontal rolling mill (hereinafter also referred to as a "rough rolling mill") 6, a rough rolling outlet thermometer 7, an edge heater 8, a rod heater 9, a finishing rolling inlet thermometer 10, a cutting head machine 11, a finishing rolling inlet descaling device 12, an F1 rolling edge mill 13, a finishing mill 14, a multi-purpose detector 15, a finishing rolling outlet thermometer 16, an output roller 17, a coiler inlet thermometer 18 and a down coiler 19.
[0056] The heating furnace 2 is a furnace for heating the rolled material (slab), and is controlled to obtain the desired slab heating pattern and the heating furnace extraction temperature. In the following description, the rolled material includes not only slabs and steel plates, but also the intermediate state before the product coil is completed. The high-pressure descaling device 3 sprays high-pressure water from the top and bottom to the rolled material coming out of the heating furnace 2 to remove the scale from the surface of the rolled material. The rough rolling inlet side thermometer 4 is arranged on the inlet side (upstream side) of the rough rolling mill 6 to measure the temperature of the surface (for example, the upper surface) of the center part of the width direction of the rolled material, that is, the rough rolling inlet side temperature. The rough rolling mill 5 performs rolling of the rolled material in the plate width direction. The rough rolling mill 6 performs rough rolling of the steel plate in the plate thickness direction. In the rough rolling mill 6, rolling is performed in multiple passes to make the rolled material become the desired thickness. Therefore, as the rough rolling mill 6, a reversible rolling mill can be used. The rough rolling outlet side thermometer 7 measures the temperature of the surface (for example, the upper surface) of the rolled material. The roughing exit thermometer 5 is disposed at the exit side (downstream side) of the roughing mill 6. When the rolled material passes through the roughing mill 6, the roughing exit thermometer 7 measures the surface temperature of the center portion in the width direction as the roughing exit temperature.
[0057] The edge heater 8 is a device that heats up the end (edge) of the rolled material in the width direction by electromagnetic induction heating or the like in order to control the temperature of the rolled material. The rod heater 9 is a device that heats up the entire rolled material by electromagnetic induction heating or the like in order to control the temperature of the rolled material. The finishing rolling entry thermometer 10 is arranged on the entry side of the finishing mill 14 to measure the temperature of the surface (for example, the upper surface) of the center portion in the width direction of the rolled material, i.e., the finishing rolling entry temperature. The head trimmer 11 cuts off the front end and the rear end of the steel plate. The finishing rolling entry descaling device 12 removes oxide scale from the surface of the steel plate on the entry side of the finishing mill 14. The F1 edge rolling mill 13 is arranged on the entry side of the finishing mill 14, and its rollers contact the rolled material from the side. The F1 edge rolling mill 13 deforms the rolled material within a range that does not buckle, so as to narrow the width of the rolled material. The finishing mill 14 is composed of a single or multiple frames, and the frames are arranged at the entry side. Figure 1 In the example shown, a tandem finishing mill is formed of seven stands. The finishing mill 14 finish-rolls the rolled material to a predetermined thickness.
[0058] The multi-purpose detector 15 is a composite measuring instrument that can perform various measurements with one device. The multi-purpose detector 15 has, for example, a form in which a plurality of X-ray detectors are arranged in the width direction of the rolled material. The multi-purpose detector 15 measures, for example, the plate thickness distribution in the width direction of the rolled material. By preparing one multi-purpose detector 15, the plate thickness, convexity, and plate width of the rolled material can be measured. In recent years, the measurement accuracy of the multi-purpose detector 15 has been greatly improved. Therefore, it is cheaper to prepare one multi-purpose detector 15 than to prepare a plate thickness meter, a convexity meter, and a plate width meter separately, and the introduction of the multi-purpose detector 15 to the hot rolling production line is being promoted. The multi-purpose detector 15 is equipped with a thermometer and a scanning pyrometer inside. The multi-purpose detector 15 measures the temperature of the rolled material, and its measured value is used to calibrate the detection value of the X-ray detector.
[0059] The finishing exit thermometer 16 measures the temperature of the surface (for example, the upper surface) of the rolled material. The finishing exit thermometer 16 is arranged on the exit side (downstream side) of the finishing mill 14. The finishing exit thermometer 16 measures the surface temperature of the central part of the width direction of the rolled material after passing through the finishing mill 14 as the finishing exit temperature. The finishing exit thermometer 12 is arranged on the exit side of the finishing mill 10. The finishing exit temperature of the rolled material is closely related to the formation of the metal structure and the material quality (tensile strength, yield stress, elongation, etc.) of the product. Therefore, it is necessary to properly manage the finishing exit temperature of the rolled material.
[0060] The output roller table 17 is a cooling device that cools the rolled material by cooling water in order to control the temperature of the rolled product. In order to control the temperature of the rolled material, the output roller table 17 supplies cooling water to the surface of the rolled material from, for example, a nozzle. The output roller table 17 is equipped with a plurality of nozzles in the length direction of the rolled material (the conveying direction of the conveyor roller table). These nozzles are divided into a plurality of groups. The nozzles are controlled for each group to control the cooling rate of the rolled material. Water cooling is performed in the group supplied with cooling water, and air cooling is performed in the group not supplied with cooling water. In addition, the rolling production line may also be equipped with cooling roller tables, forced cooling devices, etc. as cooling devices.
[0061] The coiler entry thermometer 18 is arranged at the entry side (upstream side) of the down coiler 19. After the rolled material passes through the output roller 17, the surface temperature of the central part in the width direction is measured by the coiler entry thermometer 18 as the coiling temperature. The finishing rolling exit thermometer 12 is arranged at the exit side of the finishing mill 10. The coiling temperature of the rolled material is closely related to the formation of the metal structure and the material quality (tensile strength, yield stress, elongation, etc.) of the product. Therefore, it is necessary to properly manage the coiling temperature of the rolled material.
[0062] The down coiler 19 is a device for coiling the rolled product to form a shape that is easy to transport. The conveyor roller is a device for conveying the rolled product in each process to the next process. These devices are driven by actuators such as electric motors or hydraulic devices.
[0063] Figure 1 The rolling line 1 shown is further provided with a scanning pyrometer 20. The scanning pyrometer 20 measures the temperature of the surface of the rolled material (e.g., the upper surface or the upper surface and the lower surface) at least at a plurality of locations in the width direction of the rolled material. The scanning pyrometer 20 is preferably arranged before and after the device for improving the temperature of the rolled material. Figure 1 In the example shown, the scanning pyrometers 20 are provided in front of the edge heater 8, behind the rod heater 9, and before and after the output roller table 17. In addition, the scanning pyrometer 20 disposed on the inlet side of the output roller table 17 is provided inside the multi-purpose detector 15.
[0064] [Rolling system]
[0065] Figure 2 : This is a block diagram of a rolling system 21 of Embodiment 1. The rolling system 21 is a control system of the rolling production line 1, and has a hierarchical structure of levels 0 to 3. Level 0 has a drive control device for controlling the motors of each device driving the rolling production line 1, and a hydraulic device (hydraulic device) for driving each device of the rolling production line 1. Level 1 has a control controller 24. Level 2 has a setting computer 23. Level 3 has a host computer 22 for production management. The material property prediction device 25 of the rolled product described later is connected to the setting computer 23 and can receive rolling data.
[0066] In the hot rolling process, the process conditions related to the product quality and the operating conditions, that is, the target values of various process parameters, are changed to produce the products respectively. In order to achieve the target product quality, that is, to achieve the target values of the above-mentioned various process parameters, the process control is performed by setting the computer 23.
[0067] The target value of the process parameter may be specified from the upper computer 22 of level 3 located above the setting computer 23 of level 2. In addition, the target value of the process parameter may be specified by using the steel type, plate thickness, plate width, etc. as keywords in a table in a database belonging to the setting computer 23. In addition, the target value of the process parameter may be changed during rolling by manual intervention of an operator.
[0068] The setting computer 23 has a model formula that expresses the physical phenomena of each process such as heating, rolling, cooling, and conveying in the rolling line 1. The setting computer 23 uses the model formula that expresses the physical phenomena of the process to perform setting calculations so as to achieve the target values (process conditions) of the above-mentioned various process parameters in actual operation. In the setting calculations, the calculation of the control target values of various actuators and the calculation of the state of the rolled material (the predicted value of the state of the metal material) at each stage of the process are repeatedly performed.
[0069] The control target values of the actuators are the roll gap of the rolling mills 6 and 14, the rolling speed, the conveying speed, the descaling machine, the flow rate of various sprayers, the opening / closing of the valve of the output roller table, etc. The state of the rolled material at each stage of the process (the predicted state value of the metal material) is the size, shape, temperature, microstructure, etc.
[0070] The control controller 24 receives the setting calculation results from the setting computer 23 and controls various actuators to follow the control target value. In the actual hot rolling process, various sensors are set at various locations of the rolling production line 1 to monitor and collect the actual values of parameters that affect process control, such as temperature, shape, plate thickness, plate width, and rolling load.
[0071] These actual values are used for process control, model accuracy improvement, and quality management. The target value of the process parameter is compared with the actual value obtained by various sensors or the actual calculated value recalculated by the setting computer 23 based on the actual value and the calculated value. If the target value of the process parameter is not achieved, the setting computer 23 performs the setting calculation again. Based on the results, various controls such as feedforward control, feedback control, and dynamic control are performed.
[0072] Even when the model formula of the process simulates the physical phenomenon with high accuracy, model prediction errors will actually occur. Therefore, engineers fine-tune the coefficients and constants involved in each item in the model formula to improve the prediction accuracy of the model formula. The adjustment items are the coefficients and constants of each item in the model formula. They are managed in the database belonging to the setting computer 23 for each layer using a layered table divided by factors that are prone to model errors, such as steel type, target plate thickness, target plate width, target temperature, etc. In addition to the start of the operation, the adjustment items are mainly adjusted when rolling a new type of steel and rolling with a new combination of process parameters. The adjustment items are sometimes adjusted by engineers based on experience or numerical analysis results. In recent years, statistical methods such as neural networks are sometimes used for semi-automatic adjustments. Learning items are items that perform multiplication and addition operations on the model formula in order to compensate for the error between the model output and the output of the actual process.
[0073] [Material property prediction device for rolled products]
[0074] Figure 3 2 is a block diagram showing the functions of the material property prediction device 25 of the rolled product according to the first embodiment. Figure 1 The mechanical properties of the product coil rolled in the rolling line 1 are shown. The prediction device 25 includes an approximate model creation unit 26 and a material property prediction unit 27.
[0075] Figure 4 2 is a diagram showing an example of the hardware configuration of the prediction device 25. The functions of the prediction device 25 described below can be performed by Figure 4 The processing circuit 250 shown in the figure is implemented. The processing circuit 250 can also be a dedicated hardware 251. The processing circuit can also include a processor 252 and a memory 253. The processing circuit can also be partially formed as a dedicated hardware 251 and also include a processor 252 and a memory 253. Figure 4 In the example of FIG. 2 , a part of the processing circuit 250 is formed as dedicated hardware 251 , and the processing circuit 250 further includes a processor 252 and a memory 253 .
[0076] The processing circuit 250 may also be at least one dedicated hardware 251. In this case, the processing circuit 250 corresponds to, for example, a single circuit, a complex circuit, a programmed processor, a parallel programmed processor, an ASIC, an FPGA, or a processing circuit formed by combining them.
[0077] The processing circuit 250 may also include at least one processor 252 and at least one memory 253. In this case, each function of the prediction device 25 is implemented by software, firmware, or a combination of software and firmware. Software and firmware are described as programs and stored in the memory 253. The processor 252 implements the functions of the approximate model making unit 26 and the material property prediction unit 27 by reading and executing the program stored in the memory 253. The processor 252 is also called a CPU (Central Processing Unit), a central processing unit, a processing unit, an arithmetic unit, a microprocessor, a microcomputer, or a DSP. The memory 253 is, for example, a non-volatile or volatile semiconductor memory such as a RAM, a ROM, a flash memory, an EPROM, or an EEPROM. In this way, the processing circuit 250 can implement each function of the prediction device 25 by hardware, software, firmware, or a combination thereof.
[0078] The approximate model creation unit 26 creates an approximate model 39 that comprehensively and quickly predicts the mechanical properties of the hot rolled product group manufactured in the rolling line 1, that is, the rolled products (product coils) of all steel types, sizes, and operating conditions that can be manufactured in the rolling line 1. Comprehensiveness means that the mechanical properties of not only the rolled products manufactured in the rolling line 1 but also the rolled products that may be manufactured in the rolling line 1 in the future are included. The details of the approximate model creation unit 26 are described below.
[0079] Figure 5 1 is a block diagram showing the configuration of the approximate model creation unit 26. The approximate model creation unit 26 includes a data set creation unit 28 and a model parameter determination unit 29. The data set creation unit 28 creates a data set for creating an approximate model 39. The model parameter determination unit 29 uses the data set created by the data set creation unit 28 to determine parameters that express the approximate model 39 by an optimization method or the like.
[0080] Figure 636 is a diagram for explaining a data set created by the data set creation unit 28. The data set 36 created by the data set creation unit 28 is a collection of data in which the data of the explanatory variable 37 and the data of the target variable 38 are paired (form a pair). Each pair is composed of the data of the explanatory variable 37 corresponding to a part of a certain operating condition and the data of the target variable 38 corresponding to the mechanical properties of the coil produced by the rolling line 1 under the operating condition. There are many operating conditions, but as the explanatory variable 37, an operating condition related to the mechanical properties (hereinafter referred to as the "first operating condition") is particularly selected, such as chemical composition, temperature condition, processing condition, time condition, speed condition, etc. In order to produce different steel grades, the content of chemical components in steel is adjusted. Chemical components such as C, Mn, Si, Nb, N, and Ti that contribute to mechanical properties are included in the explanatory variable 37. Similar to the chemical composition, the temperature condition is adjusted in order to produce different steel grades, and the contribution to mechanical properties is large. The temperature history of a certain part of the product coil is time series data, but when it is given as the explanatory variable 37, one-dimensional data is given by cutting out the time series data at an important point moment. For example, one-dimensional data such as heating furnace extraction temperature = 1200°C, rough rolling inlet temperature = 1100°C, rough rolling outlet temperature = 1000°C, finishing rolling inlet temperature 950°C, finishing rolling outlet temperature = 900°C, and coiling temperature = 650°C are provided to each explanatory variable 37 to express the main temperature history affecting the mechanical properties. The temperature conditions affecting the mechanical properties are selected as the explanatory variables 37 from the temperatures that can be actually measured by a thermometer or a scanning pyrometer during the rolling of the product online and the temperatures actually recalculated by the setting computer 23. Temperature conditions that may vary for each product such as the heating furnace extraction temperature and the rough rolling inlet temperature are used as the explanatory variables 37. In addition, temperature conditions that differ in the rolling direction and the width direction such as the finishing rolling outlet temperature and the coiling temperature, and temperature conditions that cause a temperature drop at the width end such as the rough rolling inlet temperature and the rough rolling outlet temperature are used as the explanatory variables 37. Furthermore, the mechanical properties are formed by adjusting the temperature path according to the cooling conditions of the run-out roller table 17. Specifically, in addition to the target temperatures of the finishing rolling outlet temperature and the coiling temperature, part of the cooling mode conditions such as the water cooling part position, water cooling speed, water cooling time, and air cooling time are set and controlled (normal cooling). Furthermore, the output roller 17 is divided into the front half and the rear half, and an intermediate thermometer is set at its midpoint. Sometimes, in addition to the above-mentioned target temperature, the target temperature of the intermediate thermometer, the front half and the rear half of the output roller 17 are also set and controlled respectively. Cooling mode conditions (distributed cooling). Therefore, the cooling mode conditions described later are also added to the explanatory variables. The processing conditions are the operating conditions such as the reduction rate and strain rate of each stand. It is known that the reduction rate and strain rate have an impact on the microstructure, and as a result, the final mechanical properties are also affected. It is added to the explanatory variables as the processing conditions of each stand.Alternatively, summary information such as the average of the entire finishing rolling and the average of the half stand after finishing rolling can also be added to the explanatory variables. Regarding time conditions, by adding the slab furnace time in the heating furnace, that is, the heating furnace furnace time, the time from rough rolling to the start of finishing rolling, that is, the rough and finishing time, etc., which are conditions that have deviations for each product and affect the mechanical properties, the prediction performance of the deviation of the mechanical properties of each product is sometimes improved. In addition, by also including the speed conditions such as the rolling speed and conveying speed at various places that change from the front end to the tail end of the rolling direction and affect the mechanical properties in the explanatory variables, the prediction performance of the deviation of the mechanical properties in the product is sometimes improved. The mechanical properties that become the target variables are yield stress, tensile strength, elongation, etc. Data set 36 includes all steel types and sizes of coils manufactured in the rolling production line 1, including the operating conditions of all places such as the front end, tail end, and end in the width direction, and the mechanical properties when rolled under the operating conditions.
[0081] Figure 7 37 is a schematic diagram for explaining an example of a cooling mode of the explanatory variable 37. Figure 7 An example of a cooling mode for normal cooling is shown in FIG. Figure 7 In the example, the upstream group of the run-out roller 17 is used for the initial cooling setting based on the setting calculation and the feedforward control using the actual value of the finishing temperature. In addition, the two groups at the most downstream are used for the feedback control using the actual value of the coiling temperature, and the group upstream of the group used for the feedback control is used for the dynamic control following the change of the conveying speed. Figure 7 For such a cooling mode, for example, the water cooling unit cooling rate, the water cooling unit cooling rate of the first half, the water cooling unit cooling rate of the second half, the average cooling rate of the first half, the average cooling rate of the second half, and the water cooling unit cooling rate of the feedback group are used as explanatory variables. cool The temperature drop caused by water cooling in the groups other than the dynamic control group and the feedback control group was calculated by the following formula (1).
[0082]
Mathematical formula 1
[0083]
[0084] Here, t W_s is the start time of water cooling except for the dynamic control group and the feedback control group, t W_e is the end time of water cooling except for the dynamic control group and the feedback control group, T W_s is the temperature at the beginning of water cooling, T W_e It is the temperature at the end of water cooling.
[0085] It is known that in the run-out roller table 17, the effects on mechanical properties are different when the front half is cooled and when the rear half is cooled. The cooling rate of the water cooling section in the front half, the cooling rate of the water cooling section in the rear half, the average cooling rate in the front half, the average cooling rate in the rear half, etc. are calculated as explanatory variables. For example, the cooling rate V of the water cooling section in the front half is E_cool The temperature drop caused by water cooling in the group from the finish rolling exit temperature to the intermediate temperature was calculated using the following formula (2).
[0086]
Mathematical formula 2
[0087]
[0088] Here, t E_s is the start time of water cooling for the first half of the group, t E_e It is the end time of the first half of the water cooling, T E_s is the temperature at the beginning of the first half of the water cooling, T E_e It is the temperature at the end of the first half of the water cooling.
[0089] In addition, the cooling rate of the water cooling unit in the second half is also calculated using the time and temperature information of the second half, and is thus included in the explanatory variables. For example, the average cooling rate V in the first half E_ave The temperature drop in the water cooling and air cooling in the group from the finish rolling exit temperature to the intermediate temperature was calculated using the following formula (3).
[0090]
Mathematical formula 3
[0091]
[0092] Here, t FDT is the time when the temperature of the finishing rolling outlet passes, t MT is the intermediate temperature passing time, T FDT is the finishing rolling outlet temperature, T MT is the intermediate temperature.
[0093] In addition, the average cooling rate of the second half is also calculated using the timing and temperature information of the second half, and is thus included in the explanatory variables. Due to the feedback control for controlling the coiling temperature, the cooling patterns of the feedback groups differ in the rolling direction of the rolled material, which may affect the mechanical properties. In order to take such interference into account, the cooling rate of the water cooling section of the feedback group is added to the explanatory variables. The cooling rate of the water cooling section of the feedback group V FB_cool The water cooling part of the feedback group is taken as the object and the calculation is performed using the following formula (4).
[0094]
Mathematical formula 4
[0095]
[0096] Here, t FB_s is the start time of water cooling of the feedback group, t FB_e is the end time of water cooling of the feedback group, T FB_s is the temperature at the start of water cooling of the feedback group, T FB_e It is the temperature when water cooling of the feedback group starts.
[0097] Figure 8 : is a diagram for explaining the structure of the data set preparation unit 28 and the preparation of the data set. The data set preparation unit 28 includes a condition setting unit 30 and a material calculation unit 31. The condition setting unit 30 sets the first operating condition related to the explanatory variable 37 of the data set 36. In addition, the condition setting unit 30 sets the operating conditions such as chemical composition, processing history, temperature history, etc. required for the metal structure calculation performed by the metal structure calculation unit 44 based on the metallurgical phenomenon model (hereinafter, the operating conditions related to the input of the metallurgical phenomenon model used by the metal structure calculation unit 44 are referred to as "second operating conditions") in the material calculation unit 31. The first operating condition and the second operating condition also include the same data, but also include different data. For example, the metallurgical phenomenon model predicts the change of microstructure at every moment. Therefore, the second operating condition includes data that is continuous in space and time at each device position and each moment of the processing history, temperature history, etc. The material calculation unit 31 calculates the metallurgical phenomenon and mechanical properties based on the set second operating condition. In addition, the material calculation unit 31 sets the data of the target variable 38 of the data set 36.
[0098] Fig. 9 This is a diagram for explaining the setting of the material calculation unit 31 and the explanatory variable 37 by the condition setting unit 30. The condition setting unit 30 is composed of, for example, an operating condition extraction unit 40, an operating condition creation unit 41, a material calculation input setting unit 42, and an explanatory variable setting unit 43.
[0099] The operating condition extraction unit 40 extracts the first operating condition related to the above-mentioned explanatory variables 37 such as chemical composition, processing history, temperature history, etc., and the second operating condition used in the material calculation unit 31 from the virtual rolling data of the virtual coil virtually rolled by the offline setting computer.
[0100] The above-mentioned offline setting computer is, for example, a device such as that disclosed in Patent Document 1, and by synchronizing the online setting computer with the process parameters, it is possible to simulate actual operation. Thus, a variety of operating conditions can be produced by offline setting calculations without applying a load to actual operation. For example, even for products and rolling conditions that have no manufacturing experience in actual operation, it is possible to simulate operations that meet various mechanical limitations.
[0101] The operating conditions and explanatory variables 37 used in the material calculation unit 31 may also include operating conditions and explanatory variables 37 extracted from actual rolling data of the product coil that is actually rolled. Alternatively, operating conditions such as processing history of strains and temperature history of various parts in the product may be prepared using more detailed analysis data than the set calculation, such as finite element analysis results.
[0102] The applicable operating conditions of the approximate model 39 should include disturbances that are not measured or calculated in actual operation. For example, disturbances refer to deviations in chemical composition, temperature unevenness in reheating of a heating furnace, temperature reduction of a rolled material caused by unexpected oscillation before finishing rolling, and failure to ensure cooling speed due to a bad sprayer. The disturbances are simulated in the above-mentioned offline setting computer. Alternatively, disturbances may be directly applied to the processing history and temperature history of actual rolling data or virtual rolling data.
[0103] In the operating condition creation unit 41, a change simulating disturbance is intentionally made to a part of the rolling data created by copying the actual rolling data or the virtual rolling data.
[0104] The material calculation input setting unit 42 transmits to the material calculation unit 31 the operating conditions extracted by the operating condition extraction unit 40, or the operating conditions obtained by the operating condition creation unit 41 by applying a change equivalent to interference to the operating conditions extracted by the operating condition extraction unit 40, or the operating conditions created by the operating condition creation unit 41 based on external data such as finite element analysis results without passing through the operating condition extraction unit 40.
[0105] The explanatory variable setting unit 43 extracts data corresponding to the explanatory variable 37 from the operating conditions extracted by the operating condition extraction unit 40, or from the operating conditions in which the operating conditions extracted by the operating condition extraction unit 40 are subjected to a change equivalent to interference by the operating condition creation unit 41, or from the operating conditions created by the operating condition creation unit 41 without passing through the operating condition extraction unit 40, and adds the data to the data set 36 as the explanatory variable 37.
[0106] Fig.10 This is a diagram for explaining the calculation of metallurgical phenomena and the setting of the target variable 38 by the material calculation unit 31. The material calculation unit 31 is composed of a metal structure calculation unit 44, a material property calculation unit 45, and a target variable setting unit 46.
[0107] The metal structure calculation unit 44 uses the rolling data such as chemical composition, processing history, temperature history, etc. set by the material calculation input setting unit 42 of the condition setting unit 11, and calculates the metal structure through the metallurgical phenomenon model that digitizes the metallurgical phenomenon. The characteristics of the metal structure to be calculated include the volume ratio of ferrite, pearlite, bainite, martensite, the grain size of ferrite, and austenite. Various metallurgical phenomenon models have been proposed, which are composed of a group of equations representing static recovery, static recrystallization, dynamic recovery, dynamic recrystallization, grain growth, and phase transformation. An example of this model is disclosed in Plastic Processing Technology Series 7 Plate Rolling (Corona Company) P198~229. By using this model, the volume ratio of ferrite, pearlite, bainite, martensite, etc., the austenite grain size, and the ferrite grain size can be calculated. In order to handle the changes in the micron-level organization, it is necessary to subdivide the calculation steps and calculation areas, and the calculation load is large. In the data set 36, it is necessary to include a large number of explanatory variables and target variables such as thousands to tens of thousands of coils. In order to implement the metal structure calculation part 44 under various operation conditions of thousands to tens of thousands of cases, the calculation is performed in an offline environment which does not affect the online calculation of the actual operation.
[0108] The material property calculation unit 45 calculates the mechanical properties based on the chemical composition and the calculated metal structure values obtained from the metal structure calculation unit 44. The mechanical properties to be measured include yield stress, tensile strength, elongation, etc. In addition, the metallurgical phenomenon model that digitizes the online setting computer metallurgical phenomenon often uses the actual value of the mechanical property obtained from the test results of the mechanical property measurement such as the tensile test implemented on a part of the product coil to automatically learn the model at any time (for example, the method of patent document 4, etc.). However, the mechanical property measured value is measured for the test piece cut from a part of the coil such as the front end or the tail end of the coil. Therefore, the measured value of the change of the mechanical property caused by the difference in the operating conditions such as the rolling direction and the temperature in the plate width direction is generally not obtained, and it cannot be reflected in the learning of the metallurgical phenomenon model. In the operating conditions where such mechanical property measured values are obtained, the mechanical property prediction results including the learned metallurgical phenomenon model become peculiar values, and the change of the mechanical property prediction results becomes uneven relative to the change of the operating conditions. In the material property calculation unit 45 , the metallurgical phenomenon model automatically learned in actual operation is not used, but a metallurgical phenomenon model that does not involve learning is used.
[0109] The target variable setting unit 46 extracts data corresponding to the target variable 20 from the mechanical properties generated by the material property calculation unit 45 , and adds the target variable 20 to the data set 36 as a pair with the explanatory variable 37 generated based on the operating conditions used by the material calculation unit 12 .
[0110] The model parameter determination unit 29 uses the data set 36 created by the data set creation unit 28 to determine model parameters constituting an approximate model 39 that simply simulates the mechanical property prediction calculation of the heavily loaded metal structure prediction in the material calculation unit 12 .
[0111] Fig.11 is a schematic diagram showing an example of an approximate model. As the approximate model 39, a machine learning model can be used. Fig.11 In the example shown, a machine learning model is constructed by a forward propagation neural network composed of an input layer, an intermediate layer, and an output layer. An explanatory variable 37 as an operation condition is input to the input layer. A target variable 20 as a mechanical property is output from the output layer.
[0112] In the example of constructing an approximate model by a forward propagation neural network, in the model parameter determination unit 29, the hyperparameters and parameters of the forward propagation neural network are determined. As hyperparameters, there are network structures (number of layers and units in the middle layer), types of activation functions, etc. Hyperparameters are determined empirically or by trial and error. Alternatively, they are determined by methods commonly used in recent years such as grid search and Bayesian optimization. Parameters are weight coefficients and deviations of each unit of the function that represents the forward propagation neural network. Although the function itself that predicts mechanical properties with high accuracy under all operating conditions is unclear, when a pair of inputs and outputs of multiple functions is given, by adjusting the parameters, a function that reproduces the input-output pair well can be produced. The input-output pair is called training data. Select parameters that make the output of the neural network when the input (explanatory variable) of the training data is assigned to the function as close as possible to the output (target variable) of the training data. The explanatory variable 37 of the data set 36 produced by the data set production unit 28 is used as input, and the target variable 20 is used as output. The explanatory variable 37 and the target variable 20 are used as output, and the parameters are adjusted using the pair of the explanatory variable 37 and the target variable 20. In learning, a part of the pairs in the data set, such as about 70% of the whole, is used as training data. As a measure of the reproducibility of the function expressed by the neural network, an error function expressed by square error, etc. is used. The neural network is learned by solving the minimization problem of the error function. In the learning of the forward propagation type neural network, as an optimization method for solving the minimization problem of the above-mentioned error function, there are known gradient descent methods and Adam as its improved methods. For example, the model parameter determination unit 29 uses Adam to learn the forward propagation type neural network. After confirming the generalization performance of the model by the retention method or cross-validation, the parameters are readjusted again as needed, sometimes after implementing the research usually carried out in the construction of the machine learning model, such as changing hyperparameters, increasing data pairs of the data set, and reconsidering data preprocessing. In addition, after the machine learning model is constructed in the above order, when a new pair of explanatory variables 37 and target variables 20 is added to the data set at any time, the gradient descent method is sometimes used to learn successively. This is the case where online rolling data is obtained at any time, and the case where offline setting calculations are performed at any time. In addition, during regular maintenance, the data set can be remade with the latest data, and the approximate model can be rebuilt and replaced. The type of explanatory variable may be changed depending on the type of mechanical property to be predicted. It is preferable to create a different approximate model for each mechanical property.
[0113] The material property prediction unit 27 uses the approximate model 39 previously created offline by the approximate model creation unit 26 to predict online the mechanical properties of each region in the three-dimensional grid of the product coil actually manufactured in the rolling line 1 .
[0114] Fig.12 2 is a block diagram showing the structure of the material property prediction unit 27. Fig.12 As shown, the material property prediction unit 27 includes a rolling data collection unit 32, a model input preparation unit 33, an approximate model calculation unit 34, and a material property output unit 35. The rolling data collection unit 32 collects rolling data of each part of the coil from the heating furnace to the coiling. The model input preparation unit 33 prepares input data for the approximate model 39 based on the rolling data collected by the rolling data collection unit 32. The approximate model calculation unit 34 calculates the mechanical properties of each area of the three-dimensional grid of the product coil based on the input data prepared by the model input preparation unit 33 and the approximate model 39. The material property output unit 35 outputs the mechanical properties calculated by the approximate model calculation unit 34 together with information indicating the position in the product.
[0115] The rolling data collection unit 13 collects rolling data such as the chemical composition, processing history, and temperature history of the product coil calculated by the setting computer 23. The rolling data is a predicted value before rolling, a predicted value and an actual recalculated value during rolling, and an actual recalculated value after rolling. The accuracy of the information varies depending on the timing of acquisition. When predicting mechanical properties for the purpose of product quality management, it is preferred to collect data at a timing that matches the actual recalculated value after the down coiler 19 completes coiling.
[0116] The model input preparation unit 33 extracts data corresponding to the explanatory variables 37 of each region in the three-dimensional grid shape of the product coil from the rolling data of the product coil collected by the rolling data collection unit 13 as input data to the approximate model 39. In the case where the rolling data collected by the rolling data collection unit 13 is insufficient to prepare the explanatory variables 37 of each region in the three-dimensional grid shape, the insufficient data is supplemented to prepare the explanatory variables 37. For example, in recent years, the temperature distribution in the thickness direction and the width direction of the rolled material is sometimes calculated based on the actual value measured by the thermometer and the model prediction, but due to the relationship of the calculation load, the calculation range is sometimes intentionally limited in the setting computer 23, for example, the temperature distribution in the width direction is calculated to the outlet side of the edge heater required for control, and the calculation is stopped downstream of the edge heater. In such a case, the actual value of the temperature distribution in the width direction measured by the multi-purpose detector at the outlet of the finishing mill, for example, downstream of the rolling data collection unit 13, the temperature distribution in the width direction at the outlet of the edge heater calculated by the model or the actual temperature distribution in the width direction of the scanning pyrometer at the outlet of the edge heater, and the temperature history calculation result of the center of the width direction from the outlet of the edge heater to the outlet of the finishing mill are used to generate the data of the explanatory variable 37 related to the temperature of each grid in the width direction, for example, the temperature at the inlet of the finishing mill. Alternatively, the temperature distribution in the width direction may be calculated using an offline setting computer. In addition, a part of the processing history such as the strain rate during rolling in the thickness direction of the plate is generally not implemented in the width direction of the plate. This is because, in the rolling in the thickness direction, the contact length between the roller and the deformation area of the material is very short compared to the width of the material, the movement of the material in the width direction is small, and the thickness reduction is mainly the elongation in the rolling direction. In this case, the data calculated in the center of the width direction can be copied in the width direction of the plate and used as the explanatory variable 37. Furthermore, for data such as chemical composition where only one value is available for the product coil, the same value is used in all grids.
[0117] The approximate model calculation unit 34 inputs the input data created by the model input creation unit 33 in advance into the approximate model 39 created by the approximate model creation unit 26 , and calculates the mechanical properties of each region of the three-dimensional mesh of the product coil.
[0118] Fig.13 : is a schematic diagram showing an example of a mesh region for which the approximate model calculation unit 34 calculates material properties. Fig.13In the figure, for simplicity, a part of the full length and one-side half of the plate width direction (the workpiece side or the drive side) are shown, but in the approximate model calculation unit 34, the full length and the full width are used as calculation objects. The setting calculation interval of the rolling direction sometimes varies according to the product plate thickness, but for example, the rolling data is produced at intervals of 2m, and the mechanical properties are predicted at the same intervals. In the plate thickness direction, for example, the mechanical properties of three points are predicted, namely, the upper surface, the center of the plate thickness, and the lower surface. In the width direction, for example, the mechanical properties of a total of five points are predicted, namely, the point 40mm away from the width end (the workpiece side and the drive side), the middle point between the width end and the center of the plate width (the 1 / 4 point of the plate width) (the workpiece side and the drive side), and the center point of the plate width.
[0119] The material property output unit 35 outputs the mechanical properties of each region calculated by the approximate model calculation unit 34, the information indicating the position of each region in the rolled product, and the information that is highly correlated with the change of the mechanical properties. The position of each region in the rolled product is the distance from the front end of the rolling direction, the distance from the tail end, the position in the width direction, and the position in the plate thickness direction. The information associated with the mechanical properties refers to the information that is highly correlated with the change of the mechanical properties, such as the finishing rolling side temperature, the coiling temperature, and the cooling rate. The output destination is a storage device such as a database and a visualization device such as an HMI. When output to a visualization device, the output data is visualized as a chart, or a numerical value is represented in a table. These are used as reference values when studying the implementation of the tensile test, the determination of the product grade, the cutting length at the dividing line, and the process parameters of the same batch of rolled products. Since the mechanical properties of each region, the information indicating the position in the rolled product, and the information that is highly correlated with the change of the mechanical properties can be obtained at a high speed, it can be reflected in the above-mentioned process parameters of the next batch of materials and the same batch of materials.
[0120] Fig.14 It is a schematic diagram showing an example of the output of the material property output unit 35. The horizontal axis represents the rolling direction position (the distance from the front end), and the vertical axis represents the change in tensile strength and coiling temperature. In addition, the position in the plate thickness direction and the width direction can be selected arbitrarily. In this example, the center of the plate thickness and the 1 / 4 point on the workpiece side in the width direction are selected. When the lower limit of the tensile strength is given, information such as the position of the rolled material with a tensile strength below the lower limit and the correlation with the coiling temperature can be obtained. When the horizontal axis is changed to the plate thickness direction and the plate width direction, the distribution of mechanical properties in each direction can be visualized. As another output method, the distribution of the mechanical properties of the entire product can also be visualized as a three-dimensional grid-like heat map. The information visualized in the table includes, for example, position information (the distance from the front end and the tail end) where the mechanical properties are not achieved, basic statistics such as the mean and standard deviation of the mechanical properties, and the correlation coefficient between the mechanical properties and the explanatory variable 37.
[0121] As described above, according to the present embodiment, a data set corresponding to rolling conditions (temperature conditions, processing conditions, time and speed conditions) that are not implemented in the actual rolling process or rolling conditions with little implementation experience is also pre-prepared and comprehensive. In this way, by using a data set prepared to include all rolling conditions including rolling conditions that are not implemented or have little implementation experience to prepare an approximate model offline, an approximate model that can correspond to the entire group of rolled products and has high approximate accuracy can be prepared. By using the approximate model prepared in this way, the computational load for online prediction of the material properties of each region of the three-dimensional grid of all rolled products can be reduced. Moreover, the material properties of all regions (parts) of the rolled product can be calculated and monitored. Thus, the result of the rolled product can also be reflected in the operation changes of the next rolled product or the rolled products in the same batch.
[0122] Implementation method 2.
[0123] Next, refer to Fig.15 as well as Fig.16 Embodiment 2 of the present invention will be described, but the description will be mainly focused on the differences from the above-mentioned Embodiment 1, and the same or corresponding parts are denoted by the same reference numerals and the description will be omitted.
[0124] Fig.15 : This is a block diagram showing the structure of the material property prediction unit 27 of the second embodiment. The mechanical property prediction unit 27 as a material property prediction unit includes a material property correction unit 47, which uses the mechanical property (hereinafter referred to as "the second mechanical property") calculated by the metallurgical phenomenon model that digitizes the metallurgical phenomenon to correct the mechanical property (hereinafter referred to as "the first mechanical property") calculated by the approximate model calculation unit 34 using the above-mentioned approximate model 39. The metallurgical phenomenon model that digitizes the metallurgical phenomenon for predicting the second mechanical property is used to perform model learning using the actual value of the mechanical property obtained through the tensile test of the past rolled product (product coil). The second mechanical property is collected from the online setting computer 23 by the rolling data collection unit 13 as a part of the rolling data.
[0125] Mechanical properties of a part of the rolled product, such as a representative point of the front end portion in the rolling direction, a representative point of the central portion, a representative point of the tail end portion, and the center of each width direction, are sometimes calculated by an online setting computer or a computer connected to the online setting computer using a metallurgical phenomenon model that digitizes the metallurgical phenomenon. In addition, the metallurgical phenomenon model that digitizes the metallurgical phenomenon uses the actual values of the mechanical properties obtained from the test results of the mechanical properties measured by tensile tests, organization observations, etc., implemented on a part of the product coils for model learning, or is adjusted to be consistent with the actual mechanical properties (for example, the method of Patent Document 4, etc.). It can be expected that the prediction accuracy of the predicted second mechanical property after such learning or adjustment is good.
[0126] An example of a correction method of the first mechanical property by the material property correction unit 47 is shown. It is assumed that the second mechanical property is obtained in the area j of the mesh area for which the approximate model calculation unit 34 calculates the first mechanical property. The following description takes the case where the mechanical property is tensile strength as an example.
[0127] First, the difference between the second mechanical property and the first mechanical property in the mesh region j where the second mechanical property is obtained is calculated by the following formula (5).
[0128] ΔTS(j)=TS MM (j)-TS ML (j)……(5)
[0129] Here, j is the index of the mesh region where the second mechanical property is obtained, TS MM (j) is the second mechanical property included in the actual rolling data, TS ML (j) is the first mechanical property calculated by the approximate model calculation unit 34, and ΔTS(j) is the difference between the second mechanical property and the first mechanical property.
[0130] Fig.15 : is a schematic diagram showing a correction method of the material property correction unit 47 according to the second embodiment of the present invention. For example, when the second mechanical property of the representative point of the front end portion, the representative point of the central portion, the representative point of the tail end portion in the rolling direction, and the center of each width direction are obtained, the difference ΔTS(i) between the second mechanical property assumed in other mesh regions i and the first mechanical property is calculated by linear interpolation or the like using ΔTS(j) of the above three points.
[0131] Next, the mesh region i is corrected by the following equation (6) using the difference ΔTS(i) between the second mechanical property and the first mechanical property.
[0132] TS comp (i) = TS ML (i)+α·ΔTS(i)……(6)
[0133] Here, i is the index of the mesh area to be corrected for tensile strength, TS ML (i) is the tensile strength calculated by the approximate model calculation unit 34 using the approximate model 39, α is the correction adjustment coefficient (=0 to 1), TS comp (i) is the corrected tensile strength.
[0134] As described above, according to this embodiment, since the material property correction unit 47 is provided, the prediction accuracy of the mechanical properties of all parts of the rolled product can be improved. Moreover, it will not affect the online calculation of the actual rolling operation. In addition, in the above, the example of predicting the mechanical properties of the material properties is described, but the same is true for the electromagnetic properties.
[0135] The embodiments of the present invention are described above, but the present invention is not limited to the above embodiments, and can be implemented in various ways without departing from the scope of the present invention. In the above embodiments, when the number, quantity, amount, range, etc. of each element is mentioned, the present invention is not limited to the mentioned quantity except for the case where it is specifically stated and the case where it is clearly determined to be the number in principle. In addition, the structures described in the above embodiments are not necessarily required by the present invention except for the case where it is specifically stated and the case where it is clearly determined to be the number in principle.
[0136] In the above-mentioned embodiment, the case where the mechanical property is predicted as the material property is described as an example, but the same is true for the case where the electromagnetic property is predicted.
[0137] Explanation of symbols:
[0138] 1: rolling production line; 25: material property prediction device for rolled products; 26: approximate model making unit; 27: material property prediction unit; 28: data set making unit; 29: model parameter determination unit; 30: condition setting unit; 31: material calculation unit; 36: data set; 37: explanatory variable; 38: target variable; 47: material property correction unit.
Claims
1. A material property prediction device for a rolled product, for predicting the material properties of a rolled product manufactured in a rolling production line, comprising: an approximate model making unit for making an approximate model offline, the approximate model comprehensively predicting the material properties of the group of rolled products manufactured in the above-mentioned rolling line; and a material property prediction unit that uses the approximate model created by the approximate model creation unit to predict online the material property of each region of the three-dimensional grid of the rolled product produced in the rolling line; The above-mentioned approximate model making department has: a data set preparation unit including a condition setting unit for setting rolling conditions for the above-mentioned rolled product group and a material calculation unit for calculating metallurgical phenomena and material properties under the above-mentioned rolling conditions, the data set preparation unit preparing a data set for preparing the above-mentioned approximate model; and The model parameter determination unit determines parameters expressing the approximate model using the data set.
2. The material property prediction device for rolled products according to claim 1, wherein: The data set creating unit creates the data set using the rolling conditions set by the condition setting unit as explanatory variables and the material properties calculated by the material calculation unit as target variables.
3. The material property prediction device for rolled products according to claim 1, wherein: The material property prediction unit includes: A rolling data collection unit, which collects online rolling data obtained when the rolled products are manufactured in the above rolling production line; A model input preparation unit, which prepares input data for the approximate model online based on the rolling data collected by the rolling data collection unit; an approximate model calculation unit that calculates material properties of each region of the three-dimensional grid of the product coil online by inputting the input data generated by the model input generation unit into the approximate model; and The material property output unit outputs the material property of each region calculated by the approximate model calculation unit, information indicating the position of each region in the rolled product, and information related to the material property.
4. The material property prediction device for rolled products according to claim 1, wherein: The above approximate model is a machine learning model.
5. The material property prediction device for a rolled product according to any one of claims 1 to 4, wherein: The material property prediction unit includes a material property correction unit that corrects the material property calculated by the approximate model calculation unit using the approximate model, using a material property result calculated using a metallurgical phenomenon model that digitizes the metallurgical phenomenon.
Citation Information
Patent Citations
Device for forecasting and controlling material quality of roll line
CN101253005A
Set value calculating device and set value calculating method
CN103920717A
Method for predicting material of steel plate
JP1993026872A
Rolling simulation device and computer readable recording medium recording rolling simulation program
JP2001025805A
Structure and material control system of rolling line
JP2008168320A
