Mandrel bar life prediction method, mandrel bar manufacturing method, and mandrel bar manufacturing condition setting device

Abstract

To provide a method for predicting a life of a mandrel bar that can extend a life of a mandrel bar, and to provide a method for manufacturing a mandrel bar and a device for setting a manufacturing condition of a mandrel bar.SOLUTION: A method for predicting a life of a mandrel bar that is manufactured through a hardening step and a tempering step includes, as input data, one or more parameters selected from attribute information of a mandrel bar, one or more parameters selected from operational parameters relating to the hardening step, and one or more parameters selected from operational parameters relating to the tempering step, and predicts a life of a mandrel bar using a life prediction model generated from machine learning which uses a life of a mandrel bar as output data.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present disclosure relates to a method for predicting the life of a mandrel bar, a method for manufacturing a mandrel bar, and an apparatus for setting manufacturing conditions for a mandrel bar. [Background technology]

[0002] Mandrel rolling is one of the processes for reducing the wall thickness of tubes. In this process, a mandrel bar, which reduces the wall thickness from the inside of the tube, is inserted into a blank tube, and both the blank tube and the mandrel bar are drawn between grooved rolls to reduce and elongate the tube. The grooved rolls are often installed in multiple stands, with a circumferential phase difference between them to ensure uniform reduction of the tube's outer circumference. Depending on the length of the tube to be produced and the number of stands, the mandrel bar, which is the internal tool, can be as long as several meters to several tens of meters. In mandrel rolling, the wall thickness is determined by the roll gap between the grooved rolls and the diameter of the mandrel bar, which reduces the tube from the inside, requiring high dimensional accuracy throughout its entire length. Furthermore, since the surface of the mandrel bar is transferred to the inner surface of the tube, good surface quality is essential. After mandrel rolling, the mandrel bar is gripped at its end by a drawing device and removed from the tube, where it is recycled for use in the mandrel rolling process again.

[0003] The mandrel bar generates high contact pressure with the inner surface of the tube being rolled. In hot rolling, the mandrel bar is exposed to high temperatures, making it difficult to apply lubricant after it is inserted into the inner surface of the tube. Due to these harsh operating conditions, the mandrel bar undergoes diameter reduction and surface seizure depending on the number of times it is used in rolling. Diameter reduction and seizure adversely affect the wall thickness and inner surface quality of the tube being produced. Therefore, the amount of diameter reduction and seizure status are controlled, and mandrel bars that adversely affect product quality are discarded. Here, mandrel bars are long and require various outer diameter sizes depending on the wall thickness of the tube being produced, resulting in high tool costs. Therefore, tool costs can be reduced by suppressing damage to the mandrel bar and extending its lifespan. For example, Patent Document 1 discloses a method for improving the toughness value of the mandrel bar while maintaining its hardness, thereby improving tool life, by performing accelerated cooling at a specific cooling rate after heating during tempering, the final stage of the heat treatment process used in mandrel bar manufacturing. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Patent No. 5835259 Summary of the Invention [Problem to be solved by the invention]

[0005] The method described in Patent Document 1 is effective in extending the tool life, but requires a device for accelerating cooling the mandrel bar at a specific cooling rate.

[0006] The present disclosure has been made in consideration of the above-described circumstances, and aims to provide a mandrel bar life prediction method, a mandrel bar manufacturing method, and a mandrel bar manufacturing condition setting device that can extend the life of the mandrel bar. [Means for solving the problem]

[0007] The inventors conducted extensive research into factors affecting the life of a mandrel bar and found that in addition to the tempering conditions during the manufacture of the mandrel bar, the quenching conditions also have a significant effect. Based on this finding, the inventors came up with a method of using a life prediction model in which input data includes attribute information on the mandrel bar, such as its components, operational parameters related to tempering, and operational parameters related to quenching, and the life of the mandrel bar is output data. By using such a life prediction model, the life of the mandrel bar can be accurately predicted. Furthermore, the inventors came up with a method of manufacturing a mandrel bar with a long life by searching for values ​​of operational parameters in the life prediction model that will extend the life of the mandrel bar and manufacturing the mandrel bar using those values. The present disclosure is based on the above findings. The gist of the disclosure is as follows.

[0008] (1) A method for predicting the life of a mandrel bar according to an embodiment of the present disclosure includes: A method for predicting the life of a mandrel bar manufactured through a quenching process and a tempering process, comprising: The life of the mandrel bar is predicted using a life prediction model generated by machine learning, which includes as input data one or more parameters selected from attribute information of the mandrel bar, one or more parameters selected from operational parameters related to the quenching process, and one or more parameters selected from operational parameters related to the tempering process, and which outputs the life of the mandrel bar.

[0009] (2) As one embodiment of the present disclosure, in (1), The operational parameters for the quenching step include at least one of a quenching temperature and a quenching time.

[0010] (3) As an embodiment of the present disclosure, in (1) or (2), The operational parameters for the tempering step include at least one of a tempering temperature and a tempering time.

[0011] (4) As an embodiment of the present disclosure, Using any one of the mandrel bar life prediction methods (1) to (3), the life of the mandrel bar is predicted using attribute information of the mandrel bar, set values ​​of operational parameters in the quenching process, and set values ​​of operational parameters in the tempering process as input data for the life prediction model, and a determination is made as to whether the predicted life prediction value of the mandrel bar is equal to or greater than a target value. If the predicted life prediction value is less than the target value, at least one operational parameter in the quenching process and the tempering process is reset, and the life prediction of the mandrel bar is continued until the predicted life prediction value is equal to or greater than the target value.

[0012] (5) A method for manufacturing a mandrel bar according to an embodiment of the present disclosure includes: (4) Heat treatment of the mandrel bar is carried out based on the heat treatment conditions when the life prediction value predicted by the mandrel bar life prediction method is equal to or greater than the target value.

[0013] (6) A mandrel bar manufacturing condition setting device according to an embodiment of the present disclosure includes: The system includes a life prediction unit that predicts the life of the mandrel bar using attribute information of the mandrel bar, set values ​​of operational parameters in the quenching process, and set values ​​of operational parameters in the tempering process as input data for the life prediction model, and a heat treatment condition setting unit that determines whether the predicted life value satisfies a target value and, if the target value is not satisfied, resets at least one of the operational parameters in the quenching process and the tempering process. [Effects of the Invention]

[0014] According to the present disclosure, it is possible to provide a mandrel bar life prediction method, a mandrel bar manufacturing method, and a mandrel bar manufacturing condition setting device that can extend the life of the mandrel bar. [Brief explanation of the drawings]

[0015] [Figure 1] FIG. 1 is a diagram for explaining a method for generating a life prediction model of a mandrel bar. [Figure 2] FIG. 2 is a diagram for explaining a method for predicting the life of a mandrel bar. DETAILED DESCRIPTION OF THE INVENTION

[0016] A mandrel bar life prediction method, a mandrel bar manufacturing method, and a mandrel bar manufacturing condition setting device according to an embodiment of the present disclosure will be described below with reference to the drawings. The present disclosure generally relates to extending the life of a mandrel bar used in mandrel rolling. According to the present disclosure, a method is provided for predicting the life (number of rolled pieces) of a mandrel bar from its manufacture until cracks develop on the surface and the bar becomes unusable after being used in rolling. Also provided are a method and device for manufacturing a mandrel bar with a long life based on the prediction.

[0017] A mandrel bar is an internal tool used in the mandrel rolling process. Hot work tool steels such as SKD6 or SKD61 specified in JIS are generally used as the steel type for manufacturing mandrel bars. Table 1 shows the steel compositions of SKD6 and SKD61. The balance not shown in Table 1 is Fe and unavoidable impurities.

[0018] [Table 1]

[0019] After being machined to a specified size, mandrel bars are manufactured through heat treatments of quenching and tempering. The quenching process is a process in which the mandrel bar is heated and then rapidly cooled to improve its strength. The tempering process is a process in which the quenched mandrel bar is placed in an appropriate temperature range to improve its toughness.

[0020] When a mandrel bar used in mandrel rolling is rolled, cracks develop in the circumferential direction on its surface as the number of rolls increases. These cracks in the mandrel bar in the circumferential direction cause defects on the inner surface of the steel pipe. The inventors conducted a detailed study of the damage mechanism of the mandrel bar and found that the quenching conditions, in addition to the tempering conditions during the manufacture of the mandrel bar, have a significant effect.

[0021] SKD6 and SKD61 contain alloying elements such as Mn, Cr, Mo, and V, which are precipitated as carbides during heat treatment to enhance their hot strength. According to the inventors' research, when the quenching time is short or the quenching temperature is low, the carbides that segregated at the grain boundaries before quenching do not dissolve but remain segregated, reducing the toughness of the material and making the mandrel bar more susceptible to cracking. This shortens the life of the mandrel bar.

[0022] Furthermore, according to the inventors' investigations, it has become clear that when the quenching time is long or the quenching temperature is high, the crystal grains become large, which reduces the toughness of the material and makes the mandrel bar more susceptible to cracks, thereby shortening the life of the mandrel bar.

[0023] In this way, the quenching conditions affect the life of the mandrel bar. Therefore, in this embodiment, a mandrel bar life prediction model is used, which includes, as input data, attribute information such as the composition of the mandrel bar, operational parameters related to tempering, and operational parameters related to quenching. The output data of the life prediction model is the life. By using such a life prediction model, it is possible to accurately predict the life of the mandrel bar.

[0024] <Mandrel bar attribute information> The attribute information of the mandrel bar can include the dimensions of the mandrel bar and the chemical composition of the mandrel bar. The chemical composition is the content of component elements such as C, Si, Mn, P, S, Cr, Mo, and V. The attribute information of the mandrel bar preferably includes the contents of C, Mn, Cr, Mo, and V. This is because C, Mn, Cr, Mo, and V precipitate as carbides and affect the toughness of the mandrel bar. The attribute information of the mandrel bar also preferably includes the contents of P and S. This is because P and S segregate at grain boundaries and may affect the toughness of the mandrel bar. The attribute information of the mandrel bar also preferably includes the content of Si. This is because Si forms oxides on the surface of the mandrel bar and affects the surface condition. In other words, Si affects the occurrence of cracks on the surface of the mandrel bar.

[0025] <Tempering operation parameters> The tempering operation parameters that can be used include the tempering temperature, tempering time, tempering heating rate, and cooling rate after tempering. The tempering operation parameters preferably include the tempering temperature. This is because the tempering temperature changes the carbides that precipitate, which affects the toughness of the mandrel bar. The tempering operation parameters preferably include the cooling rate after tempering. This is because the cooling rate after tempering changes the segregation morphology of P and S to grain boundaries, which affects the toughness of the mandrel bar.

[0026] <Quenching operating parameters> The quenching operation parameters that can be used include the quenching temperature, quenching time, quenching heating rate, and cooling rate after quenching. The quenching operation parameters preferably include the quenching temperature, because the quenching temperature changes the state of dissolution of carbides and the size of crystal grains, which affect the toughness of the mandrel bar. The quenching operation parameters preferably include the quenching time, because the quenching time changes the state of dissolution of carbides and the size of crystal grains, which affect the toughness of the mandrel bar.

[0027] <Mandrel bar life performance> The actual lifespan of the mandrel bar is stored in a host computer. The host computer may be, for example, a computer that manages the manufacture of mandrel bars and the manufacture of steel pipes and the like, including the mandrel rolling process using the mandrel bar. The lifespan of the mandrel bar may be determined based on the presence or absence of surface cracks detected visually or by an inspection device after use in mandrel rolling. When a crack occurs on the surface of the mandrel bar, the number of rolls that have been rolled up to that point may be recorded in the host computer as the actual lifespan of the mandrel bar. Here, the index indicating the lifespan of the mandrel bar is not limited to the number of rolls, and may be a value different from the number of rolls. For example, when taking the rolling load into consideration, the value obtained by multiplying the rolling load by the number of rolls may be integrated until a crack occurs, and the integrated value may be recorded in the host computer as the actual lifespan of the mandrel bar.

[0028] <Creating a mandrel bar life prediction model> FIG. 1 shows a method for generating a life prediction model for predicting the life of a mandrel bar. The model generation unit generates a life prediction model. In this embodiment, the model generation unit collects attribute information of the mandrel bar, tempering operation performance data, quenching operation performance data, and life performance data of the mandrel bar. Here, the tempering operation performance data is data of operation parameters related to the tempering process during the manufacture of the mandrel bar from which the life performance data was obtained. Furthermore, the quenching operation performance data is data of operation parameters related to the quenching process during the manufacture of the mandrel bar from which the life performance data was obtained. The model generation unit generates a life prediction model through machine learning using the collected data (performance data).

[0029] The attribute information of the mandrel bar, the tempering operation performance data, the quenching operation performance data, and the mandrel bar life performance data are sent from a host computer to the model generation unit. The model generation unit may be realized by a computer capable of communicating with the host computer and various devices that make up each operation line. The configuration of the computer is not particularly limited and may include, for example, a memory (storage device), a CPU (processing device), a hard disk drive (HDD), a communication control unit for connecting to a network, a display device, and an input device. Here, the database in FIG. 1 may be realized by a hard disk drive. The machine learning unit that generates the predictive model may be realized by a CPU. The predictive model may be stored in memory.

[0030] A plurality of data sets of performance data are collected and stored in a database. Of the performance data, the attribute information of the mandrel bar, the tempering operation performance data, and the quenching operation performance data become input performance data. Furthermore, the life performance data of the mandrel bar corresponding to the input performance data becomes output performance data. The number of data stored in the database is preferably 50 or more, more preferably 100 or more, and even more preferably 200 or more.

[0031] In this embodiment, the machine learning unit generates a life prediction model for the mandrel bar through machine learning using learning data, which is a data set of input performance data and output performance data. A plurality of pieces of learning data are used to generate the life prediction model, but the number of pieces is not limited to a specific number. Here, the input performance data includes one or more parameters selected from the attribute information of the mandrel bar, one or more parameters selected from the operation parameters related to the quenching process, and one or more parameters selected from the operation parameters related to the tempering process.

[0032] A known learning method may be applied as the machine learning method. For example, a known machine learning method such as a neural network may be used for the machine learning. Other examples of machine learning methods include decision trees, random forests, and support vector regression. Furthermore, the mandrel bar life prediction model may be updated as appropriate using learning data obtained from new performance data. The machine learning unit may store the generated life prediction model in a memory as described above.

[0033] <Mandrel bar life prediction method> The life prediction of the mandrel bar is performed using a life prediction model that has been generated in advance, after the attribute information of the slab of the mandrel bar to be predicted is obtained.

[0034] The heat treatment conditions can be reset using the results of the mandrel bar life prediction. The heat treatment conditions are operational parameters for the quenching and tempering processes. The resetting of the heat treatment conditions is performed before the quenching and tempering heat treatments of the mandrel bar to be predicted are performed.

[0035] FIG. 2 is a diagram for explaining a method for predicting the life of a mandrel bar. The life prediction unit and the heat treatment condition setting unit may be realized by a computer. Here, the computer realizing the life prediction unit and the heat treatment condition setting unit may be the same computer as the computer realizing the model generation unit, or may be a different computer. If it is a different computer, it is sufficient that it can communicate with the computer realizing the model generation unit and a higher-level computer so as to acquire a life prediction model, etc. Here, the processing of the life prediction unit and the heat treatment condition setting unit may be realized by a CPU. The predicted life value of the mandrel bar predicted using the life prediction model may be stored in memory.

[0036] The life prediction unit acquires attribute information of the mandrel bar, preset settings for tempering conditions (initial conditions), and preset settings for quenching conditions (initial conditions) from a host computer. Here, the tempering conditions are operational parameters for the tempering process. The quenching conditions are operational parameters for the quenching process. The life prediction unit uses the initial settings for the tempering and quenching operational parameters as input data for the life prediction model. The life prediction unit then obtains output data for the initial conditions, i.e., the predicted life value of the mandrel bar.

[0037] The heat treatment condition setting unit determines whether the predicted life expectancy of the mandrel bar is equal to or greater than a target value. If the predicted life expectancy of the mandrel bar is equal to or greater than the target value (i.e., satisfies the target value), the heat treatment condition setting unit determines that the set values ​​(initial conditions) are appropriate heat treatment conditions and terminates the process. If the life expectancy is less than the target value (i.e., does not satisfy the target value), the resetting unit of the heat treatment condition setting unit resets at least one operating parameter in the quenching process and the tempering process so as to extend the life expectancy of the mandrel bar.

[0038] The life prediction unit may predict the life of the mandrel bar again using the reset heat treatment conditions (operation parameters for the quenching process and the tempering process) as input data for the life prediction model. By repeatedly predicting the life of the mandrel bar in this manner, appropriate heat treatment conditions can be determined.

[0039] Here, when resetting the heat treatment conditions, the resetting unit may change, for example, at least one of the quenching temperature, quenching time, quenching heating rate, and cooling rate after quenching. Furthermore, in addition to or instead of changing the quenching conditions, the resetting unit may change, for example, at least one of the tempering temperature, tempering time, tempering heating rate, and cooling rate after tempering. Alternatively, multiple heat treatment conditions may be prepared in advance, and a life prediction for the mandrel bar may be performed under each condition. The best heat treatment condition among those that results in a life prediction value equal to or greater than the target value may be determined as the optimal heat treatment condition. Here, when preparing multiple heat treatment conditions, the conditions may be set using a method such as experimental design. Furthermore, an optimization method such as sequential quadratic programming may be used.

[0040] As described above, the mandrel bar life prediction method, mandrel bar manufacturing method, and mandrel bar manufacturing condition setting device according to the present embodiment can extend the life of the mandrel bar by using the above-described configurations or processes. In other words, according to the present disclosure, the life of the mandrel bar can be predicted with high accuracy, and operational parameters can be reset to extend the life based on the prediction, resulting in a long life of the mandrel bar.

[0041] (Example) The effects of the present disclosure will be specifically described below based on examples, but the present disclosure is not limited to the contents of the examples.

[0042] In this example, a hot mandrel rolling process was carried out in the production of seamless steel pipes.

[0043] Carbon steel with the chemical composition shown in Table 2 was melted in a vacuum melting furnace and formed into a round billet. Then, using the Mannesmann piercing method, blank tubes with diameters of 130–230 mm and wall thicknesses of 13–35 mm were produced. Mandrel rolling was performed at temperatures ranging from 1000–1150°C to reduce the wall thickness to 4–18 mm. This was performed using 250 mandrel bars until each mandrel bar reached its service life (limit of use). The mandrel bar surface was visually inspected after each rolling run, and if cracks were found, the number of rolled bars at that point was recorded as the service life. For example, if a crack appeared on the surface of one of the 250 mandrel bars after mandrel rolling 2000 blank tubes, the actual service life of the mandrel bar with the crack would be 2000.

[0044] [Table 2]

[0045] For the above 250 mandrel bars, a mandrel bar life prediction model was created using the mandrel bar attribute information, tempering operation performance data, quenching operation performance data, and mandrel bar life performance data.

[0046] The input data for the prediction model were the C, Si, Mn, P, S, Cr, Mo, and V contents of the mandrel bar, the tempering temperature, the cooling rate after tempering, the quenching temperature, and the quenching time. The C, Si, Mn, P, S, Cr, Mo, and V contents were measured during the steelmaking process when the mandrel bar material was produced. The cooling rate after tempering [°C / sec] was calculated as (tempering temperature [°C] - 100 [°C]) / (time [sec] from the tempering temperature to 100°C). The outer surface temperature of the longitudinal center of the mandrel bar during cooling was measured with a radiation thermometer, and the time required for the temperature to reach 100°C from the tempering temperature was measured.

[0047] The machine learning method used was a neural network with three intermediate layers. The activation function was a sigmoid function. The actual data for the 250 mandrel bars mentioned above was used as the data (training data) for creating the model.

[0048] Next, one mandrel bar (first mandrel bar) separate from the 250 mandrel bars described above was prepared. Using the life prediction model created above, a life prediction was performed using the attribute information and heat treatment condition settings of the first mandrel bar as input data. As a result, the life of the first mandrel bar was predicted to be 2002 pieces. The first mandrel bar was heat treated under the heat treatment conditions entered into the life prediction model. A hot mandrel rolling process was performed using the first mandrel bar to manufacture seamless steel pipes. Carbon steel having the chemical composition shown in Table 2 was melted in a vacuum melting furnace and formed into a round billet. After that, a blank pipe with a diameter of 130 to 230 mm and a wall thickness of 13 to 35 mm was manufactured using the Mannesmann piercing method. Mandrel rolling was performed at a temperature range of 1000 to 1150°C, and repeated rolling was performed to reduce the wall thickness to 4 to 18 mm. The surface of the mandrel bar was visually inspected after each rolling, and if a crack was found, the number of rolls at that point was recorded as the lifespan. As a result, the lifespan of the first mandrel bar was 2003.

[0049] Next, another mandrel bar (second mandrel bar) was prepared. Using the life prediction model created above, a life prediction was performed using the attribute information of the second mandrel bar and the heat treatment condition settings as input data. As a result, the life of the second mandrel bar was predicted to be 2,000 pieces. Setting the target value at 2,040 pieces, and repeatedly performing life predictions by changing the tempering temperature and quenching temperature, a heat treatment condition was found that predicted the life of the second mandrel bar to be 2,042 pieces. The second mandrel bar was heat treated under that heat treatment condition. A hot mandrel rolling process using the second mandrel bar was performed to manufacture seamless steel pipes. Carbon steel with the chemical composition shown in Table 2 was melted in a vacuum melting furnace and formed into a round billet. After that, a blank pipe with a diameter of 130 to 230 mm and a wall thickness of 13 to 35 mm was manufactured using the Mannesmann piercing method. Mandrel rolling was performed at a temperature range of 1000 to 1150°C, and rolling was repeated to reduce the wall thickness to 4 to 18 mm. The mandrel bar surface was visually observed after each rolling run, and if cracks were found, the number of rolled bars at that point was recorded as the lifespan. As a result, the lifespan of the second mandrel bar was 2042 bars. The results of this example confirmed that the method and apparatus described in the above embodiment can predict the lifespan of a mandrel bar with high accuracy and can achieve a long lifespan of the mandrel bar.

[0050] Although the embodiments of the present disclosure have been described based on the drawings and examples, it should be noted that those skilled in the art would easily be able to make various modifications or alterations based on the present disclosure. Therefore, it should be noted that these modifications and alterations are included within the scope of the present disclosure. For example, the functions included in each component or step can be rearranged so as not to cause logical inconsistencies, and multiple components or steps can be combined or divided into one. The embodiments of the present disclosure can also be realized as a storage medium on which a program executed by a processor included in an apparatus is recorded. It should be understood that these are also included within the scope of the present disclosure.

Claims

1. A method for predicting the life of a mandrel bar manufactured through a quenching process and a tempering process, comprising: A mandrel bar life prediction method for predicting a life of the mandrel bar using a life prediction model generated by a neural network machine learning technique, using as input data the contents of C, Si, Mn, P, S, Cr, Mo, and V of the mandrel bar, at least one of a tempering temperature and a tempering time, and at least one of a cooling rate after tempering and a quenching temperature and a quenching time, and using as output data the number of rolled pieces as the life of the mandrel bar.

2. 2. A method for predicting a life of a mandrel bar, comprising: predicting the number of rolled pieces as the life of the mandrel bar by using the method for predicting a life of a mandrel bar according to claim 1; determining whether the predicted number of rolled pieces is equal to or greater than a target value; and, if the number of rolled pieces is less than the target value, resetting at least one operational parameter in the quenching process and the tempering process, and continuing to predict the life of the mandrel bar until the number of rolled pieces is equal to or greater than the target value.

3. A method for manufacturing a mandrel bar, comprising: heat treating the mandrel bar based on heat treatment conditions when the number of rolled pieces predicted by the method for predicting the life of a mandrel bar according to claim 2 is equal to or greater than a target value.

4. A mandrel bar manufacturing condition setting device comprising: a life prediction unit that predicts the number of rolled pieces as the life of a mandrel bar, using at least one of the C, Si, Mn, P, S, Cr, Mo, and V contents, the tempering temperature and tempering time, and the cooling rate after tempering, and at least one of the quenching temperature and quenching time, of a mandrel bar manufactured through a quenching process and a tempering process as input data for a life prediction model generated by a neural network machine learning technique; and a heat treatment condition setting unit that determines whether the number of rolled pieces satisfies a target value, and if it does not satisfy the target value, resets at least one operating parameter in the quenching process and the tempering process.

Images