Manufacturing specification determination support device, manufacturing specification determination support method, steel plate manufacturing method, welded structure manufacturing method, and program

Abstract

The present invention provides a manufacturing specification determination support device, a manufacturing specification determination support method, a steel plate manufacturing method, a welded structure manufacturing method, and a program. [Solution] The manufacturing specification determination support device of the present invention comprises: an information acquisition unit that acquires the performance data of welding steel plates, the welding specifications of welded joints, and the HAZ toughness of welded joints as performance data information; a HAZ toughness estimation unit that calculates an estimated HAZ toughness value based on the steel plate characteristics and welding specifications included in the input information using a HAZ toughness prediction model constructed with past performance data including the performance data of welding steel plates, the welding specifications of welded joints, and the HAZ toughness value as training data; a search processing unit that searches for manufacturing specifications such that the estimated HAZ toughness value falls within a desired range; and an output instruction unit that provides instructions for displaying and outputting the manufacturing specifications.

Description

Technical Field

[0001] The present invention relates to a manufacturing specification determination support device, a manufacturing specification determination support method, a steel plate manufacturing method, a welded structure manufacturing method, and a program, which are particularly suitable for supporting the creation of manufacturing specifications related to steel plates for welding and welded structures.

Background Art

[0002] Steel materials such as steel plates (the steel plates include "thick steel plates") and steel pipes manufactured from steel plates are used for steel structures such as ships, offshore structures, bridges, buildings, tanks, construction machinery, and line pipes, and a welded structure is manufactured by welding a plurality of steel materials. Generally, the mechanical properties required for the welded parts of these structures are determined at the design stage of the welded joint, and the mechanical properties of the steel materials and the welded parts are defined at this design stage.

[0003] The mechanical properties of the welded part change depending on the component composition of the steel plate, the manufacturing conditions of the steel plate, and the welding conditions of the steel plate, etc., and these factors are determined in a complex entanglement. Therefore, it is extremely difficult to predict the mechanical properties of the welded part formed by welding the steel plate.

[0004] As a method for solving such problems, the techniques of Patent Documents 1 and 2 can be cited. In Patent Document 1, a method for predicting the toughness in the heat-affected zone (hereinafter referred to as "HAZ toughness"), which is the material (mechanical properties) of the welded part including the heat-affected zone (hereinafter referred to as "HAZ"), based on the steel plate component, wire component, and welding conditions has been proposed.

[0005] Further, in Patent Document 2, when joining various materials, a method has been proposed to search for and determine the optimal steel plate components and welding conditions that satisfy the HAZ toughness of the welded part, which is the joining property required for the joined materials, by a neural network.

Prior Art Documents

Patent Documents

[0006] [Patent Document 1] Japanese Patent Publication No. 2004-004034 [Patent Document 2] Japanese Patent Publication No. 2003-039180 [Overview of the Initiative] [Problems that the invention aims to solve]

[0007] As described above, conventional methods for predicting the mechanical properties of welds have been proposed. However, these prediction methods only used the component composition of the welding material as input information for predicting the mechanical properties of the weld. Furthermore, conventional prediction methods do not take metallurgical phenomena into account, and therefore could not consider the behavior of inclusions when subjected to heat from welding, which affects the HAZ toughness of the weld. For these reasons, conventional techniques have the problem of not being able to predict the HAZ toughness of welds with high accuracy.

[0008] This invention was completed in view of the above problems, and its purpose is to provide a manufacturing specification determination support device, a manufacturing specification determination support method, and a program that can improve the accuracy of predicting HAZ toughness, specifically by considering metallurgical phenomena during welding when creating manufacturing specifications for steel plates (e.g., welding steel plates) and welded structures. It also aims to provide a method for manufacturing steel plates and welded structures using the created manufacturing specifications. [Means for solving the problem]

[0009] To achieve the above objectives, the inventors diligently investigated methods for improving the accuracy of predicting HAZ toughness in welded joints. As a result, they found that it is effective to estimate HAZ toughness by considering the metallurgical phenomena that occur in welded joints when welding multiple steel plates, and then to search for the optimal manufacturing specifications for welded steel plates and welded structures based on this estimated HAZ toughness.

[0010] This invention was completed based on the above findings, and its gist is as follows. [1] An information acquisition unit that acquires as performance data information the performance of steel plates for welding, including the steel plate component composition, steel plate manufacturing conditions and steel plate characteristic values, the performance of welding specifications for welded joints, including welding conditions, and the performance of HAZ toughness of the said welded joints. A HAZ toughness estimation unit calculates an estimated HAZ toughness value based on the steel plate characteristics and welding specifications included in the input information, using a HAZ toughness prediction model constructed by relating the training data with the HAZ toughness results, which includes past performance data including the performance of the steel plate for welding, the welding specifications of the welded joint, and the HAZ toughness results, as training data. A search processing unit that searches for manufacturing specifications such that the calculated HAZ toughness estimate falls within a desired range, An output instruction unit that provides instructions for displaying and outputting the searched manufacturing specifications, A manufacturing specification determination support device characterized by comprising the following features. [2] The manufacturing specification determination support apparatus according to [1] above, wherein the HAZ toughness estimation unit selects the training data according to the steel plate thickness and welding heat history predicted from the welding specifications included in the input information. [3] The manufacturing specification determination support device according to [1] or [2] above, characterized in that the search processing unit further searches for the manufacturing specifications such that the calculated HAZ toughness estimate satisfies the manufacturing cost threshold. [4] An information acquisition step to acquire as performance data information the performance of welding steel plates including steel plate component composition, steel plate manufacturing conditions and steel plate characteristic values, the performance of welding specifications for welded joints including welding conditions, and the performance of HAZ toughness of the said welded joints, A HAZ toughness estimation step involves using past performance data, including the performance of the welding steel plate, the welding specifications of the welding joint, and the HAZ toughness performance, as training data, and using a HAZ toughness prediction model constructed by relating the training data with the HAZ toughness performance, to calculate an estimated HAZ toughness value based on the steel plate characteristics and welding specifications included in the input information. A search process step for searching for manufacturing specifications such that the calculated HAZ toughness estimate falls within a desired range, An output instruction step that provides instructions for displaying and outputting the searched manufacturing specifications, A method for supporting the determination of manufacturing specifications, characterized by comprising the following features. [5] The manufacturing specification determination support method according to [4] above, characterized in that the HAZ toughness estimation step selects the training data according to the steel plate thickness and welding heat history predicted from the welding specifications included in the input information. [6] The manufacturing specification determination support method according to [4] or [5] above, characterized in that the search processing step further searches for the manufacturing specifications such that the calculated HAZ toughness estimate satisfies a manufacturing cost threshold. [7] A method for manufacturing a steel sheet, characterized by comprising a step of manufacturing a steel sheet for welding based on the manufacturing specifications of the steel sheet for welding searched by the manufacturing specification determination support device described in any one of [1] to [3] above. [8] A method for manufacturing a welded structure, comprising the step of manufacturing a welded structure based on the manufacturing specifications of the welded joint found by the manufacturing specification determination support device described in any one of [1] to [3] above. [9] A program used to determine manufacturing specifications, An information acquisition process that acquires, as performance data information, the performance of welding steel plates including steel plate component composition, steel plate manufacturing conditions, and steel plate characteristic values, the performance of welding specifications for welded joints including welding conditions, and the performance of HAZ toughness of said welded joints. A HAZ toughness estimation process calculates an estimated HAZ toughness value based on the steel plate characteristics and welding specifications included in the input information, using a HAZ toughness prediction model constructed by associating the training data with the HAZ toughness performance, which includes past performance data including the performance of the steel plate for welding, the welding specifications of the welded joint, and the HAZ toughness performance, as training data. A search process to find manufacturing specifications such that the calculated HAZ toughness estimate falls within a desired range, An output instruction process that provides instructions for displaying and outputting the searched manufacturing specifications, A program characterized by causing a computer to execute it. [Effects of the Invention]

[0011] According to the present invention, as a metallurgical phenomenon during welding, considering the behavior of inclusions when receiving heat from welding that affects the HAZ toughness of the welded part, the mechanical properties of the welded part can be predicted. Thereby, it is possible to provide a manufacturing specification determination support device, a manufacturing specification determination support method, and a program that improve the prediction accuracy of HAZ toughness. Further, it becomes possible to manufacture a steel plate or a welded structure using the obtained manufacturing specifications.

Brief Description of the Drawings

[0012] [Figure 1] FIG. 1 is a block diagram showing a manufacturing specification determination support device according to an embodiment of the present invention. [Figure 2] FIG. 2 is a flowchart for explaining the processing of the manufacturing specification determination support method in the manufacturing specification determination support device shown in FIG. 1. [Figure 3] FIG. 3 is a flowchart showing an example of a prediction model creation process in an embodiment of the present invention. [Figure 4] FIG. 4 is a flowchart showing a process of calculating an estimated value of HAZ toughness as an example of the prediction model creation process shown in FIG. 3. [Figure 5] FIGS. 5(A) and (B) are schematic diagrams showing an example of the overall outline of a system in which the present invention is applied to the manufacturing processes of a steel plate and a welded structure. [Figure 6] FIG. 6 is a graph showing the verification results of the prediction accuracy of HAZ toughness in the manufacturing specifications searched by the present invention. [Figure 7] FIG. 7 is a graph showing the verification results of the prediction accuracy of HAZ toughness in the manufacturing specifications searched by the present invention. [Figure 8] FIG. 8 is a graph showing the verification results of the prediction accuracy of HAZ toughness in the manufacturing specifications searched by the present invention. [Figure 9] FIG. 9 is a graph showing the verification results of the prediction accuracy of HAZ toughness in the manufacturing specifications searched by a conventional prediction method.

Embodiments for Carrying Out the Invention

[0013] Embodiments of the present invention will be described below with reference to the figures. However, the present invention is not limited to these embodiments.

[0014] As described above, the manufacturing specification determination support device and manufacturing specification determination support method of the present invention can be suitably used to support the creation of manufacturing specifications for steel plates (e.g., welding steel plates) and manufacturing specifications for welded structures formed by welding said welding steel plates. Examples of applications will be described later, but here, as one embodiment of the present invention, System 1 is configured to include a process computer or distributed control system (DCS) 2 that manages each process constituting a manufacturing line (also referred to as a manufacturing process), and a manufacturing specification determination support device 3 (see Figure 5). Each process is managed by the process computer or distributed control system (hereinafter referred to as "process computer, etc.") 2.

[0015] First, the manufacturing specification determination support device of the present invention will be described with reference to Figure 1. Figure 1 shows a block diagram of the manufacturing specification determination support device 3 of the present invention.

[0016] As shown in Figure 1, the manufacturing specification determination support device 3 of the present invention comprises an input unit 31, a storage unit 32, an output unit 33, a communication unit 34, and a device body 35. The arithmetic processing unit 351 included in the device body 35 includes an information acquisition unit 3a, a HAZ toughness estimation unit 3b, a search processing unit 3c, and an output instruction unit 3d, which will be described later. The device body 35 transmits and receives necessary information and performs information processing by communicating with the input unit 31, storage unit 32, output unit 33, and communication unit 34 via a bus 36. In the example shown in Figure 1, a wired connection is used, but the connection method is not limited to this, and a wireless connection or a combination of wired and wireless connections is also possible.

[0017] In general, the manufacturing specification determination support device 3 in System 1 of this embodiment uses a HAZ toughness prediction model constructed based on past performance data in the manufacturing process of welded structures and the manufacturing process of welding steel plates used in said welded structures to calculate an estimated HAZ toughness value based on the information input from the input unit 31 (specifically, steel plate characteristics and welding specifications), and searches for manufacturing specifications for welding steel plates and welded structures such that the estimated HAZ toughness value falls within the range of the information input from the input unit 31 (specifically, the target characteristic value of the welded joint, which is referred to as the "desired value").

[0018] The searched manufacturing specification information is transmitted, for example, by the communication unit 34 to the process computer 2. The process computer 2 gives instructions to each process based on the received manufacturing specifications and, based on those instructions, appropriately controls each piece of equipment such as heating furnaces and rolling mills to carry out each process.

[0019] The input unit 31 is any input interface capable of detecting operations performed by the system administrator, such as a keyboard, pen tablet, touchpad, and mouse. The input unit 31 receives operations that serve as instructions for each process of the main unit 35.

[0020] The memory unit 32 is a recording medium such as a hard disk drive, semiconductor drive, optical disk drive, magneto-optical disk drive, and magnetic tape, and is a device that stores information necessary for System 1. The memory unit 32 stores past performance data. For example, it stores the component composition, steel plate ID, manufacturing specification, and mechanical properties of each steel plate manufactured in the past, as well as welding specification data linked to welding conditions for each welded joint manufactured in the past (i.e., each welded joint formed by welding each of the steel plates), and also stores the HAZ toughness data of each welded joint as an evaluation index corresponding to the welding specification data.

[0021] The output unit 33 is any display, such as a liquid crystal display or an organic EL display. The output unit 33 is capable of displaying a screen based on output data and signals. The output unit 33 may also be a print output, a printer, or a plotter.

[0022] The communication unit 34 receives past performance data for each process transmitted from the process computer, etc. 2 that controls the manufacturing process, and transmits it to the main unit 35 of the device. The communication unit 34 also transmits manufacturing specification information obtained through processing in the main unit 35 to the process computer, etc. 2. Alternatively, the communication unit 34 may directly transmit and receive data from sensors, etc. that acquire data at each process without going through the process computer, etc. 2.

[0023] As shown in Figure 1, the main unit 35 of the device includes an arithmetic processing unit 351, a ROM 352, and a RAM 354. The ROM 352 stores the program 353, which will be described later. The arithmetic processing unit 351, ROM 352, and RAM 354 are connected to each other by a bus 36. The arithmetic processing unit 351 includes one or more processors, such as a general-purpose processor or a dedicated processor specialized for a specific process. The arithmetic processing unit 351 reads the program 353 from the ROM 352 and uses the RAM 354, which is a temporary storage unit, to implement specific functions, which will be described later. The arithmetic processing unit 351 controls the operation of the entire main unit 35 of the device.

[0024] The arithmetic processing unit 351 includes an information acquisition unit 3a, a HAZ toughness estimation unit 3b, a search processing unit 3c, and an output instruction unit 3d.

[0025] The information acquisition unit 3a acquires, at a minimum, the actual performance data of welding steel plates, including the steel plate component composition, steel plate manufacturing conditions, and steel plate characteristic values ​​stored in the storage unit 32, the actual welding specifications of welded joints, including welding conditions, and the actual HAZ toughness of said welded joints, as actual performance data information. Here, an example is given of using past performance data that has been previously stored in the storage unit 32, but it is not limited to this example, and it is also possible to use various types of performance data provided by manufacturers and distributors (hereinafter referred to as "vendors") that manufacture and sell steel plates and welded joints. The data provided by these vendors can be stored in the storage unit 32 and used.

[0026] The "performance data for welding steel sheets" mentioned above refers to performance data for each steel sheet manufactured in the past, and includes at least the steel sheet composition, manufacturing conditions, and characteristic values. The "steel sheet manufacturing conditions" described above include hot rolling conditions, heat treatment conditions, including slab thickness, slab heating temperature, rolling start temperature, rolling end temperature, cooling start temperature, cooling rate, cooling stop temperature, and product thickness. The "steel sheet property values" mentioned above refer to information on the mechanical properties of steel sheets obtained through mechanical property tests of steel sheets under various test conditions, test specimen sampling locations, and test specimen sampling directions. These steel sheet property values ​​include, for example, yield strength, tensile strength, yield ratio, elongation, toughness, hardness, and bendability.

[0027] Depending on the welding conditions and the location of the HAZ toughness evaluation, the microstructure and mechanical properties of the steel plate before welding may affect the microstructure and mechanical properties of the HAZ of the welded joint, and as a result, the HAZ toughness may be significantly affected. In other words, by including "steel plate manufacturing conditions" and "steel plate characteristic values" in the prediction, as in the present invention, the accuracy of predicting the HAZ toughness of the welded joint can be improved. For example, the cooling rate during steel plate manufacturing greatly affects the transformation behavior from austenite in steel, so by including this, the influence of the microstructure of the steel plate before welding can be reflected, and as a result, an improvement in the accuracy of predicting the HAZ toughness of the welded joint caused by the microstructure of the steel plate can be expected.

[0028] In addition, when obtaining steel sheet manufacturing specifications, it is possible to optionally include in the actual data information, such as the type and mixing ratio of raw materials used, adjustments to the component composition in the refining process (converter and secondary refining equipment), adjustments to the slab heating temperature, slab time in the furnace, and slab extraction temperature after casting, slab arrangement, heating gas flow rate, heating gas components, product dimensions, rolling conditions, and temperature conditions in the rolling process (rolling mill), descaling conditions, cooling adjustments in the cooling process (accelerated cooling equipment), and material test items and test results. Including one or more of these selected types of information will further improve accuracy.

[0029] The "Welding Specification Record for Welded Joints" mentioned above refers to the performance data for each welded joint previously manufactured and linked to the steel plate described above. This welding specification record includes information such as the chemical composition and mechanical properties of the welding wire and flux, welding method, groove shape, root gap, preheating temperature, current, voltage, gas flow rate, welding speed, number of passes, and post-heat treatment conditions.

[0030] The "HAZ toughness performance" mentioned above refers to performance data related to evaluation indicators for each welded joint. The HAZ toughness performance of a welded joint is the test result obtained by conducting various tests on the welded joint of a welded joint where steel plates are welded under arbitrary welding conditions. HAZ toughness is a value obtained by taking a test specimen from a location specified by, for example, a ship classification standard, and conducting tests such as a Charpy test. This evaluation indicator includes data such as the joint Charpy absorbed energy and the joint Charpy vTrs.

[0031] The information acquisition unit 3a processes when the calculation processing unit 351 receives instructions from the input unit 31 to acquire welding specification data for welded joints, data for welding steel plates, or data for HAZ toughness. However, the trigger for executing the information acquisition unit 3a processes is not limited to the input unit 31. For example, the information acquisition unit 3a processes may be triggered when the communication unit 34 receives data on each performance value from sensors or other devices that acquire data in each process, as described above.

[0032] The HAZ toughness estimation unit 3b uses past performance data, including the performance of welding steel plates, the welding specifications of welded joints made from said welding steel plates, and the HAZ toughness performance of said welded joints, as training data. Using a HAZ toughness prediction model constructed by relating this training data with the HAZ toughness performance, it performs inverse analysis based on the prediction model to calculate an estimated HAZ toughness value based on at least the steel plate characteristics and welding specifications included in the input information from the input unit 31. As described above, each piece of performance information constituting the training data is read from the storage unit 32 by the information acquisition unit 3a.

[0033] In the present invention, it is preferable that the HAZ toughness estimation unit 3b selects training data according to the welding heat history predicted from the steel plate thickness and welding specifications included in the input information.

[0034] The reason is as follows: By selecting training data according to the welding heat history, the optimal model can be used from separate prediction models that have been pre-constructed for each classification of welding heat history. This makes it possible to predict the HAZ toughness of the weld by taking into account the behavior of inclusions when subjected to welding heat, which affects the toughness of the HAZ in the weld. As a result, the prediction accuracy of HAZ toughness can be further improved. For example, by considering the melting temperatures of major inclusions that are thought to have an effect in each steel type and excluding steel types that contain many chemical components that form precipitates that melt at any given maximum temperature from the training data, the prediction accuracy can be improved.

[0035] As a specific method, for example, regarding the inclusion TiN, which is thought to significantly affect the HAZ toughness of welded joints, the solid solution temperature range of TiN is calculated using Thermo-calc, an integrated thermodynamic calculation software, based on the Ti and N addition amounts within the steel plate composition range of the training data. Then, the training data is selected into two categories: those where the maximum temperature reached during welding is within the temperature range where TiN is sufficiently dissolved, and those where TiN does not dissolve and precipitates. This improves the accuracy of predicting the HAZ toughness of the welded joint.

[0036] The phrase "according to the welding heat history predicted from the steel plate thickness and welding specifications" above refers to predicting the temperature change over time during welding and applying the optimal model according to the predicted maximum temperature reached and cooling rate.

[0037] The HAZ toughness estimation unit 3b should be able to use a HAZ toughness prediction model constructed using the aforementioned historical data as training data. Statistical methods and machine learning models such as local regression, support vector machines, neural networks, and random forests are created as the HAZ toughness prediction model.

[0038] In addition, the present invention may include a prediction model creation unit that performs the prediction model creation process described later, and may use the HAZ toughness prediction model created by this prediction model creation unit. This prediction model creation unit may be provided within the calculation processing unit 351, or it may be provided within the system 1 as a prediction model creation device independent of the main unit 35. Alternatively, a prediction model may be obtained by reading an existing prediction model provided in advance by a vendor or the like, without directly creating the prediction model.

[0039] The search processing unit 3c searches for manufacturing specifications such that the HAZ toughness estimate calculated by the HAZ toughness estimation unit 3b falls within a desired range. These manufacturing specifications refer to the manufacturing specifications for steel plates for welding and the manufacturing specifications for welded structures. In the case of manufacturing specifications for steel plates for welding, the specifications take into account that the steel plates will be welded under specified conditions afterward, and in the case of manufacturing specifications for welded structures, the specifications take into account the HAZ toughness of the welded joint after welding.

[0040] Specifically, the search processing unit 3c uses the target characteristic value of the welded joint (i.e., the desired value) input from the input unit 31 as a reference and performs feedforward or feedback calculations on the necessary control amount so that the calculated HAZ toughness estimate falls within the range of the desired value.

[0041] In this context, feedforward calculation refers to setting or modifying the manufacturing specifications of one or more subsequent processes, such as joining process s7 and post-weld heat treatment process s8, so that the estimated HAZ toughness value calculated from the input information (steel plate properties and welding specifications) for a welded joint that has progressed to the pre-treatment process s5 and preheating process s6 falls within the desired HAZ toughness range, based on the target characteristic value of the welded joint (see Figure 5(B)).

[0042] Furthermore, feedback calculation refers to using a HAZ toughness prediction model constructed with training data from welding steel plate performance, welding specifications, and HAZ toughness performance for welded joints that have already been manufactured (i.e., manufactured in the past). Based on the input steel plate characteristics and welding specifications, an estimated value of HAZ toughness is calculated, and when manufacturing the next welded joint, the steel plate manufacturing specifications are set or changed based on the target characteristic value of the welded joint (input value) so that the HAZ toughness falls within the desired range, i.e., the root mean square error (RMSE value) of the estimated value is 30 J or less.

[0043] In the present invention, it is preferable that the search processing unit 3c searches for manufacturing specifications for steel plates and welded structures that satisfy a threshold for the manufacturing cost of the steel plates and welded structures (hereinafter referred to as the "manufacturing cost threshold"), in addition to conditions such as the calculated HAZ toughness estimate being within the range of the desired target toughness value for welded joints. This has the effect of excluding manufacturing conditions that result in high costs and excessive characteristic values. The above-mentioned manufacturing cost threshold refers to the threshold at which the manufacturing cost achieves the target value.

[0044] The output instruction unit 3d displays and outputs the manufacturing specifications found by the search processing unit 3c. Upon receiving this instruction, the output unit 33 displays the manufacturing specifications received from the output instruction unit 3d. These manufacturing specifications are stored in the storage unit 32 as actual steel plate manufacturing specifications for each process and actual welded structure manufacturing specifications.

[0045] Next, the manufacturing specification determination support method of the present invention will be described. Referring to Figure 2, the information processing performed by the manufacturing specification determination support device 3 of the present invention described above will be explained. Note that explanations that overlap with those of the manufacturing specification determination support device 3 will be omitted.

[0046] Figure 2 shows an example of a flowchart of the manufacturing specification determination support method of this embodiment. As shown in Figure 2, the manufacturing specification determination support method of the present invention comprises an information acquisition step (S301) of acquiring actual data information relating to welding steel plates and welding joints from a storage unit 32, a HAZ toughness estimation step (S302) of calculating an estimated HAZ toughness value, a search processing step (S303) of searching for manufacturing specifications using the calculated estimated HAZ toughness value, and an output instruction step (S304) of giving instructions such as displaying the searched manufacturing specifications.

[0047] The processing of the manufacturing specification determination support method of the present invention is initiated by an instruction from the calculation processing unit 351, triggered by an operation of the input unit 31. Alternatively, as described above, the processing may be triggered by the reception of actual value information acquired by sensors, etc., in each process via the communication unit 34.

[0048] First, information acquisition processing is performed. The information acquisition unit 3a acquires performance data information for welded joints and welding steel plates used in previously manufactured welded joints, which has been stored in advance (step S301). This "performance data" refers to various data related to the HAZ toughness prediction model, and includes data including at least the performance of the welding steel plate, the welding specification performance of the welded joint to which the welding steel plate was welded, and the HAZ toughness performance. This performance data is stored in the storage unit 32, and the information acquisition unit 3a reads the data from there.

[0049] Next, the HAZ toughness estimation process is performed. The HAZ toughness estimation unit 3b uses the HAZ toughness prediction model constructed based on the various data acquired in step S301, performs an inverse analysis on the HAZ toughness prediction model, and estimates the HAZ toughness value of the welded joint (step S302).

[0050] Specifically, past performance data, including the performance of welding steel plates, welding specifications for welded joints, and HAZ toughness performance of said welded joints, read in step S301, is used as training data. Using a HAZ toughness prediction model constructed by relating this training data with HAZ toughness performance, an estimated HAZ toughness value of the welded joint is calculated based on the input information of steel plate characteristics and welding specifications entered in the input unit 31.

[0051] Next, a search process is performed. The search processing unit 3c searches for a manufacturing specification such that the estimated HAZ toughness value calculated in step S302 falls within the range of the desired value, based on the target characteristic value of the welded joint, i.e., the desired value (step S303).

[0052] Specifically, the search processing unit 3c uses the HAZ toughness value desired by the user as the data of the weld joint target characteristic value included in the input information received from the input unit 31, and searches for a manufacturing specification in which the absolute value of the difference between the estimated HAZ toughness value and the desired value is within a certain threshold. The phrase "the absolute value is within a certain threshold" refers to a case where the RMSE value is 30J or less. A RMSE value of 25J or less is preferable.

number

[0053] Thus, in the exploration process, an inverse analysis is performed on the HAZ toughness prediction model, and the desired manufacturing specifications (e.g., manufacturing specifications for welded structures and steel plates) are obtained based on the target characteristic values ​​of the welded joint, thereby determining the manufacturing specifications for welded structures and steel plates. In other words, by considering the behavior of inclusions when subjected to heat from welding, which affects the HAZ toughness of the weld as a metallurgical phenomenon during welding, and readjusting accordingly, the mechanical properties of the weld can be predicted with high accuracy. This results in obtaining more appropriate manufacturing specifications.

[0054] Next, output instruction processing is performed. The output instruction unit 3d gives instructions for displaying and outputting the desired manufacturing specifications found in step S303 (step S304). The output unit 33 displays the manufacturing specifications received from the output instruction unit 3d. This completes the processing of the manufacturing specification determination support method.

[0055] The desired manufacturing specifications thus discovered are transmitted to the process computer, etc. 2 via the communication unit 34. The process computer, etc. 2 controls each step of the manufacturing process based on these acquired manufacturing specifications, thereby producing welding steel plates and welded structures.

[0056] However, the present invention is not limited to this embodiment. As described later, the calculation processing unit 351 may be equipped with a prediction model creation unit, and the HAZ toughness prediction model may be created in advance by the prediction model creation unit. Alternatively, the HAZ toughness prediction model may be created in advance by a prediction model creation device independent of the main unit 35, or a HAZ toughness prediction model provided by a vendor or the like may be purchased separately. The prediction model obtained by these methods may be stored in the storage unit 32 and used in the HAZ toughness estimation step S302.

[0057] Thus, according to the present invention, steel plates can be manufactured or welded using the obtained manufacturing specifications. This makes it possible to produce welded structures that satisfy the HAZ toughness of the welded joint.

[0058] Next, with reference to Figure 3, an example of the prediction model creation unit and the prediction model creation process will be described.

[0059] The predictive model creation unit in the manufacturing specification determination support device 3 of this embodiment creates a predictive model (i.e., a HAZ toughness predictive model) that is constructed and linked using the steel plate performance data, welding specification performance data, and HAZ toughness performance data stored in the memory unit 32 as training data.

[0060] Figure 3 shows a flowchart of the predictive model creation process. When the arithmetic processing unit 351 receives an instruction to create a predictive model based on the operation of the input unit 31, it executes the information reading process (step S201), preprocessing (step S202), characteristic learning process (step S203), and result display process (step S204) shown in Figure 3. Here, a predictive model creation unit is provided within the arithmetic processing unit 351 (it is not shown in Figure 1).

[0061] As shown in Figure 3, when the calculation processing unit 351 receives an instruction to create a prediction model, the prediction model creation unit of the calculation processing unit 351 reads the welding steel plate performance data and welding specification performance data from the storage unit 32, and then reads the HAZ toughness performance data of the welded part corresponding to the read steel plate performance data and welding specification performance data from the storage unit 32. Specifically, based on the steel plate ID (identification information), it identifies the welding steel plate performance data for the target steel plate (step S201).

[0062] Next, in order to create a HAZ toughness prediction model that is built using the welding steel plate data read in step S201 as training data and linked to HAZ toughness data, this data is processed into evaluation data to be used in the prediction model creation unit (step S202). Specifically, the welding steel plate data, welding specification data, and HAZ toughness data are normalized, and noise from missing and abnormal data is removed to create evaluation data.

[0063] Next, predictive models are created using statistical and machine learning models (step S203). Specifically, the predictive model creation unit uses multiple or a single statistical method and machine learning model such as local regression, support vector machines, neural networks, and random forests, selects the combination or model with the best accuracy, and further optimizes hyperparameters for neural networks and other models that require adjustment, and performs characteristic learning.

[0064] Next, the result of creating the prediction model (i.e., the HAZ toughness prediction model) created by characteristic learning in step S203 is transmitted to the arithmetic processing unit 351 and displayed in the output unit 33 (step S204).

[0065] Here, we will refer to Figure 4 and explain in detail an example of the prediction model creation process described above.

[0066] Figure 4 is an example of a flowchart showing the process by which the manufacturing specification determination support device 3 uses a prediction model to calculate an estimated value of toughness in the HAZ of a welded joint. Figure 4 shows the processing flow when using the prediction model in this system. The prediction model provided by the manufacturing specification determination support device 3 according to this embodiment includes, in order from the input side, an input layer IL, a model M such as a mathematical model or a machine learning model, and an output layer OL.

[0067] The input layer IL stores evaluation data, including normalized steel plate properties between 0 and 1, welding specifications, toughness in the HAZ of the weld, target characteristic values ​​of the welded joint, and, if necessary, steel plate manufacturing specifications and steel plate mechanical properties. In other words, the input layer IL stores category names other than actual or numerical information, normalized between 0 and 1.

[0068] A portion of the evaluation data stored in the input layer IL can be appropriately extracted, and the extracted evaluation data can be treated as steel plate property data, welding specification data, toughness data in the HAZ of the welded joint, and, if necessary, steel plate manufacturing specification data and steel plate mechanical property data. These data can then be used to create models M, such as mathematical models and machine learning models.

[0069] Of these performance data, the toughness performance data for the HAZ of a welded joint can be stored in the input layer IL if the measurement results of the welding mechanical properties of welded joints manufactured in the past are known. On the other hand, if the measurement results of the welding mechanical properties of a welded joint are not known, the toughness in the HAZ of the welded joint can be calculated based on the steel plate properties, welding specifications, and, if necessary, the steel plate manufacturing specifications and steel plate mechanical properties, and the calculated toughness in the HAZ of the welded joint can be treated as the toughness performance data for the HAZ of the welded joint.

[0070] Furthermore, the explanatory variables for the steel plate properties, welding specifications, toughness in the HAZ of the weld, target characteristic values ​​of the welded joint, and, if necessary, the steel plate manufacturing specifications and mechanical properties of the steel plate, stored in the input layer IL, should preferably be explanatory variables related to the toughness in the HAZ of the welded joint. However, the correlation between the number of these explanatory variables and the toughness in the HAZ of the weld is arbitrary. An example of an explanatory variable related to the mechanical properties of the welded joint is the Charpy absorption energy.

[0071] The output layer OL contains an estimated value of the toughness of the weld in the HAZ of the final welded joint, calculated using a mathematical or machine learning model. Furthermore, the output layer OL also contains an estimated value of the welding efficiency of the welded joint with the weld, calculated using a mathematical or machine learning model. Through this process, based on the input steel plate properties and welding specifications, the target properties of the welded joint, and optionally the steel plate manufacturing specifications and steel plate mechanical properties, the model M is selected, and the desired welding specifications and steel plate manufacturing specifications are obtained, based on the output estimated value of the toughness in the HAZ of the weld, the measured actual toughness value in the HAZ of the weld, and the target property value of the welded joint. The process of obtaining the desired welding specifications is performed by searching so that the estimated value of the toughness in the HAZ of the weld asymptotically approaches the desired value or falls within the range of the desired value.

[0072] Furthermore, based on the estimated welding efficiency calculated and output through this process, and the measured welding efficiency, the selection of model M and the acquisition of welding specifications are performed. Here, the acquisition of welding specifications involves searching so that the estimated toughness value in the HAZ of the welded part asymptotically approaches the desired value or falls within the desired range, using the target characteristic value of the welded joint as a reference.

[0073] Furthermore, the process for acquiring steel plate manufacturing specifications searches for a solution that uses the target characteristic value of the welded joint as a reference, so that the estimated toughness value in the HAZ of the welded joint asymptotically approaches a desired value or falls within a desired range. The prediction model creation unit inputs the created evaluation data into model M and obtains the calculated toughness value in the HAZ of the welded joint for the evaluation data as an estimated value.

[0074] Next, the prediction model creation unit stores the training data, evaluation data, the type of mathematical or machine learning model and its parameters, as well as the model's output results for the training data and evaluation data, in the storage unit 32. It also transmits the training data, evaluation data, the type of mathematical or machine learning model and its parameters, as well as the model's output results for the training data and evaluation data, to the output unit 33, which then displays the results. The output unit 33 outputs, for example, the calculated toughness of the HAZ of the weld, welding efficiency, welding cost, number of welding passes, etc., in tabular format.

[0075] Next, we will describe a manufacturing line to which the manufacturing specification determination support device and manufacturing specification determination support method of the present invention can be applied. As described above, the present invention can be applied to support the creation of manufacturing specifications for welding steel plates and welded structures. Therefore, we will describe an example of applying the present invention to a system that manages a manufacturing line for welding steel plates and a manufacturing line for welded structures that are welded to the steel plates manufactured on this manufacturing line.

[0076] [First Embodiment] An example of applying the present invention to a manufacturing line for welding steel sheets will be described. Figure 5(A) shows an overall diagram of the system that manages this manufacturing line.

[0077] The manufacturing line (manufacturing process) for welding steel plates shown in Figure 5(A) includes multiple processes, such as a steelmaking process (s1), a heating process (s2), a rolling process (s3), and a cooling process (s4).

[0078] The steelmaking process (S1) is the process of producing an intermediate material called a slab by casting molten steel. The subsequent heating process (S2) is the process of heating the manufactured slab in a heating furnace. The subsequent rolling process (S3) is the process of rolling the heated slab using a rolling mill. The subsequent cooling process (S4) is the process of cooling the rolled hot-rolled sheet using an accelerated cooling device. Through these processes, a welding steel sheet (in this embodiment, a thick steel sheet; hereinafter simply referred to as "steel sheet") is manufactured as a product.

[0079] Specifically, the manufacturing specification determination support device 3 performs information processing in steps S301 to S304, thereby searching for the optimal manufacturing specifications for welding steel sheets based on the input information values. The searched manufacturing specifications for welding steel sheets are transmitted to the process computer, etc. 2 via the communication unit 34 of the manufacturing specification determination support device 3. The process computer, etc. 2 controls each process (s1 to s4) based on the manufacturing specifications for welding steel sheets to manufacture the welding steel sheets.

[0080] [Second Embodiment] An example of applying the present invention to a manufacturing line for welded structures will be described. Figure 5(B) shows an overall diagram of the system for managing this manufacturing line. This second embodiment also shows a process of joining multiple steel plates manufactured through the above-described manufacturing process for welding steel plates.

[0081] The manufacturing line (manufacturing process) for the welded structure shown in Figure 5(B) includes multiple processes, such as a pre-treatment process (s5), a preheating process (s6), a joining process (s7), and a post-weld heat treatment process (s8).

[0082] The pre-treatment step (s5) is the process of selecting the base material, which is a steel plate, and the welding material suitable for that base material. In this step, the base material is beveled, cut, or bent. The base material is cut by thermal cutting or mechanical cutting. The base material is bent by mechanical methods such as presses and rollers, or by thermal methods such as linear heating with a gas burner. The base material is beveled into V-grooves, X-grooves, U-grooves, etc., by thermal cutting or mechanical cutting. The grooves are cleaned and corrected to maintain groove accuracy and mounting accuracy. After that, the cut base material and the beveled base material are assembled. In the case of butt welding, tab plates may be attached to the base material at the start and end points of the weld.

[0083] The subsequent preheating process (s6) is a process of preheating (heating) the assembled steel plates as a measure to prevent welding cracks. Preheating the steel plates suppresses hardening of the steel plates due to rapid cooling of the heat-affected zone after the joining process described later. The preheating temperature is not simply determined by the type of base material, but is determined considering the size of the base material, plate thickness, welding method, and welding conditions. In the case of multi-layer welding, the inter-pass welding temperature is set.

[0084] The subsequent joining process (s7) is the process of joining steel plates together under predetermined welding conditions. The welding conditions are determined according to the type of steel plate, plate thickness, size of the steel plate, welding method, welding material, welding equipment, and groove shape of the steel plate. For example, when using arc welding, the above-mentioned "predetermined welding conditions" are determined by considering the welding current, arc voltage, welding speed, shielding gas flow rate, number of electrodes, current polarity, and number of welding passes, in addition to the conditions mentioned above. These welding conditions affect deformation of the welded joint, residual stress, and the occurrence of welding defects.

[0085] The subsequent post-weld heat treatment process (s8) is a process of applying heat treatment to the welded joint formed by joining steel plates. This heat treatment includes "immediate heating," which is performed immediately after welding, and "post-weld heat treatment (PWHT)," which is aimed at removing residual stress from welding. "Immediate heating" is performed in welding thick steel plates, which are prone to low-temperature cracking, to prevent phenomena caused by the rapid cooling of the weld by heating the weld and its surroundings.

[0086] As shown in Figure 5(B), the system 1 according to this embodiment includes a process computer 2 for managing the manufacturing process of steel plates for welding (s1 to s4) and the manufacturing process of welded structures formed by welding the steel plates (s5 to s8), and a steel plate manufacturing specification determination support device 3, which will be described later. Each step in the manufacturing process is controlled by this process computer, etc. 2.

[0087] In this embodiment, the manufacturing specification determination support device 3 performs information processing in steps S301 to S304, thereby searching for the optimal manufacturing specifications for welding steel plates and welded structures based on the input information values. The searched manufacturing specifications are transmitted to the process computer etc. 2 via the communication unit 34 of the manufacturing specification determination support device 3. The process computer etc. 2 controls each process (s1 to s8) based on the manufacturing specifications to manufacture the welded structure.

[0088] Furthermore, the manufacturing line for welded structures may use pre-prepared steel plates. In this case, the steel plate manufacturing specification determination support device 3 can be used to search for the manufacturing specifications of the welded structures from s5 to s8. [Examples]

[0089] The effects of the present invention will be specifically described below based on examples. However, the present invention is not limited to these embodiments.

[0090] In this embodiment, an inverse analysis was performed to obtain the manufacturing specifications for a 50mm thick steel plate (hereinafter referred to as "steel plate") subjected to electrogas arc welding (EGW) using the system. The prediction model was created by applying a machine learning method called XGBoost to the training data as pre-training. At this time, training data was selected according to the welding heat history predicted from the steel plate thickness and welding specifications, and a HAZ toughness prediction model was constructed. The training data used was the actual steel plate for welding, the actual welding specifications, and the actual HAZ toughness of the welded area. The system equipped with the above-mentioned manufacturing specification determination support device used the HAZ toughness prediction model constructed and linked using these training data as training data. Table 1 shows the variable ranges of each explanatory variable and the dependent variable.

[0091] [Table 1]

[0092] As shown in Table 1, the explanatory variables for the steel sheet composition were the component composition (C, Si, Mn, P, S, Al, Ni, Ti, N, O, Ca, Cu, Cr, Mo, Nb, V, B, Mg) and the carbon equivalent (Ceq), ACR, Pcm, and Ti / N derived from the component composition. The explanatory variables for the steel sheet manufacturing conditions were the slab thickness, slab heating temperature, rolling start temperature, rolling end temperature, cooling start temperature, cooling rate, cooling stop temperature, and product thickness. The explanatory variables for the welding specifications were the groove angle, groove length, and heat input. The explanatory variables for the steel sheet characteristic values ​​were the YS of 1 / 2t base material (where "t" is the plate thickness), TS of 1 / 2t base material, Charpy absorption energy (average of 3 measurements) and vTrs of 1 / 2t base material at a test temperature of -60°C, and joint Charpy absorption energy (average of 3 measurements) and vTrs at a test temperature of -40°C.

[0093] For each steel grade, welded joints were prepared using joint test plates taken from steel plates. The surface layer of the resulting joint was defined as the surface layer of the test specimen, and a notch was made at the fusion zone (FL) where the weld metal and base metal each comprised 50% of the material. An NK U4 impact test specimen was then taken. A Charpy impact test was performed on the collected specimens at a test temperature of -40°C. The average value of the absorbed energy of three specimens performed under the same conditions, vE-40°C (unit: J), was defined as the toughness of the HAZ.

[0094] Of the 632 samples of data after removing noise such as missing data, 442 samples were used for training, and the remaining 190 samples were used as validation data. The HAZ toughness prediction model was created using machine learning techniques with XGBoost. The hyperparameters were optimized to minimize the root mean square error (RMSE) of the validation data, and the resulting prediction model was used.

[0095] Based on the present invention, a prediction model was constructed by selecting training data into three levels based on the maximum heating temperature, corresponding to the welding heat history predicted from the steel plate thickness and welding specifications: 1350°C or more and less than 1400°C, 1400°C or more and less than 1450°C, and 1450°C or more and less than 1500°C. The results of verifying the prediction accuracy are shown in Figures 6 to 8. The maximum heating temperatures were 1350°C in Figure 6, 1400°C in Figure 7, and 1450°C in Figure 8. The RMSE of the verification data, based on the absorbed energy of the weld at -40°C, were 18.0 J (see Figure 6), 21.9 J (see Figure 7), and 19.8 J (see Figure 8), respectively, confirming that the model has sufficient prediction accuracy.

[0096] On the other hand, Figure 9 shows the results of verifying the prediction accuracy by constructing a model without selecting training data as in the conventional method. "As in the conventional method" refers to the conventional method described above, which uses only the component composition of the welding material as input values. The RMSE of the verification data was 41.3 J for the absorbed energy of the weld at -40°C, confirming that the prediction accuracy was low.

[0097] Applying the present invention, a predictive model suitable for the plate thickness and welding specifications was used to search for welding specifications that would increase the Charpy absorption energy at the weld (on the melting line) after the preheating process s6. The searched welding specifications were feedforward controlled and reflected in the welding specifications. The target Charpy absorption energy was set to 64 J or higher, and steel plates were manufactured with a steel plate composition and manufacturing specifications that achieved the target Charpy absorption energy. As a result of evaluating the HAZ toughness, the desired toughness range of RMSE ≤ 30 J was achieved. From this, it was found that metallurgical phenomena (particularly the behavior of inclusions due to welding heat that affects HAZ toughness) could be taken into consideration. Furthermore, it was possible to explore the conditions that would minimize manufacturing costs based on the achieved steel plate manufacturing specifications and welding specifications.

[0098] Thus, by applying the present invention, it was confirmed that when manufacturing steel plate welded joints, it is possible to reverse-search based on a predictive model to find welding specifications and steel plate manufacturing specifications that satisfy a predetermined HAZ toughness, and to manufacture welded joints and steel plates based on welding specifications and steel plate manufacturing specifications that enable efficient welding.

[0099] On the other hand, using a predictive model constructed with conventional methods, and similarly setting the target Charpy absorption energy to 64 J or higher, steel plates were manufactured with steel plate component composition and manufacturing specifications that achieved the target Charpy absorption energy. The results showed that less than 50% of the conditions for achieving the joint Charpy absorption energy within the desired toughness range were met. This indicates that the metallurgical phenomena were not considered accurately, and the model construction was not appropriate. Furthermore, the conditions under which the target was achieved resulted in excessively high manufacturing costs. [Explanation of Symbols]

[0100] 1 System 2. Process computer or distributed control system 3. Manufacturing Specification Determination Support Device 3a Information acquisition section 3b HAZ toughness estimation part 3c Search Processing Unit 3d output instruction section 31 Input section 32 Storage section 33 Output section 34 Communications Department 35 Main unit of the device 351 Arithmetic Processing Unit 352 ROM 353 Programs 354 RAM 36 bus

Claims

1. An information acquisition unit acquires, as performance data information, the performance of welding steel plates including steel plate component composition, steel plate manufacturing conditions, and steel plate characteristic values, the performance of welding specifications for welded joints including welding conditions, and the performance of HAZ toughness of said welded joints. A HAZ toughness estimation unit calculates an estimated HAZ toughness value based on the steel plate characteristics and welding specifications included in the input information, using a HAZ toughness prediction model constructed by associating the training data with the HAZ toughness results, which includes past performance data including the performance of the steel plate for welding, the welding specifications of the welded joint, and the HAZ toughness results as training data. A search processing unit that searches for manufacturing specifications such that the calculated HAZ toughness estimate falls within a desired range, An output instruction unit that provides instructions for displaying and outputting the searched manufacturing specifications, A manufacturing specification determination support device characterized by comprising the following features.

2. The manufacturing specification determination support device according to claim 1, characterized in that the HAZ toughness estimation unit selects the training data according to the steel plate thickness and welding heat history predicted from the welding specifications included in the input information.

3. The manufacturing specification determination support device according to claim 1, characterized in that the search processing unit further searches for the manufacturing specifications such that the calculated HAZ toughness estimate satisfies the manufacturing cost threshold.

4. The manufacturing specification determination support device according to claim 2, characterized in that the search processing unit further searches for the manufacturing specifications such that the calculated HAZ toughness estimate satisfies the manufacturing cost threshold.

5. An information acquisition step to acquire, as performance data information, the performance of welding steel plates including steel plate component composition, steel plate manufacturing conditions, and steel plate characteristic values, the performance of welding specifications for welded joints including welding conditions, and the performance of HAZ toughness of said welded joints. A HAZ toughness estimation step involves using past performance data, including the performance of the steel plate for welding, the welding specifications of the welded joint, and the HAZ toughness performance, as training data, and using a HAZ toughness prediction model constructed by relating the training data with the HAZ toughness performance to calculate an estimated HAZ toughness value based on the steel plate characteristics and welding specifications included in the input information. A search process step for searching for manufacturing specifications such that the calculated HAZ toughness estimate falls within a desired range, An output instruction step that provides instructions for displaying and outputting the searched manufacturing specifications, A method for supporting the determination of manufacturing specifications, characterized by comprising the following features.

6. The manufacturing specification determination support method according to claim 5, characterized in that the HAZ toughness estimation step involves selecting the training data according to the steel plate thickness and welding heat history predicted from the welding specifications included in the input information.

7. The manufacturing specification determination support method according to claim 5 or 6, characterized in that the search processing step further searches for the manufacturing specifications such that the calculated HAZ toughness estimate satisfies a manufacturing cost threshold.

8. A method for manufacturing a steel sheet, characterized by comprising a step of manufacturing a steel sheet for welding based on the manufacturing specifications of the steel sheet for welding, which have been searched by a manufacturing specification determination support device according to any one of claims 1 to 4.

9. A method for manufacturing a welded structure, characterized by comprising a step of manufacturing the welded structure based on the manufacturing specifications of the welded joint, which have been searched by the manufacturing specification determination support device described in any one of claims 1 to 4.

10. A program used to determine manufacturing specifications, An information acquisition process that obtains, as performance data information, the performance of welding steel plates including steel plate component composition, steel plate manufacturing conditions, and steel plate characteristic values, the performance of welding specifications for welded joints including welding conditions, and the performance of HAZ toughness of said welded joints. A HAZ toughness estimation process calculates an estimated HAZ toughness value based on the steel plate characteristics and welding specifications included in the input information, using a HAZ toughness prediction model constructed by associating the training data with the HAZ toughness results, which includes past performance data including the performance of the steel plate for welding, the welding specifications of the welded joint, and the HAZ toughness results, as training data. A search process to find manufacturing specifications such that the calculated HAZ toughness estimate falls within a desired range, An output instruction process that provides instructions for displaying and outputting the searched manufacturing specifications, A program characterized by causing a computer to execute it.

Images