Method for predicting welded part properties, method for generating prediction model, and prediction model

By using prediction models to analyze electrical information during resistance spot welding, the method addresses inefficiencies in selecting welding conditions for multi-layer metal plate assemblies, enhancing the accuracy and efficiency of nugget diameter and joint strength prediction.

JP7578213B1Active Publication Date: 2024-11-06JFE STEEL CORP
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
JP2024534780
Authority / Receiving Office
JP · JP
Patent Type
Patents
Current Assignee / Owner
Priority Date
2023-02-28
Filing Date
2024-02-27
Publication Date
2024-11-06
Estimated Expiration
2044-02-27

Smart Images

  • Figure 0007578213000001
    Figure 0007578213000001
  • Figure 0007578213000002
    Figure 0007578213000002
  • Figure 0007578213000003
    Figure 0007578213000003
Patent Text Reader

Abstract

The weld characteristic prediction method is a method for predicting the characteristics of a weld 5 obtained by resistance spot welding in which an electric current is passed between two electrodes 3, 4 that sandwich a plate set of three or more overlapping metal plates, and includes the steps of acquiring electrical information output during resistance spot welding, inputting the electrical information into a first prediction model 221 that predicts the characteristics of a first weld 51 and acquiring a predicted result of the characteristics of the first weld 51 output from the first prediction model 221, and inputting the electrical information into a second prediction model 222 that predicts the characteristics of a second weld 52 and acquiring a predicted result of the characteristics of the second weld 52 output from the second prediction model 222.
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present disclosure relates to a method for predicting properties of a weld obtained by resistance spot welding, a method for generating a predictive model of the properties of a weld, and a predictive model of the properties of a weld. [Background technology]

[0002] Resistance spot welding is performed to assemble the bodies of automobiles and the like. Resistance spot welding is performed at several thousand locations on one vehicle body. Resistance spot welding is a welding method in which two or more steel sheets are stacked together, and a nugget of a predetermined size is formed between the steel sheets by clamping them between a pair of upper and lower welding electrodes and applying pressure while passing an electric current between the welding electrodes, thereby obtaining a welded joint. The joint strength of a welded portion by resistance spot welding is determined according to the nugget diameter. In welding that requires high joint strength, such as welding of automobile parts, it is important to form a nugget of a predetermined diameter or more between the steel sheets. A manufacturing method for a spot welded joint having a nugget diameter of a required size is known (see, for example, Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2005-262259 A Summary of the Invention [Problem to be solved by the invention]

[0004] In order to select the conditions for resistance spot welding that can produce a spot welded joint with a required nugget diameter, an experiment is performed under various conditions to weld steel plates, and the cross section of the welded joint obtained by the experiment is observed to measure the nugget diameter. The more the number of conditions for the experiment increases, the more the number of welded joints for which the nugget diameter is measured. The more time is spent on the measurement of the nugget diameter of the welded joint, the lower the efficiency of the selection of the conditions for resistance spot welding. In particular, it is known that it is difficult to form a nugget between the thinnest metal plate and the thickest metal plate in a plate assembly in which three or more metal plates are overlapped and the plate thickness ratio, which is the value obtained by dividing the total plate thickness of the plate assembly by the plate thickness of the thinnest metal plate among the metal plates, is 5 or more. In Patent Document 1, a welding method that stabilizes nugget formation by multi-stage current application and pressure application is developed, but the efficiency of the selection of the welding conditions decreases due to the increase in welding parameters. Therefore, it is required to further improve the efficiency of the selection of the conditions for resistance spot welding in a plate assembly in which the plate thickness ratio of three or more metal plates overlapping each other is 5 or more.

[0005] In view of the above, an object of the present disclosure is to provide a weld property prediction method, a prediction model generation method, and a prediction model that enable efficient selection of resistance spot welding conditions. [Means for solving the problem]

[0006] In order to achieve the above object, a method for determining welding conditions for resistance spot welding according to the present disclosure is as follows.

[0007] [1] A method for predicting the characteristics of a welded portion of a plate assembly including three or more overlapping metal plates, the plate assembly including at least a first metal plate, a second metal plate, and a third metal plate, obtained by resistance spot welding in which a current is passed between two electrodes that sandwich the plate assembly, the method comprising: acquiring electrical information output during resistance spot welding, the electrical information including at least one of a current flowing between the two electrodes, a voltage between the two electrodes, a resistance between the two electrodes, or a heat generation amount between the two electrodes; inputting the electrical information into a first prediction model that predicts characteristics of a first welded portion between the first metal plate and the second metal plate, and acquiring a prediction result of the characteristics of the first welded portion output from the first prediction model; inputting the electrical information into a second prediction model that predicts characteristics of a second weld between the second metal plate and the third metal plate, and acquiring a prediction result of the characteristics of the second weld output from the second prediction model; A method for predicting weld characteristics, comprising:

[0008] [2] The plate thickness ratio, which is the value obtained by dividing the total plate thickness of the plate group by the plate thickness of the thinnest metal plate in the plate group, is 5 or more, The first metal plate is thinner than the second metal plate. The method for predicting weld properties described in [1] above.

[0009] [3] The method for predicting weld properties described in [2] above, wherein the first prediction model is a nonlinear model.

[0010] [4] A method for predicting weld properties described in any one of [1] to [3] above, further comprising a step of generating the first prediction model and the second prediction model by performing learning using the electrical information as learning data.

[0011] [5] The method for predicting weld properties described in any one of [1] to [4] above, further comprising a step of obtaining a predicted result of at least one of the nugget diameter, tensile shear strength, cross tensile strength, or fracture strength by mechanical peel test of the weld property of the weld.

[0012] [6] The method for predicting welded portion characteristics described in any one of [1] to [5] above, wherein in the step of acquiring the electrical information, a total amount of heat generated from the start of current flow to the end of current flow in resistance spot welding is acquired as the electrical information.

[0013] [7] The method for predicting weld characteristics described in [6] above, wherein in the step of acquiring electrical information, the resistance between the two electrodes at the end of current flow is acquired as the electrical information.

[0014] [8] The resistance spot welding includes a first stage of passing a first current and a second stage of passing a second current; The method for predicting weld characteristics according to the above [6] or [7], wherein in the step of acquiring electrical information output during the resistance spot welding, the amount of heat generated during the first stage of the period from the start of current flow to the end of current flow is acquired as the electrical information.

[0015] [9] A method for generating a model predicting characteristics of a weld, the weld including a first weld between the first metal plate and the second metal plate and a second weld between the second metal plate and the third metal plate, obtained by resistance spot welding in which three or more metal plates including at least a first metal plate, a second metal plate, and a third metal plate are overlapped and a current is passed between two electrodes that sandwich the three or more metal plates, the method comprising: acquiring electrical information output during resistance spot welding, the electrical information including at least one of a current flowing between the two electrodes, a voltage between the two electrodes, a resistance between the two electrodes, or a heat generation amount between the two electrodes; obtaining an analysis result of the characteristics of the weld when the resistance spot welding is performed; executing machine learning based on learning data that associates the electrical information with an analysis result of the characteristics of the welded portion when the resistance spot welding is performed, to generate a first prediction model that predicts characteristics of the first welded portion and a second prediction model that predicts characteristics of the second welded portion; generating a prediction model including the first prediction model and the second prediction model; A method for generating a predictive model, comprising:

[0016]

[10] A predictive model for predicting the properties of a weld, generated by the predictive model generation method described in [9] above. Effect of the Invention

[0017] According to the present disclosure, a method for predicting weld characteristics, a method for generating a prediction model, and a prediction model that enable efficient selection of resistance spot welding conditions are provided. [Brief description of the drawings]

[0018] [Figure 1] FIG. 1 is a schematic diagram illustrating an overview of resistance spot welding. [Diagram 2] FIG. 1 is a block diagram showing a configuration example of a prediction system according to the present disclosure. [Diagram 3] FIG. 2 is a block diagram showing an example of the configuration of a prediction model. [Figure 4A] 1 is a graph showing an example of a change over time in inter-electrode resistance value in resistance spot welding. [Figure 4B] 1 is a graph showing an example of a change over time in accumulated heat generation amount in resistance spot welding. [Diagram 5] 1 is a graph showing an example of a correlation between total heat generation amount and nugget diameter. [Figure 6] 1 is a graph showing an example of a correlation between an initial heat generation amount and a nugget diameter. [Figure 7] 13 is a graph showing an example of a pattern in which a current flowing between electrodes is divided into two stages. [Figure 8A] 1 is a cross-sectional view showing an example of a welded portion when a first-stage current is passed through a plate assembly in which three metal plates are overlapped. FIG. [Figure 8B] 1 is a cross-sectional view showing an example of a welded portion when a second-stage current is passed through a plate assembly in which three metal plates are stacked. FIG. [Figure 9] 1 is a flowchart illustrating an example of a procedure for a prediction model generation method. [Figure 10]1 is a flowchart showing an example of a procedure of a welded portion characteristic prediction method. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0019] Hereinafter, embodiments of the weld property prediction method, prediction model generation method, and prediction model according to the present disclosure will be described with reference to the drawings. Each drawing is schematic and may differ from the actual one. In addition, the following embodiments are illustrative of an apparatus or method for embodying the technical idea of ​​the present disclosure, and are not intended to specify the configuration as described below. In other words, the technical idea of ​​the present disclosure can be modified in various ways within the technical scope described in the claims.

[0020] By carrying out the weld property prediction method according to the present disclosure, the properties of the weld obtained by resistance spot welding are predicted. Resistance spot welding, which is performed to obtain the weld that is the prediction target of the method according to the present disclosure, is a method in which a plate set in which an upper steel plate 6, a middle steel plate 7, and a lower steel plate 8 are overlapped is sandwiched between a pair of electrodes 3 and 4 and pressed while a current is passed between the electrodes 3 and 4 to join them, as shown in FIG. 1. The upper steel plate 6 is also referred to as the first metal plate. The middle steel plate 7 is also referred to as the second metal plate. The lower steel plate 8 is also referred to as the third metal plate. In the plate set illustrated in FIG. 1, the upper steel plate 6 and the middle steel plate 7 come into contact with each other, and the middle steel plate 7 and the lower steel plate 8 come into contact with each other. By passing a current through the sheet pair between the electrode 3 and the electrode 4, the temperature of the sheet pair rises due to heat generation caused by the electrical resistance of each of the upper steel plate 6, the middle steel plate 7, and the lower steel plate 8, as well as heat generation caused by the contact resistance between the upper steel plate 6 and the middle steel plate 7 and the contact resistance between the middle steel plate 7 and the lower steel plate 8. When a current is passed between the electrode 3 and the electrode 4, the high-temperature parts of the sheet pair melt and solidify after the current is stopped, forming a weld 5 that joins the upper steel plate 6, the middle steel plate 7, and the lower steel plate 8 that make up the sheet pair. The shape of the weld 5 can be a disk shape with the line connecting the electrode 3 and the electrode 4 as the axis of rotation. The weld 5 is also called a "nugget."

[0021] The joint strength of the welded portion 5 obtained by performing resistance spot welding is determined by the area of ​​the welded portion 5 joining the upper steel plate 6 and the middle steel plate 7, and the area of ​​the welded portion 5 joining the middle steel plate 7 and the lower steel plate 8. In other words, the joint strength of the welded portion 5 is determined by the size of the welded portion 5 in the plan view of the steel plates constituting the plate assembly. The size of the welded portion 5 in the plan view of the steel plates constituting the plate assembly may be expressed as the diameter of the disk-shaped welded portion 5. The diameter of the disk-shaped welded portion 5 is also called the nugget diameter. As shown in FIG. 1, the nugget diameter of the welded portion 5 is measured separately as a nugget diameter d1 at the contact surface between the upper steel plate 6 and the middle steel plate 7, and a nugget diameter d2 at the contact surface between the middle steel plate 7 and the lower steel plate 8. When the welded portion 5 is formed in a disk shape with the normal line of each steel plate constituting the plate assembly as the rotation axis, the nugget diameters d1 and d2 are measured at the cross section of the plate assembly drawn to pass through the rotation axis. When high joining strength is required for automobile parts and the like, it is important that the welded portion 5 is formed so that the nugget diameter is equal to or greater than a predetermined diameter.

[0022] The welded portion 5 is located inside a plate assembly including an upper steel plate 6, a middle steel plate 7, and a lower steel plate 8. In order to confirm that the nugget diameter of the welded portion 5 is equal to or larger than a predetermined diameter, it is necessary to conduct a test in which the plate assembly is actually welded, take a sample from the welded joint obtained in the test, and measure the nugget diameter by cross-sectional observation. In order to determine the welding conditions under which the nugget diameter of the welded portion 5 is equal to or larger than a predetermined diameter, it is necessary to repeat the test and cross-sectional observation while changing the welding conditions. However, a great deal of man-hours and costs are consumed in the cross-sectional observation.

[0023] The joint strength of the welded portion 5 is expressed as the tensile shear strength or cross tensile strength of the welded portion 5. When high joint strength is required for automobile parts, etc., it is important that the welded portion 5 is formed so that the tensile shear strength or cross tensile strength of the welded portion 5 is increased. In order to confirm the tensile shear strength or cross tensile strength of the welded portion 5, it is necessary to conduct a test in which the plate assembly is actually welded, and a fracture test in which a load is applied to the welded joint obtained in the test. In order to determine the welding conditions under which the joint strength of the welded portion 5 is the required strength, it is necessary to repeat the test and the fracture test while changing the welding conditions. However, a great deal of man-hours and costs are consumed for the fracture test.

[0024] Therefore, in order to reduce the man-hours and costs required for determining welding conditions and to make the selection of welding conditions more efficient, it is required to predict the nugget diameter of the welded portion 5 without performing cross-sectional observation and to predict the strength of the welded portion 5 without performing a fracture test. According to the welded portion characteristic prediction method according to the present disclosure, the characteristics of the welded portion 5, including the nugget diameter or strength of the welded portion 5, are predicted without performing cross-sectional observation or fracture test of a welded joint obtained by testing. As a result, the selection of welding conditions is made more efficient.

[0025] (Example of configuration of prediction system 100) As shown in FIG. 2, a prediction system 100 according to an embodiment of the present disclosure includes a prediction model generation device 10, a prediction device 20, a welding device 30, and an analysis device 40.

[0026] The welding device 30 performs resistance spot welding to join steel plates by sandwiching a plate set of three or more overlapping metal plates, such as an upper steel plate 6, a middle steel plate 7, and a lower steel plate 8 in FIG. 1, between two electrodes 3 and 4 (see FIG. 1) and applying pressure while passing current between the two electrodes 3 and 4. The welding device 30 is configured to be capable of controlling the pressure and the welding current flowing between the electrodes during welding. The welding device 30 may be of various types, such as a stationary type or a robot gun. The power source of the welding device may be either a DC power source or an AC power source. When the power source is an AC power source, a process is performed in which the calculation at the point where V=0 is ignored or a moving average is obtained so that the resistance does not become infinite, so that the resistance does not become zero. The shape of the electrodes 3 and 4 of the welding device 30 may be a shape having a curved tip as exemplified in FIG. 1, or may be of various other shapes. The welding device 30 outputs electrical information during resistance spot welding. The electrical information output from the welding device 30 during resistance spot welding includes at least one of the current flowing between the two electrodes when performing resistance spot welding, the voltage between the two electrodes when performing resistance spot welding, the resistance between the two electrodes when performing resistance spot welding, or the amount of heat generated between the two electrodes when performing resistance spot welding. The electrical information output during resistance spot welding is hereinafter also referred to as electrical information during resistance spot welding.

[0027] The plate assembly is formed by stacking three or more steel plates. In particular, when selecting welding conditions for a plate assembly in which one or more of the three or more steel plates constituting the plate assembly are high-tensile steel plates of 590 MPa class or more and the plate thickness ratio, which is the value obtained by dividing the total plate thickness of the plate assembly by the plate thickness of the thinnest metal plate, is 5 or more, it takes a lot of time to check the state of welding by analysis. Therefore, by selecting welding conditions for a plate assembly including high-tensile steel plates and having a plate thickness ratio of 5 or more by the method of the present disclosure, the effect of the present disclosure of improving the efficiency of the selection of welding conditions is greater. The number of steel plates constituting the plate assembly is not limited to three as exemplified in FIG. 1, but may be four or more. The plate assembly may be formed by stacking other types of metal plates, not limited to steel plates. At least one metal plate among the metal plates constituting the plate assembly may be a steel plate having a plating layer. Two or more metal plates may be steel plates having a plating layer. The steel plate having a plating layer may have a plating layer only on one side or on both sides. The steel sheet having a plating layer may be, for example, a zinc-based plated steel sheet.

[0028] The analysis device 40 analyzes the characteristics of the welded portion of the plate assembly obtained by performing resistance spot welding with the welding device 30, and outputs the analysis results of the welded portion characteristics. The analysis device 40 may include a cross-sectional observation device or a non-destructive inspection device that measures the nugget diameter of the welded portion. The analysis device 40 may include a fracture test device that measures the bonding strength of the welded portion. The welded portion characteristics include at least one of the nugget diameter, tensile shear strength, and cross tensile strength of the welded portion.

[0029] The prediction model generating device 10 includes a data acquiring unit 12, a database 14, and a model generating unit 16. The data acquiring unit 12 acquires electrical information during resistance spot welding in a test in which resistance spot welding is actually performed on a plate assembly of two or more overlapping metal plates to weld the plate assembly from the welding device 30. The data acquiring unit 12 acquires an analysis result of the welded portion characteristics of a sample of a welded joint obtained in the welding test from the analysis device 40. The data acquiring unit 12 stores data in which the electrical information during resistance spot welding in the welding test and the analysis result of the welded portion characteristics of the sample of a welded joint obtained in the welding test are associated with each other in the database 14. The model generating unit 16 performs learning using the data stored in the database 14 as learning data. The model generating unit 16 performs learning to generate a model that predicts the welded portion characteristics. The model that predicts the welded portion characteristics is configured to output a prediction result of the welded portion characteristics when electrical information during resistance spot welding is input and resistance spot welding is performed with the electrical information. The model generating unit 16 outputs the generated model to the prediction device 20.

[0030] Prediction device 20 uses the model acquired from prediction model generation device 10 as prediction model 22. Prediction device 20 acquires electrical information during resistance spot welding when a welding test is performed by welding device 30, and inputs the information to prediction model 22. Prediction device 20 acquires the prediction results of the welded portion characteristics output from prediction model 22, and outputs the results to the outside.

[0031] The predictive model generating device 10, the prediction device 20, the welding device 30, or the analysis device 40 are connected to each other so as to be able to communicate with each other. The predictive model generating device 10, the prediction device 20, the welding device 30, or the analysis device 40 may have a communication interface based on a wired or wireless communication standard. For example, the wireless communication standard may include a cellular phone communication standard such as 3G, 4G, or 5G. Also, for example, the wireless communication standard may include IEEE802.11 or Bluetooth (registered trademark). The communication interface may support one or more of these communication standards. The communication interface is not limited to these examples and may communicate with other devices or input and output data based on various standards. The predictive model generating device 10, the prediction device 20, the welding device 30, or the analysis device 40 may be connected via a network or directly (for example, peer-to-peer (P2P)).

[0032] The prediction device 20 may be communicatively connected to an external device so as to transmit the prediction result of the welded portion characteristics to the external device. The prediction device 20 may include an output device that outputs the prediction result of the welded portion characteristics. The prediction device 20 may include a display device for displaying the prediction result of the welded portion characteristics as the output device. The display device may include, for example, an LCD (Liquid Crystal Display), an organic EL (Electro-Luminescence) display or an inorganic EL display, or a PDP (Plasma Display Panel). The display device is not limited to these displays and may include various other types of displays.

[0033] The prediction model generation device 10 or the prediction device 20 may be configured to include at least one processor, such as a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit). The prediction model generation device 10 or the prediction device 20 may be configured with one processor or multiple processors. The processor constituting the prediction model generation device 10 or the prediction device 20 may realize the functions of the prediction model generation device 10 or the prediction device 20 by reading and executing a program stored in a storage unit described later.

[0034] The prediction model generating device 10 or the prediction device 20 may include a storage unit. The storage unit stores various information or data. The storage unit may store, for example, a program executed in the prediction model generating device 10 or the prediction device 20, or data or a processing result used in the processing executed in the prediction model generating device 10 or the prediction device 20. The storage unit may also function as a work memory for the prediction model generating device 10 or the prediction device 20. The storage unit may include, but is not limited to, a semiconductor memory or the like. For example, the storage unit may be configured as an internal memory of a processor used as the prediction model generating device 10 or the prediction device 20, or may be configured as a hard disk drive (HDD) accessible from the prediction model generating device 10 or the prediction device 20. The storage unit may be configured as a non-transient readable medium. The storage unit may be configured integrally with the prediction model generating device 10 or the prediction device 20, or may be configured separately from the prediction model generating device 10 or the prediction device 20.

[0035] The prediction system 100 may not include the prediction model generation device 10. When the prediction system 100 does not include the prediction model generation device 10, the prediction device 20 acquires the prediction model 22 from an external device. The prediction system 100 may not include the welding device 30. When the prediction system 100 does not include the welding device 30, the prediction model generation device 10 or the prediction device 20 acquires electrical information during resistance spot welding in a welding experiment from an external device that performs welding. The prediction system 100 may not include the analysis device 40. When the prediction system 100 does not include the analysis device 40, the prediction model generation device 10 acquires an analysis result of the welded portion characteristics of a sample of a welded joint obtained in a welding experiment from an external device that performs analysis.

[0036] As shown in FIG. 3, the prediction model 22 may include a first prediction model 221 and a second prediction model 222. The first prediction model 221 is configured as a model for predicting the characteristics of a first welded portion 51 (see FIG. 8A) located between an upper steel plate 6 and a middle steel plate 7. The second prediction model 222 is configured as a model for predicting the characteristics of a second welded portion 52 (see FIG. 8B) located between a middle steel plate 7 and a lower steel plate 8. The prediction model 22 may predict the characteristics of the first welded portion 51 and the characteristics of the second welded portion 52 separately using the first prediction model 221 and the second prediction model 222, and output the prediction results of the welded portion characteristics as a prediction result by combining the respective prediction results. The number of models constituting the prediction model 22 is not limited to two as illustrated in FIG. 3. For example, the prediction model 22 applied to a plate assembly in which four steel plates are overlapped may be configured as a model divided into three. The number of models constituting the prediction model 22 to be applied to a plate assembly in which three or more steel plates are overlapped is not limited to the number of steel plates minus one, but may be the number of steel plates minus two or more.

[0037] (Example of operation of prediction system 100) An example of the operation of the prediction system 100 will now be described in detail.

[0038] The data acquisition unit 12 of the prediction model generation device 10 acquires electrical information during resistance spot welding performed in a welding experiment from the welding device 30. The data acquisition unit 12 acquires from the analysis device 40 the analysis results of the welded portion characteristics of the welded joint samples obtained in the welding experiment.

[0039] The data acquiring unit 12 selects one or more pieces of performance data from the electrical information during resistance spot welding and stores the data in the database 14. The performance data of the electrical information during resistance spot welding includes at least one of the current flowing between two electrodes or the voltage applied between two electrodes when resistance spot welding is being performed, the resistance between two electrodes when resistance spot welding is being performed, or the amount of heat generated between two electrodes when resistance spot welding is being performed.

[0040] The data acquisition unit 12 selects one or more pieces of performance data from the analysis results of the welded portion characteristics and stores them in the database 14. The performance data of the welded portion characteristics includes at least one of the nugget diameter, tensile shear strength, and cross tensile strength of the welded portion 5 (see FIG. 1). The analysis results of the welded portion characteristics are information obtained by offline testing in the analysis device 40.

[0041] The data acquisition unit 12 stores the result data of the electrical information during resistance spot welding together with the associated information such as the welding control number in the database 14 so that the result data of the electrical information and the welded portion characteristics during resistance spot welding can be associated with each other. The result data of the electrical information during resistance spot welding and the result data of the welded portion characteristics are associated with each other by the associated information such as the welding control number and are stored in the database 14 as a set of data sets. One data set corresponding to one welding condition of a welding experiment is stored in the database 14. The number of data sets used to generate the prediction model 22 is preferably 100 or more, more preferably 200 or more. The number of data sets is further preferably 300 or more.

[0042] The model generation unit 16 generates a trained prediction model 22 by executing machine learning using one or more performance data selected from the electrical information during resistance spot welding among the data sets stored in the database 14 as input performance data. The machine learning method executed by the model generation unit 16 may be a known learning method. The machine learning method may be any method as long as the trained prediction model 22 can obtain sufficient prediction accuracy of the welded portion characteristics for practical use. For example, a known machine learning method using a neural network including deep learning, a convolutional neural network (CNN), or a recurrent neural network (RNN) may be used. As other methods, regression tree learning, support vector regression, Gaussian process, or the like may be used. Also, an ensemble model combining multiple models may be used. Also, the prediction model 22 may be appropriately updated by relearning using the latest learning data after being generated by executing machine learning.

[0043] <Actual data on electrical information during resistance spot welding> As described above, in the prediction system 100 according to the present embodiment, one or more pieces of actual data are selected from the electrical information during resistance spot welding and used to predict the welded portion characteristics. The one or more pieces of actual data selected from the electrical information during resistance spot welding include, for example, the welding current value I, the interelectrode voltage V, the interelectrode resistance R, or the heat generation amount Q. The welding current value I is the value of a current flowing between the two electrodes 3 and 4 (see FIG. 1) during the execution of resistance spot welding. The interelectrode voltage V is the voltage applied between the two electrodes 3 and 4 during the execution of resistance spot welding. The interelectrode resistance R is the electrical resistance value between the two electrodes 3 and 4, and is calculated by R=V / I using the interelectrode voltage V and the welding current value I. The heat generation amount Q is the amount of heat generated between the two electrodes 3 and 4, and is calculated by Q=(I·R)^2 / (r·t^2) using the welding current value I, the interelectrode resistance R, and the resistivity r and sheet thickness t of the metal sheets constituting the sheet assembly. When the metal plate is a steel plate, the resistivity r is the specific resistance value of pure iron at 1200°C.

[0044] During resistance spot welding, the resistance of the metal plate changes as the metal plate melts. As a result, as illustrated in FIG. 4A, the inter-electrode resistance R changes over time during resistance spot welding. Also, as illustrated in FIG. 4B, the accumulated value of the heat generation increases over time during resistance spot welding. In the graphs of FIG. 4A and FIG. 4B, the horizontal axis represents the welding time of the resistance spot welding. The vertical axis of the graph of FIG. 4A represents the inter-electrode resistance R. The vertical axis of the graph of FIG. 4B represents the accumulated heat generation value, which is the accumulated value of the heat generation value between the electrodes.

[0045] The symbols indicating specific times on the horizontal axis of FIGS. 4A and 4B are defined as follows. Tm: The time when the interelectrode resistance R reaches its maximum value (Rmax) Ts: A predetermined time has elapsed since the start of current flow in resistance spot welding Tf: Time when current flow for resistance spot welding ends Tml: The intermediate time between the time (Tm) when the interelectrode resistance R reaches its maximum value (Rmax) and the time (Tf) when the current flow ends, calculated as Tm + (Tf - Tm) / 2

[0046] The symbols indicating particular resistance values ​​on the vertical axis of FIG. 4A are defined as follows: Rfirst: The value of the interelectrode resistance R when a certain time (Ts) has elapsed since the start of current flow in resistance spot welding Rmax: Maximum interelectrode resistance R Rfm: The value of the interelectrode resistance R at half the time (Tm) when the interelectrode resistance R reaches its maximum value (Rmax) Rlast: The value of the interelectrode resistance R at the end of resistance spot welding current Rml: The value of the interelectrode resistance R at the midpoint between the time (Tm) when the interelectrode resistance R reaches its maximum value (Rmax) and the time when the current flow ends (Tf)

[0047] The symbols indicating specific heat values ​​on the vertical axis of FIG. 4B are defined as follows: Qm: Cumulative heat generation up to the time (Tm) when the interelectrode resistance R reaches its maximum value (Rmax) Qi: The cumulative heat generated from the start of current flow during resistance spot welding until a specified time (Ts) has elapsed Qt: Cumulative heat generation from the start of current application to the end of current application in resistance spot welding Qi is also called the initial heat generation amount. Qt is also called the total heat generation amount.

[0048] The actual data of the electrical information during resistance spot welding is specified by the values ​​indicated by the above-mentioned symbols. That is, the actual data of the electrical information during resistance spot welding includes the values ​​indicated by the above-mentioned symbols. It has been found that, among the actual data of the electrical information during resistance spot welding, the total heat generation amount Qt has a large effect on the welded portion characteristics. For example, as shown in FIG. 5, the total heat generation amount Qt correlates with the nugget diameters d1 and d2 (see FIG. 1). The horizontal axis of the graph in FIG. 5 represents the total heat generation amount Qt. The vertical axis represents the nugget diameter. Based on the correlation between the total heat generation amount Qt and the nugget diameter, the total heat generation amount Qt may be adopted as the actual data used to generate the prediction model 22 among the actual data of the electrical information during resistance spot welding and the actual data input to the prediction model 22 to predict the welded portion characteristics.

[0049] It has also been found that the initial heat generation amount Qi has a large effect on the nugget growth process at the beginning of current application in resistance spot welding. In other words, it has been found that the initial heat generation amount Qi among the actual data of electrical information during resistance spot welding has a large effect on the welded portion characteristics. The initial heat generation amount Qi is the amount of heat generated during a part of the period from the start of current application to the end of current application in resistance spot welding. The predetermined time Ts defining the initial heat generation amount Qi may be set to, for example, 200 ms (milliseconds). The initial heat generation amount Qi may be calculated by excluding the amount of heat generated during a very short period from the start of current application in resistance spot welding to the end of current application, for example, 10 ms. For example, as shown in FIG. 6, the initial heat generation amount Qi correlates with the nugget diameter d1 (see FIG. 1). The horizontal axis of the graph in FIG. 6 represents the initial heat generation amount Qi. The vertical axis represents the nugget diameter. Based on the correlation between the initial heat generation amount Qi and the nugget diameter, when the initial heat generation amount Qi is adopted as the actual data used to generate the prediction model 22 from among the actual data of electrical information during resistance spot welding and as the actual data input to the prediction model 22 to predict the weld characteristics, the prediction accuracy is improved.

[0050] It has also been found that various values ​​of interelectrode resistance R (Rfirst, Rfm, Rmax, Rml or Rlast, or Rmean, which is the average value of interelectrode resistance R over the entire period from the start to the end of welding) among the actual data of electrical information during resistance spot welding have a significant effect on the weld characteristics. The value of interelectrode resistance R is significantly affected by the current diameter. The current diameter is affected by the nugget diameter. For example, the value of interelectrode resistance R at the end of welding (Rlast) correlates with the final nugget diameter. Therefore, the value of interelectrode resistance R at the end of welding (Rlast) may be used to predict the nugget diameter of the weld.

[0051] <Welded part characteristics> Examples of the welded portion characteristics include the nugget diameter d of the welded portion 5 (see FIG. 1), cross tensile strength, shear tensile strength, or fracture strength by a mechanical peel test. The welded portion characteristics are measured by taking a test piece from a welded joint obtained by performing resistance spot welding on a sheet assembly in which two or more metal sheets are stacked. Among the welded portion characteristics, the nugget diameter may be measured by the method described in JIS Z3140, or by a non-destructive test such as an ultrasonic test or an eddy current test. The cross tensile strength may be measured by the method described in JIS Z3137. The shear tensile strength may be measured by the method described in JIS Z3136. The fracture strength by a mechanical peel test may be measured by the method described in ISO 14270.

[0052] <Welding conditions> The welding conditions of the resistance spot welding to be the target of the prediction of the welded portion characteristics in the prediction system 100 according to the present embodiment are not particularly limited. The welding conditions may be set, for example, so that a constant current flows from the start to the end of the resistance spot welding. The welding conditions may be set so that the period from the start to the end of the resistance spot welding is divided into a first stage and a second stage, and the welding conditions are different in the first stage period and the second stage period, resulting in two-stage welding. In particular, when welding a plate assembly in which three or more metal plates are overlapped and the plate thickness ratio, which is the value obtained by dividing the total plate thickness of the plate assembly by the plate thickness of the thinnest metal plate among the metal plates, is 5 or more, it is preferable to perform the two-stage welding under the following conditions.

[0053] In the first stage welding, the welding pressure P1 (kN) is preferably set so as to satisfy the following formula (1) in relation to the thickness tm of the thinnest metal plate among the multiple metal plates. The thickness of the thinnest metal plate corresponds to the thickness (mm) of the upper steel plate 6 in Fig. 6. 0.8tm≦P1≦5tm ………(1)

[0054] If the welding pressure P1 in the first stage is 5 tm (kN) or more, the welding pressure becomes too high, the heat generated by contact resistance becomes small, and a nugget is not formed between the upper steel plate 6 and the middle steel plate 7. On the other hand, if the welding pressure P1 is 0.8 tm or less, the contact resistance between the electrode tip and the upper steel plate 6 becomes large, making it easier for sparks to occur and for splashing to occur between the upper steel plate 6 and the middle steel plate 7.

[0055] In the first stage welding, the current application time T1 (ms) is preferably set so as to satisfy the following formula (2). 30≦T1≦120 ………(2)

[0056] If the current flow time T1 is 30 ms or less, the current flow time is too short, and a nugget of the desired size is not formed between the upper steel sheet 6 and the middle steel sheet 7. On the other hand, if the current flow time T1 is longer than 120 ms, expulsion occurs.

[0057] In the first-stage welding, the welding current I1 (kA) is preferably set so as to satisfy the following formula (3) in relation to the thickness tm of the thinnest metal plate among the multiple metal plates. The thinnest plate thickness corresponds to the thickness (mm) of the upper steel plate 6 in Fig. 6. 3tm+5≦I1 ………(3)

[0058] When the welding current I1 in the first stage welding is equal to or less than (3tm+5), the contact resistance heat cannot be effectively utilized, and a nugget is not formed between the upper steel sheet 6 and the middle steel sheet 7.

[0059] For this reason, in the first stage of welding, it is preferable to set the welding pressure P1, welding current I1, or current flow time T1 so as to satisfy the above-mentioned equations (1), (2), and (3) in order to form a nugget of the desired size between the upper steel plate 6 and the middle steel plate 7.

[0060] In the second stage welding, the welding pressure P2 (kN) is preferably set so as to satisfy the following formula (4) in relation to the welding pressure P1 in the first stage welding. P1≦P2≦10P1 ………(4)

[0061] Moreover, it is preferable that the second-stage welding current I2 (kA) is set so as to satisfy the following formula (5) in relation to the first-stage welding current I1. 0.5≦I2≦I1 ………(5)

[0062] Moreover, it is preferable that the second-stage welding current application time T2 (ms) is set so as to satisfy the following formula (6) in relation to the first-stage welding current application time T1. T1≦T2≦10T1 ………(6)

[0063] If the second stage welding conditions are outside the above range, it becomes difficult to prevent the occurrence of expulsion or to obtain a predetermined nugget diameter. In addition, if the pressure is excessively increased, problems such as large heat marks or lifting may occur. When performing two-stage welding, it is desirable to set the predetermined time (Ts) among the performance data of the electrical information during the resistance spot welding described above to the time when the first stage period ends and the second stage period starts.

[0064] When selecting welding conditions for performing welding with two-stage current supply as described above, the number of combinations of conditions increases. Therefore, by applying the prediction system 100 and prediction method according to the present embodiment to select welding conditions for performing welding with two-stage current supply, the efficiency of the welding condition selection work is improved.

[0065] Furthermore, the amount of heat generated during a first stage of resistance spot welding may be acquired as the electrical information during the resistance spot welding. By using the amount of heat generated during the first stage of resistance spot welding as the electrical information during the resistance spot welding, the prediction accuracy of the weld characteristics is improved.

[0066] The prediction system 100 and the prediction method according to the present embodiment may be applied not only to the selection of welding conditions in the above-mentioned welding methods, but also to the selection of welding conditions in various other welding methods. The prediction system 100 and the prediction method according to the present embodiment may be applied to the selection of welding conditions in a method such as one-stage current conduction in which the current value and the pressure are constant, multi-stage current conduction in which the set value of the current value or the pressure is changed midway, or adaptive control in which the current value is controlled according to a reference heat quantity. Furthermore, the prediction system 100 and the prediction method according to the present embodiment may be applied to the selection of welding conditions in a method including post-current conduction after the main current conduction for forming a welded portion.

[0067] <Example> A specific embodiment will be described below. In this embodiment, a single-phase AC resistance spot welding machine of a servo motor pressure type is used as the welding device 30. The spot welding machine is equipped with a chromium copper DR type electrode having a tip diameter of 6 mm and a curvature radius of 40 mm. In this embodiment, the welding conditions of a plate set in which three steel plates including an upper steel plate 6, a middle steel plate 7, and a lower steel plate 8 are overlapped as illustrated in FIG. 1 are determined. A steel plate having a tensile strength of 270 MPa and a plate thickness of 0.7 mm is used as the upper steel plate 6. A steel plate having a tensile strength of 1470 MPa and a plate thickness of 1.6 mm is used as the middle steel plate 7 and the lower steel plate 8. That is, the plate set for which the welding conditions are determined in this embodiment is configured so that the first metal plate is thinner than the second and third metal plates, and the second and third metal plates have the same plate thickness.

[0068] The above-mentioned welding test of the plate assembly was carried out by the above-mentioned welding device 30. In this embodiment, the welding conditions of the welding test were set to two stages of current flow in which the welding current flowing between the electrodes is different in a first stage period represented by time T1 and a second stage period represented by time T2, as exemplified in Fig. 7. The horizontal axis of Fig. 7 represents the current flow time of the resistance spot welding. The vertical axis represents the current flowing between the electrodes. The welding current flowing between the electrodes in the first stage period is represented as I1. The welding current flowing between the electrodes in the second stage period is represented as I2.

[0069] As shown in FIG. 8A, the temperature rises at the contact surface between the upper steel plate 6 and the middle steel plate 7, which is thinner than the middle steel plate 7 and the lower steel plate 8, earlier than the contact surface between the middle steel plate 7 and the lower steel plate 8, and the steel plate melts to form a first welded portion 51. As shown in FIG. 8B, the temperature rises at the contact surface between the middle steel plate 7 and the lower steel plate 8, and the steel plate melts to form a second welded portion 52. As a result, as shown in FIG. 1, one welded portion 5 is formed to join three steel plates. The nugget diameter of the welded portion 5 is measured separately into a nugget diameter d1 at the contact surface between the upper steel plate 6 and the middle steel plate 7 and a nugget diameter d2 at the contact surface between the middle steel plate 7 and the lower steel plate 8, as shown in FIG. 1. The nugget diameters d1 and d2 are the diameters of the welded portion 5 when each steel plate constituting the plate assembly is viewed in plan. When the welded portion 5 is formed in a disk shape with the normal line of each steel plate constituting the sheet assembly as the rotation axis, the nugget diameters d1 and d2 are measured on a cross section of the sheet assembly drawn to pass through the rotation axis.

[0070] Specifically, the welding conditions were set such that during the first stage, the electrode pressure was 3-5 kN (kilonewtons), the welding current I1 was 5-13 kA (kiloamperes), and the current flow time T1 was 60 ms (milliseconds). Also, during the second stage, the welding conditions were set such that the electrode pressure was 3-5 kN, the welding current I2 was 5-10 kA, and the current flow time T2 was 360 ms.

[0071] The electrode pressure and the welding current were each changed within the above-mentioned ranges to set 180 combinations of welding conditions. Resistance spot welding was performed on the plate pairs for which the welding conditions were determined in this example, applying each of the 180 welding conditions, and 180 welded joints were produced.

[0072] In this embodiment, the nugget diameter d1 of the first welded portion 51 and the nugget diameter d2 of the second welded portion 52 were analyzed as the welded portion characteristics of 180 welded joints. Specifically, the nugget diameters were measured by cutting the welded portion of the welded joint obtained by welding, etching the cross section, and then observing it with an optical microscope. The nugget diameter d1 of the first welded portion 51 was measured as the nugget diameter d1 between the upper steel plate 6 and the middle steel plate 7 illustrated in FIG. 1. The nugget diameter d2 of the second welded portion 52 was measured as the nugget diameter d2 between the middle steel plate 7 and the lower steel plate 8 illustrated in FIG. 1.

[0073] The data acquiring unit 12 of the predictive model generating device 10 acquires performance data of electrical information during resistance spot welding to which each of the 180 welding conditions is applied, and performance data of nugget diameters d1 and d2 of the 180 welded joints. The data acquiring unit 12 associates the performance data of electrical information during resistance spot welding corresponding to each welding condition with the performance data of nugget diameters d1 and d2, and stores them in the database 14 of the predictive model generating device 10.

[0074] The model generation unit 16 of the prediction model generation device 10 acquires data associating actual data of electrical information during resistance spot welding with actual data of nugget diameters d1 and d2 from the database 14. The model generation unit 16 executes machine learning using the acquired data as learning data, and generates a first prediction model 221 that outputs a prediction result of the nugget diameter d1 of the first welded portion 51 located between the upper steel plate 6 and the middle steel plate 7, and a second prediction model 222 that outputs a prediction result of the nugget diameter d2 of the second welded portion 52 located between the middle steel plate 7 and the lower steel plate 8, when the actual data of electrical information during resistance spot welding is input. That is, in this embodiment, a different prediction model is generated for each combination of two steel plates that contact each other.

[0075] In the plate combination exemplified in FIG. 1, the plate thickness ratio of the combination of the upper steel plate 6 and the middle steel plate 7, which are in contact with each other, is about 2.3. The plate thickness ratio of the combination of the middle steel plate 7 and the lower steel plate 8, which are in contact with each other, is 1. When the plate thickness ratio differs for each combination of two steel plates, it has been found that the prediction accuracy of the nugget diameter in each combination is improved by applying a different model to each combination. The plate thickness ratio of the combination of the upper steel plate 6 and the middle steel plate 7 is also referred to as the first plate thickness ratio. The plate thickness ratio of the combination of the middle steel plate 7 and the lower steel plate 8 is also referred to as the second plate thickness ratio. A different model may be applied to each combination in which the first plate thickness ratio and the second plate thickness ratio are different from each other.

[0076] Specifically, it has been found that when the thicknesses of the steel plates in contact with each other are close, that is, when the plate thickness ratio is close to 1, the prediction accuracy of the nugget diameter is improved by using a linear model. On the other hand, when the thicknesses of the steel plates in contact with each other are significantly different, that is, when the plate thickness ratio is large, the prediction accuracy of the nugget diameter is improved by using a nonlinear model. As the linear model, for example, Ridge regression or Elastic-Net may be used. As the nonlinear model, a neural network or a regression tree model may be used. The models applied as the linear model and the nonlinear model are not limited to the above-mentioned models, and may be appropriately selected according to the configuration of the plate assembly for which the welding conditions are determined. In this embodiment, it is assumed that LightGBM, which is classified as a nonlinear model, is applied as the first prediction model 221 for predicting the characteristics of the first welded portion 51 between the upper steel plate 6 and the middle steel plate 7, which have a large plate thickness ratio. On the other hand, it is assumed that Ridge regression, which is classified as a linear model, is applied as the second prediction model 222 for predicting the characteristics of the second welded portion 52 between the middle steel plate 7 and the lower steel plate 8, which have a plate thickness ratio of 1.

[0077] The machine learning method executed by the model generation unit 16 to generate the prediction model 22 may be a known learning method regardless of whether it is a linear model or a nonlinear model. The machine learning method may be any method as long as the learned first prediction model 221 or second prediction model 222 can obtain a prediction accuracy of the characteristics of the first welded portion 51 or the second welded portion 52 that is sufficient for practical use. In particular, there are many methods that can be applied regarding nonlinear regression. For example, support vector regression, Gaussian process, or an ensemble model combining multiple models may be used. In addition, the first prediction model 221 or the second prediction model 222 may be appropriately updated by re-learning using the latest learning data after being generated by executing machine learning.

[0078] As described above, the prediction device 20 can predict the nugget diameter of the welded portion of a welded joint obtained in a welding test performed by applying new welding conditions, without observing the cross section of the welded joint, by using the prediction model 22 including the first prediction model 221 and the second prediction model 222 generated by the prediction model generating device 10. In this embodiment, 50 new welding conditions were set and 50 welded joints were created.

[0079] When the results of predicting nugget diameters d1 and d2 of the welds of 50 welded joints using a prediction model 22 including a first prediction model 221 to which LightGBM is applied and a second prediction model 222 to which Ridge regression is applied were compared with the results of actually measuring the nugget diameters d1 and d2 of the welds of 50 welded joints, the nugget diameter d1 between the upper steel plate 6 and the middle steel plate 7 was predicted with an error of ±6%. Also, the nugget diameter d2 between the middle steel plate 7 and the lower steel plate 8 was predicted with an error of ±2%.

[0080] On the other hand, as a comparative example, when the prediction model 22 in which Ridge regression was applied to the first prediction model 221 was used, the nugget diameter d1 between the upper steel plate 6 and the middle steel plate 7 was predicted with an error of ±9%. In other words, the prediction accuracy of the nugget diameter d1 when the LightGBM classified as a nonlinear model is applied to the first prediction model 221 is improved compared to the prediction accuracy of the nugget diameter d1 when the Ridge regression classified as a linear model is applied to the first prediction model 221.

[0081] As another example, a prediction model 22 was generated by performing learning in which an initial heat generation amount Qi was further added to the learning data as actual data of electrical information during resistance spot welding. The initial heat generation amount Qi is the cumulative heat generation amount from the start of current flow until 60 ms (milliseconds). When using the prediction model 22 generated by adding the initial heat generation amount Qi, the nugget diameter d1 between the upper steel plate 6 and the middle steel plate 7 was predicted with an error of ±7%. Also, the nugget diameter d2 between the middle steel plate 7 and the lower steel plate 8 was predicted with an error of ±2%. By adopting the initial heat generation amount Qi as actual data to be input to the prediction model 22, the prediction accuracy of the nugget diameter d1 between the upper steel plate 6 and the middle steel plate 7 was improved. In other words, by adopting the initial heat generation amount Qi as actual data to be input to the prediction model 22, the prediction accuracy of the nugget diameter at the contact surface between the thin steel plate and the thick steel plate was improved.

[0082] As described in the examples, the prediction system 100 according to the present disclosure predicts welded portion characteristics such as nugget diameter with high accuracy based on actual data of electrical information during resistance spot welding. By predicting the nugget diameter with high accuracy, the process of exposing a cross section of a welded joint created in a welding test and actually measuring the nugget diameter is omitted. The labor and cost consumed in the process of exposing a cross section of a welded joint and actually measuring the nugget diameter is reduced. As a result, the number of welding conditions set in a welding test is increased, and the welding conditions are efficiently selected.

[0083] The prediction system 100 and the prediction method according to the present disclosure are executed on a plate assembly in which two or more metal plates are overlapped, thereby improving the efficiency of the selection of welding conditions for the plate assembly in which two or more metal plates are overlapped. The prediction system 100 and the prediction method according to the present disclosure are executed on a plate assembly in which three or more metal plates are overlapped, thereby improving the efficiency of the selection of welding conditions for the plate assembly in which three or more metal plates are overlapped. The selection of welding conditions for a plate assembly in which three or more metal plates are overlapped is more difficult than the selection of welding conditions for a plate assembly in which two metal plates are overlapped. Therefore, the effect of improving the efficiency of the selection of welding conditions is further increased by executing the prediction system 100 and the prediction method according to the present disclosure on a plate assembly in which three or more metal plates are overlapped.

[0084] In the above-mentioned embodiment, the thickness of the upper steel plate 6 was 0.7 mm, and the thickness of the middle steel plate 7 was 1.6 mm. That is, the thickness ratio, which is the value obtained by dividing the total thickness of the plate group by the thickness of the thinnest metal plate in the plate group, was about 5.6. It is known that the larger the thickness ratio of the plate group to be subjected to resistance spot welding, the more difficult it is to form a weld having a required nugget diameter at the contact surface between the thin steel plate and the thick steel plate. In particular, in resistance spot welding of a plate group having a thickness ratio exceeding 5, it is difficult to determine welding conditions for forming a weld having a required nugget diameter. The prediction system 100 according to the present disclosure can also predict with high accuracy the weld characteristics for a plate group in which steel plates are stacked in a combination with a larger thickness ratio, or in particular in a combination with a thickness ratio of 5 or more, and can improve the efficiency of the work of determining welding conditions.

[0085] <Flowchart example> In the prediction system 100 according to this embodiment, the prediction model 22 may be generated by executing a flowchart including the steps illustrated in FIG.

[0086] The welding device 30 performs a welding test on the plate assembly for which the welding conditions for resistance spot welding are to be determined (step S1). The welding conditions for the welding test may be set as appropriate.

[0087] The data acquisition unit 12 of the predictive model generating device 10 acquires electrical information during resistance spot welding performed by the welding device 30 in the welding test (step S2). The data acquisition unit 12 of the predictive model generating device 10 also acquires analysis results of the characteristics of each of the first welded portion 51 and the second welded portion 52 for the welded joint sample obtained by the welding test (step S3). The analysis results of the characteristics of each of the first welded portion 51 and the second welded portion 52 are obtained by the analysis device 40 analyzing the welded joint sample obtained by the welding test offline. The data acquisition unit 12 stores data in the database 14 that associates the electrical information during the resistance spot welding with the analysis results of the welded portion characteristics.

[0088] The model generation unit 16 of the prediction model generation device 10 performs learning using the data stored in the database 14 as learning data, and generates a first prediction model 221 that predicts the characteristics of the first welded zone 51 and a second prediction model 222 that predicts the characteristics of the second welded zone 52 (step S4). After completing the procedure of step S4, the prediction system 100 ends the execution of the flowchart in FIG.

[0089] In the prediction system 100 according to this embodiment, a flowchart including the procedure illustrated in FIG. 10 may be executed to determine welding conditions by welding tests.

[0090] Test conditions for conducting a welding test on a plate assembly for which welding conditions for resistance spot welding are to be determined are set (step S11). The test conditions may be set by a person who performs the welding test, or the like.

[0091] Prediction device 20 acquires electrical information during resistance spot welding performed by welding device 30 in a welding test from welding device 30 (step S12). Prediction device 20 predicts the characteristics of each of first welded portion 51 and second welded portion 52 of the welded joint obtained by the welding test, based on the electrical information during resistance spot welding (step S13). Prediction device 20 may input the electrical information during resistance spot welding to each of first prediction model 221 and second prediction model 222, and acquire a prediction result of the characteristic of first welded portion 51 output from first prediction model 221 and a prediction result of the characteristic of second welded portion 52 output from second prediction model 222.

[0092] It is determined whether the prediction results of the characteristics of the first welded portion 51 and the second welded portion 52 made by the prediction device 20 are good (step S14). The prediction results of the characteristics of the first welded portion 51 and the second welded portion 52 may be determined by a person who performs the welding test, or by the prediction device 20 or an external device that acquires the prediction results from the prediction device 20. If the prediction results of the characteristics of the first welded portion 51 and the second welded portion 52 are not good (step S14: NO), the process returns to the test condition setting procedure in step S11 and the test is repeated.

[0093] If the predicted results of the respective characteristics of the first welded portion 51 and the second welded portion 52 are good (step S14: YES), the welded joint obtained by the test is actually analyzed, and it is determined whether the analysis results are good (step S15). The determination of the analysis results of the respective characteristics of the first welded portion 51 and the second welded portion 52 may be performed by the person who performs the welding test, or may be performed by a device such as analysis device 40. If the analysis results of the respective characteristics of the first welded portion 51 and the second welded portion 52 are not good (step S15: NO), the process returns to the test condition setting procedure of step S11 and the test is repeated.

[0094] If the analysis results of the characteristics of the first welded portion 51 and the second welded portion 52 are good (step S15: YES), the test conditions under which good characteristics were obtained are determined as the welding conditions (step S16). After completing the procedure of step S16, the prediction system 100 ends the execution of the flowchart in FIG.

[0095] The prediction device 20 may further perform a step of predicting the characteristics of the welded portion 5 including the first welded portion 51 and the second welded portion 52 based on the prediction result of the characteristics of the first welded portion 51 and the prediction result of the characteristics of the second welded portion 52.

[0096] (summary) As described above, according to the prediction system 100 of the present disclosure, the welded joint characteristics of the welded joint are predicted based on the actual data of electrical information during resistance spot welding when a welding test is performed on a plate assembly for which welding conditions are to be determined, without actually analyzing the welded joint obtained by the welding test. By being able to predict the welded joint characteristics without actually analyzing them, it is possible to easily increase the number of welding tests with different welding conditions. As a result, the efficiency of the work of selecting suitable welding conditions for the welded joint to have a predetermined nugget diameter or a predetermined strength when the plate assembly to be welded is improved.

[0097] In addition, in a sheet assembly formed by overlapping at least a first metal plate, a second metal plate, and a third metal plate, the characteristics of a first welded portion 51 between the first metal plate and the second metal plate and the characteristics of a second welded portion 52 between the second metal plate and the third metal plate are predicted by separate models, thereby improving the prediction accuracy of the welded portion characteristics.

[0098] In addition, when the first plate thickness ratio between the first metal plate and the second metal plate and the second plate thickness ratio between the second metal plate and the third metal plate are different, the characteristics of the first welded portion 51 and the characteristics of the second welded portion 52 are predicted using separate models, thereby improving the prediction accuracy of the welded portion characteristics.

[0099] Furthermore, when the first plate thickness ratio is large, the characteristics of the first welded portion 51 are predicted by applying a nonlinear model, thereby improving the prediction accuracy of the welded portion characteristics.

[0100] 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 can make various modifications or alterations based on the present disclosure. Therefore, it should be noted that these modifications or alterations are included in the scope of the present disclosure. For example, the functions included in each component or each step can be rearranged so as not to be logically inconsistent, and multiple components or steps can be combined into one or divided. The embodiments of the present disclosure can also be realized as a program executed by a processor included in the device or a storage medium on which a program is recorded. It should be understood that these are also included in the scope of the present disclosure. [Explanation of symbols]

[0101] 100 Prediction System 3, 4 electrodes 5 Welded part (51: first welded part, 52: second welded part) 6 Upper steel plate 7 Medium steel plate 8 Lower steel plate 10 Prediction model generating device (12: data acquisition unit, 14: database, 16: model generating unit) 20 Prediction device (22: prediction model, 221: first prediction model, 222: second prediction model) 30 Welding equipment 40 Analysis equipment

Claims

1. A method for predicting characteristics of a welded portion of a plate set including three or more overlapping metal plates, the plate set including at least a first metal plate, a second metal plate, and a third metal plate, obtained by resistance spot welding in which a current is passed between two electrodes sandwiching the plate set, the method comprising: acquiring electrical information output during resistance spot welding, the electrical information including at least one of a current flowing between the two electrodes, a voltage between the two electrodes, a resistance between the two electrodes, or a heat generation amount between the two electrodes; inputting the electrical information into a first prediction model that predicts a characteristic of a first welded portion between the first metal plate and the second metal plate, and acquiring a prediction result of the characteristic of the first welded portion output from the first prediction model; inputting the electrical information into a second prediction model that predicts characteristics of a second weld between the second metal plate and the third metal plate, and acquiring a prediction result of the characteristics of the second weld output from the second prediction model; A method for predicting weld characteristics, comprising:

2. A plate thickness ratio, which is a value obtained by dividing the total plate thickness of the plate set by the plate thickness of the thinnest metal plate in the plate set, is 5 or more, The first metal plate is thinner than the second metal plate. The method for predicting weld properties according to claim 1 .

3. The method of claim 2 , wherein the first prediction model is a non-linear model.

4. The method for predicting weld characteristics according to claim 1 , further comprising the step of generating the first prediction model and the second prediction model by performing learning using the electrical information as learning data.

5. The method of claim 1 , wherein the properties of the weld include at least one of a nugget diameter, a tensile shear strength, a cross tensile strength, or a fracture strength by a mechanical peel test of the weld.

6. 6. The method for predicting welded portion characteristics according to claim 1, wherein in the step of acquiring the electrical information, a total amount of heat generated from a start of current flow to an end of current flow in resistance spot welding is acquired as the electrical information.

7. The method for predicting welded portion characteristics according to claim 6 , wherein in the step of acquiring electrical information, a resistance at the end of current flow between the two electrodes is acquired as the electrical information.

8. The resistance spot welding includes a first stage of passing a first current and a second stage of passing a second current; 7. The method for predicting welded portion characteristics according to claim 6, wherein in the step of acquiring electrical information output during the resistance spot welding, a heat generation amount during the first stage period within a period from a start of current flow to an end of current flow is acquired as the electrical information.

9. A method for generating a model predicting characteristics of welds including a first weld between the first metal plate and the second metal plate and a second weld between the second metal plate and the third metal plate, the welds being obtained by resistance spot welding in which three or more metal plates including at least a first metal plate, a second metal plate, and a third metal plate are overlapped and a current is passed between two electrodes that sandwich the three or more metal plates, the method comprising: acquiring electrical information output during resistance spot welding, the electrical information including at least one of a current flowing between the two electrodes, a voltage between the two electrodes, a resistance between the two electrodes, or a heat generation amount between the two electrodes; obtaining an analysis result of the characteristics of the weld when the resistance spot welding is performed; executing machine learning based on learning data that associates the electrical information with an analysis result of the characteristics of the welded portion when the resistance spot welding is performed, to generate a first prediction model that predicts characteristics of the first welded portion and a second prediction model that predicts characteristics of the second welded portion; A method for generating a predictive model, comprising:

Citation Information

Patent Citations

  • Resistance spot welding method

    JP1996318377A

  • Resistance spot welding method

    JP2019171450A

  • Spot weldment

    JP2020062655A

  • Manufacturing method of spot welded joint and spot welding device

    JP2022140236A

  • Manufacture method of cutter for paper shredder

    KR102180877B1