JP7676014B2 - WELDING JUDGMENT DEVICE, WELDING JUDGMENT DEVICE SYSTEM, WELDING JUDGMENT METHOD, AND WELDING JUDGMENT PROGRAM - Google Patents

Description

The present invention relates to a welding judgment device, a welding judgment device system, a welding judgment method, and a welding judgment program, and in particular to a welding judgment device and welding judgment device system, as well as a welding judgment method and a welding judgment program, that can accumulate welding conditions detected from a welding device and judge the suitability of welding based on fluctuations in the welding conditions.

In the field of welding machines, there is a trend toward automating the monitoring of welding conditions and the determination of welding conditions, and thus there are welding devices (robots) and systems equipped with machine learning and artificial intelligence (AI) technology. Examples of such devices include a spot welding device that determines the welding condition (Patent Document 1), a welding monitoring system (Patent Document 2), and a spot welding quality diagnosis system (Patent Document 3), and a welding monitoring system and welding monitoring method for a resistance welding machine (Patent Document 4).

However, in previously disclosed welding equipment and systems, the content and variety of data showing the state during welding was limited, and the amount of data stored in the welding equipment was limited. As a result, when a welding defect occurred during welding, the stored data was insufficient to provide a basis for determining the cause of the defect. As a result, when a welding defect occurred, the cause could not be immediately identified, and it took time to determine the cause and take measures such as changing the welding state. As a result, equipment operation rates and productivity at manufacturing sites decreased.

Furthermore, taking the example of welding automobile bodies, in addition to the fact that more types of steel plates are used than before, aluminum plates and other auto body materials have been added to the list of welding objects. As the number of types of welding objects increases, the welding conditions have become more complex. Specifically, welding conditions such as multi-stage current flow, multi-stage pressure, upslope and downslope are used in spot welding. This has increased the burden of managing welding conditions that are suited to the intended welding objects. Due to this increased burden, manufacturing sites are burdened with the need to allocate a large number of personnel when setting up new manufacturing equipment and maintaining manufacturing equipment that is currently in operation.

JP 2018-034172 A JP 2018-001184 A JP 2016-203246 A JP 2020-179406 A

As a result of this series of events, the inventors have earnestly investigated methods to increase the types of welding conditions to be managed, thereby improving the precision of welding management, and reducing the burden of managing welding conditions, in order to respond to the diversification of welding objects and welding conditions.

The present invention has been made in consideration of the above points, and provides a welding judgment device, welding judgment device system, welding judgment method, and welding judgment program that can reduce the burden of managing welding conditions while increasing the variety of welding conditions, and also improve the accuracy of welding management.

That is, the welding judgment device judges the welding state of a welding device that welds workpieces, and the welding judgment device is characterized by having an object information acquisition unit that acquires welding object information related to the workpieces themselves, a welding condition storage unit that accumulates accumulated information that associates the welding quality with each of multiple types of welding conditions controlled by the welding device when welding the workpieces in response to the welding object information, an appropriateness information generation unit that generates and accumulates the association of welding conditions corresponding to the welding object information associated with the welding quality as appropriateness information by a statistical method using the accumulated information, a sequential acquisition unit that acquires the same type of welding conditions as the multiple types of welding conditions controlled by the welding device when welding the workpieces corresponding to the welding object information as sequential welding condition information, a comparison unit that compares the sequential welding condition information with the appropriateness information, a judgment unit that judges whether the sequential welding condition information deviates from the appropriateness information and generates a judgment result, and an output unit that outputs the judgment result.

Furthermore, the welding condition storage unit of the welding judgment device may acquire multiple types of welding conditions controlled by the welding device for each welding of the workpiece.

Furthermore, the statistical method may be machine learning using training data in which the welding quality when welding the workpiece is used as training data and multiple types of welding conditions are used as input data.

Furthermore, the statistical method may be a multivariate analysis in which the welding quality is the objective variable and multiple types of welding conditions are the explanatory variables.

Furthermore, the welding judgment device may be provided with a display unit that displays a time series graph of the welding conditions corresponding to each of the multiple types of welding conditions included in the suitability information, and a predetermined precision may be set for the fluctuation range of the time series graph.

Furthermore, the welding judgment device may be provided with a display unit that displays a time series graph of the welding conditions corresponding to each of the multiple types of welding conditions included in the suitability information, and a threshold value may be set for the fluctuation range of the time series graph.

Furthermore, the judgment unit may generate the judgment result using a judgment formula obtained by a statistical method using the judgment result and the suitability information.

Furthermore, the judgment unit of the welding judgment device may be configured to judge whether or not at least one of the multiple welding conditions included in the sequential welding condition information deviates from the appropriate information.

Furthermore, the welding device may be a resistance welding machine.

Furthermore, the welding judgment device may be implemented in each welding device.

The welding judgment device of the present invention includes an object information acquisition unit that acquires welding object information related to the workpiece itself, a welding condition storage unit that accumulates accumulated information that associates the welding quality with each of multiple types of welding conditions controlled by the welding device when welding the workpiece in response to the welding object information, an appropriateness information generation unit that generates and accumulates the association of welding conditions corresponding to the welding object information associated with the welding quality as appropriateness information using a statistical method using the accumulated information, a sequential acquisition unit that acquires the same type of welding conditions as the multiple types of welding conditions controlled by the welding device when welding the workpiece corresponding to the welding object information as sequential welding condition information, a comparison unit that compares the sequential welding condition information with the appropriateness information, a judgment unit that judges whether the sequential welding condition information deviates from the appropriateness information and generates a judgment result, and an output unit that outputs the judgment result. This reduces the burden of managing the welding conditions while increasing the number of welding conditions, and also improves the accuracy of welding management.

1 is a schematic diagram showing a configuration of a welding judgment system including a welding judgment device according to an embodiment; 1 is a schematic block diagram showing a configuration of a welding judgment device; FIG. 2 is a schematic block diagram showing a functional section of a welding judgment device. FIG. 1 is a schematic diagram showing the main parts of a resistance welding machine. FIG. 2 is a schematic cross-sectional view showing a main portion of a resistance welded portion. FIG. 2 is a schematic diagram showing an example of welding conditions for a resistance welding machine. FIG. 1 is a schematic diagram showing a feedback flow of a previous weld defect. FIG. 2 is a schematic diagram showing an overview of a process in a welding judgment device. FIG. 2 is a first exemplary schematic diagram showing an example of a process in the welding judgment device; FIG. 11 is a second exemplary schematic diagram showing an example of a process in the welding judgment device. 4 is a first flowchart showing a processing procedure in the welding judgment device of the embodiment. 5 is a second flowchart showing the processing procedure in the welding judgment device of the embodiment. 10 is a third flowchart showing the processing procedure in the welding judgment device of the embodiment.

The schematic diagram in FIG. 1 shows a welding judgment system 1 that includes multiple welding devices 2, 3, and 4, each of which is connected to a welding judgment device 100. The welding device illustrated as an embodiment is a welding device known as a resistance welding machine or a spot welding machine. In the explanation of the embodiment, a resistance welding machine will be used as an example of the welding device.

The welding judgment device 100 is installed in each of the multiple welding devices 2, 3, and 4, and stores and monitors various welding conditions (see Figures 5 and 6 below) for the workpiece when operating the welding device (resistance welding machine) required for welding, such as the current and voltage in the welding devices 2, 3, and 4 (resistance welding machine). The welding judgment device 100 then plays a role in judging the welding condition of the welding device, such as poor welding of the workpiece. In addition, since the welding judgment device 100 is installed in each of the multiple welding devices 2, 3, and 4, it becomes an edge computing device implemented in each welding device.

In the welding judgment system 1, the welding judgment device 100 installed in each of the multiple welding devices 2, 3, and 4 is connected to an in-line measuring device (not shown), a resistance welding controller, and a programmable logic controller, and is further connected to a monitoring device 500. A server, a display, etc. are also connected to the monitoring device 500 as appropriate. The welding judgment device 100 and the monitoring device 500 have substantially the same functions. A welding judgment device 100 is installed for each individual welding device. In other words, the monitoring device 500 monitors the welding conditions of each of all of the connected welding devices 2, 3, and 4.

In the illustration, the number of welding devices is three as shown. The number of welding devices is not limited to three, and it is possible to have one or to increase the number to four or more. Of course, this configuration example is just one example, and each device may be added or omitted as necessary. Details of the welding judgment device 100 will now be described with reference to the illustration.

The welding judgment device 100 installed in each of the multiple welding devices 2, 3, and 4 is a computer that monitors the welding conditions of the installed welding device and judges the suitability of welding based on the deviation from the existing accumulated welding conditions. As shown in the schematic block diagram of Figure 2, the welding judgment device 100 is equipped with a CPU 101, a ROM 102, a RAM 103, a storage unit 104, an input unit 105, an output unit 106, etc. in terms of hardware. It also includes a main memory, an LSI, etc. The monitoring device 500 also has a similar hardware configuration to that shown in the schematic block diagram of Figure 2. The welding judgment device 100 and the monitoring device 500 are various electronic computers (computational resources) such as personal computers (PCs), mainframes, workstations, tablet terminals, and smartphones.

The input unit 105 and the output unit 106 are known input/output interfaces, and are connected to the welding device 2 disclosed in FIG. 1 and the like. In the case of the monitoring device 500, the output unit 106 is connected to a display (such as a liquid crystal display device or an organic EL display device) as the display unit 107. These are merely examples, and may be combined as appropriate and selected optimally.

The memory unit 104 of the welding judgment device 100 (monitoring device 500) is a known storage device such as an HDD or SSD. In addition, each functional unit that performs various calculations in the welding judgment device 100 (monitoring device 500) is a computing element such as the CPU 101. As shown in the schematic block diagram of FIG. 3, each functional unit in the CPU 101 of the welding judgment device 100 (monitoring device 500) includes a target information acquisition unit 110, a welding condition storage unit 120, an appropriateness information generation unit 130, a sequential acquisition unit 140, a comparison unit 150, a judgment unit 160, an output unit 170, a machine learning unit 180, etc. The operation and execution of the welding judgment device 100 (monitoring device 500) is realized in software terms by a welding monitoring program for a resistance welding machine loaded into the main memory, etc.

When each functional unit of the welding judgment device 100 (monitoring device 500 of the welding judgment system 1) in FIG. 1 is realized by software, the welding judgment device 100 (monitoring device 500) is realized by executing instructions of a program, which is software that realizes each function. The recording medium that stores this program can be a "non-transient tangible medium," such as a CD, DVD, semiconductor memory, or programmable logic circuit. In addition, this program may be supplied to the welding judgment device 100 (monitoring device 500) via any transmission medium (communication network, broadcast waves, etc.) that can transmit the program.

The schematic diagram in FIG. 4 shows the main parts of the welding device 2. Since the welding devices 2, 3, and 4 in FIG. 1 are the same machine, the welding device 2 is shown as a representative. A resistance welding portion 11 is provided at the tip of the welding device 2 (resistance welding machine). The resistance welding portion 11 shown in the figure is a portion with an inverted C-shaped clamp structure. The shape of the resistance welding portion 11 is appropriately selected according to the target portion of the resistance welding. Electrode portions 12 and 13 are connected to the resistance welding portion 11. The electrode portions 12 and 13 are consumables and can be attached and detached freely, and are replaced from the resistance welding portion 11 due to wear or the like. In addition, a welding judgment device 100 is connected to the resistance welding portion 11 to measure the welding conditions of the resistance welding portion 11. The welding device 2 has an arm portion connected by a joint portion, and the joint portion is equipped with a servo motor (not shown).

As can be seen from the schematic cross-sectional view of FIG. 5, the workpieces W1 and W2 are placed between the electrode portion 12 and the electrode portion 13 of the resistance welding portion 11. Then, as the electrode portion 12 advances toward the electrode portion 13, the electrode portion 12 and the electrode portion 13 come into contact with the workpieces W1 and W2 and are pressed against them. Then, resistance heating occurs due to the passage of electricity from the electrode portion 12 and the electrode portion 13 of the resistance welding portion 11 to the workpieces W1 and W2, causing the metal of the workpieces W1 and W2 to partially melt, and a metal molten area 14 is formed between the workpieces W1 and W2.

As shown in the schematic diagram of FIG. 6, for example, when the welding device 2 (resistance welding machine) is of AC or DC type, the supply voltage is stepped down by a transformer (not shown). At this time, the stepped-down voltage on the secondary side of the transformer is measured through a voltage detection line wired to the resistance welding portion 11. In addition, a non-contact ammeter 22 such as a CT type ammeter is attached to the resistance welding portion 11. The thin dashed lines in FIG. 6 show the actual wiring flow of the voltage detection line measured when the resistance welding portion 11 is energized (during welding). The thick dashed lines in FIG. 6 show the actual wiring flow of the non-contact ammeter 22 measured by the non-contact ammeter 22 when the resistance welding portion 11 is energized (during welding). Since the resistance welding portion 11 is equipped with a voltage detection line, a non-contact ammeter 22, etc., the welding conditions are constantly measured (constantly monitored) each time resistance welding is performed. Of course, the measurement method shown and described is just one example, and the measurement is appropriately performed according to the method of the welding device 2 (resistance welding machine).

In the example of FIG. 6, the current and voltage during operation of the welding device 2 (resistance welding machine) are exemplified as the welding conditions. Furthermore, the welding conditions include quantified information such as the current flow time, the pressure applied to the workpiece in the case of a resistance welding machine, the resistance value, the distortion value, the amount of heat input, the amount of water for cooling the device, the control amount of the motor that drives the resistance welding section (see FIG. 4), and the air temperature during welding. The welding conditions can be graphs (waveforms) that change with time on the horizontal axis. As an example, in the case of pressure, a graph showing the change over time in pressure from the start to the end of welding can be used as the welding conditions. Various welding conditions are acquired by the welding judgment device 100.

In addition to the information acquired on the welding device side, information on the workpiece side (welding target information, peripheral information) is acquired by the welding judgment device 100. The welding target information is, for example, information on the shape of the workpiece, the material of the workpiece (type of steel plate, aluminum material, etc.), the name of the welding process (welding area), the number of welding points, plate assembly, date and time, etc.

In other words, both the welding conditions acquired on the welding equipment side and the welding target information provided on the workpiece side are acquired by the welding judgment device 100 (monitoring device 500), making it easier for the welding judgment device 100 (monitoring device 500) to find the correlation between the welding conditions and the welding target information.

From now on, we will explain each functional part of the welding judgment device 100 (monitoring device 500) (its CPU 101) in order using the above-mentioned Figures 2 and 3.

The object information acquisition unit 110 acquires welding object information related to the workpiece itself. As described above, the welding object information includes information such as the material of the workpiece, the process name (welding portion) during welding, the number of welding points, plate assembly, date and time, etc. In other words, if the material of the workpiece, the welding portion, etc. are different, the welding conditions on the welding device side will obviously change. Therefore, information related to the workpiece itself, which is the premise of welding, is acquired in advance. Note that the acquisition of welding object information is performed every time the workpiece is changed. The acquisition of welding object information is input in advance by a worker or the like according to the workpiece to be welded. As an example, the input is completed by reading a two-dimensional barcode or the like attached to each workpiece through a reader. Alternatively, the information may be input to the monitoring device 500 and then transmitted to each welding judgment device 100.

The welding condition storage unit 120 stores accumulated information that associates the quality of welding with multiple types of welding conditions controlled by the welding device when welding the workpiece in response to the welding object information. For example, the welding object information acquired by the object information acquisition unit 110 may include information such as the material of the workpiece being SS400, resistance welding of the door frame part of an automobile vehicle, and 15 welding points. In response to the welding object information, multiple types of welding conditions that can be managed and controlled by the welding device 2 (resistance welding machine) are comprehensively acquired, such as current, voltage, current flow time, pressure applied to the workpiece, resistance value, distortion value, and input heat amount. A collection of these multiple types of welding conditions becomes accumulated information. The welding conditions that form the basis of the accumulated information depend on the welding object information, the type of welding device itself, and information that can be acquired by the welding device.

The welding condition storage unit 120 acquires multiple types of welding conditions controlled by the welding device for each welding of the workpiece. Since multiple types of welding conditions are acquired for each welding, accidental omissions are avoided, and the reliability of the accumulated welding conditions is improved.

When constantly storing multiple types of welding conditions, if extremely small fluctuations are continually stored, the storage capacity of the storage unit 104 of the welding judgment device 100 (monitoring device 500) may not be able to keep up. In this case, the data volume can be compressed and the calculation burden reduced by the method described below. First, a time series graph of the welding conditions is created for each of the multiple types of welding conditions (the resistance waveform, heat waveform, and expansion amount waveform in Figures 9 and 10 are referenced). The time series graph may be displayed on the display of the display unit 107, etc.

A specified precision is set for the fluctuation range of a time series graph. In the case of a graph with large up and down fluctuations, the measurement waveform is flattened appropriately, so the amount of storage in the memory unit 104 is reduced. To achieve the specified precision, processing such as the use of moving average lines for the time series graph and statistically excluding abnormal values from the detected values is performed.

In addition, a predetermined threshold is set for the fluctuation range of the time series graph. An appropriate threshold (range) is set in advance for each of the multiple types of welding conditions, and values above and below the threshold are removed from the graph. This eliminates the apparent extreme fluctuation range from the time series graph, and the measurement waveform is appropriately flattened.

The suitability information generating unit 130 generates and stores, as suitability information, a link between welding conditions corresponding to welding object information associated with the quality of welding, using a statistical method using accumulated information. In the suitability information generating unit 130, a link is made between the welding conditions corresponding to the welding object information acquired through the welding condition storing unit 120 and the welding quality corresponding to each welding condition. In other words, for welding performed under welding conditions corresponding to certain welding object information, the suitability (goodness or badness) of the welding quality is fed back (acquired and reflected) from the user's viewpoint, etc. Then, the suitability of the welding quality according to various welding conditions is collected. Each of these links between the welding conditions and the quality of welding quality is generated as suitability information. In this way, the links between the welding conditions and the quality of welding quality are generated as suitability information, and the suitability information is stored.

For example, when the material of the welding object is aluminum, the thermal conductivity is higher and the base material resistance is lower than that of iron, so the most important welding conditions that affect the welding are the pressure applied to the workpiece and the resistance value. The appropriate information is multiple types of welding conditions that are used when good welding is performed on the workpiece. In other words, the appropriate information is a collection of multiple types of welding conditions that are used when the welding equipment is operating normally and the workpiece is welded appropriately.

As a statistical method used to process the accumulated information, supervised machine learning is performed using the welding quality when welding the workpiece as the training data and multiple types of welding conditions as the input data. For example, the resistance waveform, heat waveform, expansion waveform, etc. acquired from the accumulated information are compared with a pass (good) welding quality (welding condition) acquired by sampling inspection or non-destructive inspection of the actual welded workpiece. Note that inspection of the actual welded workpiece is only performed in the initial stage for generating the appropriate information. Then, the accumulated information is associated (linked, annotated) with the welding quality. In this way, the training data is generated and machine learning with the training data is performed.

To go into more detail about the statistical method used to process the accumulated information, it is carried out as a multivariate analysis with welding quality as the objective variable and multiple types of welding conditions as explanatory variables. Multivariate analysis is highly convenient as a statistical method because it can derive relationships between multiple variables at once.

The suitability information generating unit 130 is equipped with a machine learning unit 180 as the operating entity of the statistical method (machine learning). The machine learning unit 180 executes the machine learning and multivariate analysis with the teacher data described above. In addition to the above-mentioned methods, the machine learning analysis method in the machine learning unit 180 includes regression analysis such as linear regression, logistic regression, and support vector machine. In this way, the suitability information generating unit 130 generates and accumulates suitability information that serves as the standard for determining whether welding is good or bad, based on multiple types of welding conditions that correspond to welding quality.

By going through the process up to this point, the welding equipment (resistance welding machine, spot welding machine) is actually operated to weld various workpieces, and suitability information corresponding to the welding object information is generated and stored.

From this point on, the quality of the welding of the workpieces is judged individually based on the accumulated information. The sequential acquisition unit 140 acquires, as sequential welding condition information, the same types of welding conditions as the multiple types of welding conditions controlled by the welding device when welding the workpieces corresponding to the welding object information. For example, in the case where the material of the workpieces is SS400, resistance welding is performed on the door frame part of an automobile vehicle, and the number of welding points is 15, the current, voltage, current flow time, pressure applied to the workpieces, resistance value, distortion value, input heat amount, etc. are acquired as sequential welding condition information corresponding to the welding object information. The sequential acquisition unit 140 comprehensively acquires multiple types of welding conditions that can be managed and controlled by the welding device 2 (resistance welding machine) as sequential welding condition information.

The collation unit 150 compares the sequential welding condition information with the appropriateness information. In this process, the sequential welding condition information (multiple types of welding conditions) is compared with the appropriateness information, which is multiple types of welding conditions when the welding device operates normally and appropriately welds the workpieces. In other words, multiple types of welding conditions that are common between the welding of the workpieces currently being performed and the welding of the workpieces previously performed are selected.

The judgment unit 160 judges whether the sequential welding condition information deviates from the appropriate information and generates a judgment result. Here, the welding of the workpiece currently being performed is compared with the welding of the workpiece previously performed, and it is judged whether the welding conditions are appropriate. The judgment result is a conclusion that the welding of the workpiece currently being performed is "normal" or "abnormal". Furthermore, in generating the judgment result in the judgment unit 160, a judgment formula obtained by a statistical method using the judgment result and the appropriateness information is used. The statistical method referred to here is the establishment of a correlation such as the least squares method, and the level of the correlation coefficient or the coefficient of determination, and the judgment formula is a correlation formula such as the least squares method.

As mentioned above, the welding conditions can be understood as a graph (waveform) that changes with time on the horizontal axis, and are acquired and stored as waveforms. In this case, deviations may be determined by comparing the waveforms of multiple types of welding conditions included in the sequential welding condition information with multiple types of welding conditions included in the suitability information. Furthermore, a threshold value may be set for each of the multiple welding conditions included in the suitability information, and a deviation may be determined when an excess over the threshold value is detected.

As a practical matter, it is impossible for the multiple welding conditions included in the suitability information to match the multiple types of welding conditions included in the sequential welding condition information in terms of all values and waveforms. Fluctuations can occur due to measurement errors and other factors. Therefore, as a practical solution, a threshold is set, and if the results converge within the threshold, it can be treated as if there are no problems with welding. The threshold can be set on both the upper and lower sides of the graph of the welding conditions over time, or on either the upper or lower side.

Furthermore, the determination unit determines whether at least one of the multiple welding conditions included in the sequential welding condition information deviates from any of the multiple welding conditions included in the appropriateness information. Both the appropriateness information and the sequential welding condition information each include multiple welding conditions. For example, if there are three types of welding conditions, and one of them deviates (exceeds a threshold value), it will be determined that there is a welding abnormality even if there is no problem with the other two types of welding conditions.

The output unit 170 outputs the judgment result. Specifically, the conclusion of whether the welding of the currently being welded workpiece is "normal" or "abnormal" is displayed on the display (not shown) of the display unit 107 as to whether the welding is normal or abnormal. If the judgment result is abnormal, the product cannot be supplied to the production line with poor welding, and appropriate measures such as stopping the production line and welding again are implemented.

As can be seen from this series of processes, in the welding judgment device 100 (monitoring device 500), the welding of the workpiece currently being performed is monitored and accumulated through multiple welding conditions during that welding, and is then compared and judged against suitability information generated through machine learning based on statistical methods. In particular, it becomes possible to comprehensively monitor and judge multiple welding conditions. This makes it possible to move away from relying on the experience of on-site workers to select welding conditions. In addition, because of constant monitoring, the burden of inspections to confirm weld strength, such as sampling after welding and hitting the welded area with a hammer, is greatly reduced.

The use of the welding judgment device 100 (monitoring device 500) of the embodiment will be explained using the schematic diagram of resistance welding in Figure 7. The schematic diagram of Figure 7 explains the actual situation before the introduction of the welding judgment device 100 (monitoring device 500). Normally, in resistance welding, a metal melting area 14 is generated between the workpieces W1 and W2 (see Figures 5 and 7). The metal melting area 14 is a circular area called a nugget. Here, if the metal melting area 14 (nugget) does not meet the specified diameter, the area of the metal melting area 14 is insufficient, resulting in insufficient welding strength and poor welding.

In this case, in consideration of the result of poor welding, a number of welding conditions that can be monitored by the welding device 2, etc., are acquired as a cause analysis, along with the material, shape, etc. of the workpiece. In the figure, a normal waveform (statistical waveform) and an abnormal waveform are presented. In addition, the temperature, humidity, date, time, etc. during welding are also acquired.

Then, from the multiple welding conditions presented in the cause analysis, the welding conditions are selected based on the worker's manual effort and the worker's rule of thumb, and are fed back as the welding conditions to be managed. The worker's rule of thumb takes into account external disturbances, changes in the conditions of the workpiece or equipment, distortion of the workpiece itself, abnormalities in the equipment itself, etc. Therefore, since feedback to improve the welding conditions is given only after a welding defect occurs, the responsiveness of the work is not satisfactory. In addition, even if the workers are experienced, it is difficult for them to share the welding conditions to be managed among themselves, and it is also not easy to convey this to workers with little experience.

Therefore, by introducing the welding judgment device 100 (welding judgment system 1 equipped with the monitoring device 500), welding object information and multiple welding conditions during normal welding are constantly acquired, and suitability information serving as a sample of a good product is generated in advance, making it easy to compare and judge with newly performed welding. In addition, since the suitability information is generated through machine learning in the welding judgment device 100, the variability in the multiple welding conditions that should be adopted by each worker is eliminated without the involvement of the worker. In particular, since judgment is possible at the stage of newly performed welding, a quick response is possible even if a welding defect occurs.

FIG. 8 is a schematic diagram showing an overview of the processing when the welding judgment device 100 (welding judgment system 1 including the monitoring device 500) is introduced.
As described above, welding determination device 100 (monitoring device 500) measures and acquires welding object information and a plurality of corresponding welding conditions (S1).
In the embodiment, since the welding machine is a resistance welding machine, a plurality of welding conditions corresponding to the welding object information are stored and accumulated for each welding point (S2).
Normal welding conditions are learned as training data for machine learning from a plurality of welding conditions accumulated over a predetermined period of time (S3).
Suitability information is generated by linking the welding object information with a plurality of corresponding welding conditions (S4).The suitability information is then used for judgment.
Since the suitability information is generated through machine learning by associating the welding object information with a plurality of corresponding welding conditions, even when adopting new welding points or plate combinations, the suitability information once generated can be utilized, so there is no need to redo everything, and it is possible to reduce expenses required for production preparation, such as setting up new welding equipment (S5).

Figure 9 is a schematic diagram showing an overview of the introduction of the welding judgment device 100 (welding judgment system 1 equipped with a monitoring device 500). Figure 9 shows the flow from the generation of the aforementioned suitability information to the output of suitable welding conditions.

First, multiple welding conditions are acquired along with the acquisition of welding object information. In the figure, the resistance waveform, heat amount waveform, and expansion amount waveform of the welding device 2 (resistance welding machine) are shown as welding conditions. Next, each welding condition is summarized with the horizontal axis representing the current flow time and the vertical axis representing various values, and waveform analysis is performed using the statistical method described below. This process corresponds to data editing, or so-called visualization. Then, with the horizontal axis representing the current flow time and the vertical axis representing various values, optimal welding conditions (judgment logic) corresponding to the welding object information are generated through machine learning as appropriate information (appropriate conditions). Since the process from the acquisition of multiple initial welding conditions to the generation of appropriate information is performed automatically, labor savings are achieved. If the appropriate information were to be generated manually, the number of repetitions would be large, and the analysis burden would be large, making it impossible to achieve.

Figure 10 is a schematic diagram showing an overview of the introduction of the welding judgment device 100 (welding judgment system 1 equipped with a monitoring device 500). Figure 10 shows the flow from the generation of the aforementioned suitability information to the actual execution of the judgment.

First, multiple welding conditions are obtained along with the acquisition of welding object information. In the figure, the resistance waveform, heat amount waveform, and expansion amount waveform of the welding device 2 (resistance welding machine) are shown as welding conditions. Next, each welding condition is summarized with the horizontal axis representing the current flow time and the vertical axis representing various values, and waveform analysis is performed using the statistical method described below. This process corresponds to data editing, or so-called visualization. Then, for example, as multiple welding conditions, a comparison is made between graphs of the current flow time and resistance value of normal waveforms and abnormal waveforms, and a comparison is made between graphs of the current flow time and expansion amount of normal waveforms and abnormal waveforms. This corresponds to the processing in the collation unit 150, judgment unit 160, etc. described above. Then, in the case of abnormality detection, the abnormality judgment result is output by the output unit 170.

When performing the waveform analysis using the statistical method in the explanation of Figures 9 and 10, a judgment formula obtained by the statistical method is used. The statistical method referred to here is the establishment of a correlation such as the least squares method, and the magnitude of the correlation coefficient or the coefficient of determination, and the judgment formula is a correlation formula such as the least squares method. The welding conditions can be understood as a graph (waveform) that changes with time on the horizontal axis, and are acquired and stored as waveforms. In this case, deviations may be determined by comparing the waveforms of multiple types of welding conditions. Furthermore, a threshold value may be set for each of the multiple welding conditions included in the suitability information, and a deviation may be determined when an excess of the threshold value is detected.

It is impossible for multiple types of welding conditions to match in all of their values and waveforms. Fluctuations can occur due to measurement errors and other factors. In reality, a threshold is set, and if the results converge within the threshold, they can be treated as normal. The threshold can be set on both the upper and lower sides of the graph showing the welding conditions over time, or on either the upper or lower side.

The welding judgment method and welding judgment program of the embodiment will now be described with reference to the flowcharts of Figures 11 and 12. The welding judgment method is executed by the CPU 101 (computer) of the welding judgment device 100 (monitoring device 500) based on the welding judgment program of the welding judgment device 100 (monitoring device 500). The welding judgment program causes the CPU 101 (computer) of Figures 2 and 3 to execute a target information acquisition function, a welding condition accumulation function, an appropriate information generation function, a sequential acquisition function, a comparison function, a judgment function, an output function, and a machine learning function. Each function overlaps with the description of the welding judgment device 100 described above, so details will be omitted.

The flowchart in FIG. 11 shows the flow of the welding judgment method of the welding judgment device of the embodiment, and includes various steps of a target information acquisition step (S110), a welding condition accumulation step (S120), and a proper information generation step (S130). In addition, the welding judgment method of the embodiment also includes various steps not shown that are necessary as appropriate, such as storing the calculation results, calling them up, other calculations, input, output, and storage.

The object information acquisition function acquires welding object information related to the workpiece itself (S110; information acquisition step). The welding condition accumulation function acquires and accumulates, as accumulated information, multiple types of welding conditions controlled by the welding device when welding the workpiece in response to the welding object information over a predetermined period of time (S120; welding condition accumulation step). The suitability information generation function generates and accumulates, by a statistical method using the accumulated information, a link between welding conditions corresponding to the welding object information that is associated with the suitability of the welding quality as suitability information, as suitability information, using the accumulated information (S130; suitability information generation step).

The flowchart in FIG. 12 is a continuation of the flow of the welding judgment method of the welding judgment device of the embodiment, and includes various steps of a sequential acquisition step (S140), a collation step (S150), a judgment step (S160), and an output step (S170). In addition, the welding judgment method of the embodiment also includes various steps not shown that are necessary as appropriate, such as storing the calculation results, calling them up, other calculations, input, output, and storage.

The sequential acquisition function acquires, as sequential welding condition information, the same types of welding conditions as the multiple types of welding conditions controlled by the welding device when welding the workpiece corresponding to the welding object information (S140; sequential acquisition step). The collation function compares the sequential welding condition information with the appropriate information (S150; collation step). The determination function determines whether the sequential welding condition information deviates from the appropriate information and generates a determination result (S160; determination step). The output function outputs the determination result (S170; output step).

The flowchart in FIG. 13 is a continuation of the flow of the welding judgment method of the welding judgment device of the embodiment, and in particular shows a case where the machine learning unit 180 operates when performing machine learning in the suitability information generation 130. The suitability information generation step (S130) includes a machine learning step (S180).

The welding judgment program of the embodiment can be implemented using, for example, a scripting language such as ActionScript, JavaScript (registered trademark), Python, or Ruby, or a compiler language such as C, C++, C#, Objective-C, Swift (registered trademark), or Java (registered trademark).

REFERENCE SIGNS LIST 1 Welding judgment system 2, 3, 4 Welding device 11 Resistance welded portion 12, 13 Electrode portion 14 Metal melting portion 20 Welding condition measuring portion 21 Voltage measuring device 22 Non-contact ammeter 100 Welding judgment device 101 CPU
102 ROM
103 RAM
104 Memory unit 105 Input unit 106 Output unit 107 Display unit 110 Target information acquisition unit 120 Welding condition storage unit 130 Appropriateness information generation unit 140 Sequential acquisition unit 150 Collation unit 160 Determination unit 170 Output unit 180 Machine learning unit W1, W2 Workpieces to be welded

Claims (12)

A welding judgment device for judging a welding state in a resistance welding machine that welds workpieces, comprising:
The welding judgment device includes:
an object information acquisition unit that acquires, as welding object information , a shape of the workpiece, a material of the workpiece, and a number of welding points related to the workpiece itself;
an input section for inputting whether the welding quality is acceptable or not;
a welding condition storage unit that acquires a plurality of types of welding conditions for each welding point when the plurality of types of welding conditions controlled by the resistance welding machine in welding the workpieces in response to the welding object information are defined as a current and a voltage, a current application time, a pressure applied to the workpieces, and a resistance value during operation of the resistance welding machine, and that stores, as accumulated information, information in which the acquired plurality of types of welding conditions are associated with the quality of the weld inputted via the input unit ;
an appropriateness information generating unit that generates and accumulates, as appropriateness information, associations between a plurality of types of welding conditions corresponding to the welding object information that are associated with the acceptability of the welding quality by a statistical method using the accumulated information;
a sequential acquisition unit that sequentially acquires, as welding condition information, the same types of welding conditions as the plurality of types of welding conditions controlled by the resistance welding machine when welding the workpiece corresponding to the welding object information;
A collation unit that compares the sequential welding condition information with the appropriate information;
a determination unit that determines whether the sequential welding condition information deviates from the appropriate information and generates a determination result;
and an output unit that outputs the judgment result. The welding judgment device according to claim 1, wherein the statistical method is machine learning using training data in which the welding quality when the workpiece is welded is used as training data and the multiple types of welding conditions are used as input data. The welding judgment device according to claim 1, wherein the statistical method is a multivariate analysis in which the welding quality is an objective variable and the multiple types of welding conditions are explanatory variables. a display unit that displays a time series graph of the welding conditions corresponding to each of the plurality of types of welding conditions included in the suitability information;
2. The welding judgment device according to claim 1, wherein a predetermined fluctuation range is set for the time series graph. a display unit that displays a time series graph of the welding conditions corresponding to each of the plurality of types of welding conditions included in the suitability information;
The welding judgment device according to claim 1 , wherein a threshold value is set for a fluctuation range of the time series graph. The welding judgment device according to claim 1, wherein the judgment unit generates the judgment result using a judgment formula obtained by a statistical method using the judgment result and the suitability information. The welding judgment device according to claim 1, wherein the judgment unit judges whether at least one of the plurality of welding conditions included in the sequential welding condition information deviates from the appropriate information. The welding judgment device according to claim 1 , wherein the resistance welding machine is a plurality of machines, and the welding judgment device is mounted in each of the resistance welding machines . A welding judgment system for judging a welding condition in a resistance welding machine that welds a workpiece, comprising:
The welding judgment system includes:
an object information acquisition unit that acquires, as welding object information , a shape of the workpiece, a material of the workpiece, and a number of welding points related to the workpiece itself;
an input section for inputting whether the welding quality is acceptable or not;
a welding condition storage unit that acquires a plurality of types of welding conditions for each welding point when the plurality of types of welding conditions controlled by the resistance welding machine in welding the workpieces in response to the welding object information are defined as a current and a voltage, a current application time, a pressure applied to the workpieces, and a resistance value during operation of the resistance welding machine, and that stores, as accumulated information, information in which the acquired plurality of types of welding conditions are associated with the quality of the weld inputted via the input unit ;
an appropriateness information generating unit that generates and accumulates, as appropriateness information, associations between a plurality of types of welding conditions corresponding to the welding object information that are associated with the acceptability of the welding quality by a statistical method using the accumulated information;
a sequential acquisition unit that sequentially acquires, as welding condition information, the same types of welding conditions as the plurality of types of welding conditions controlled by the resistance welding machine when welding the workpiece corresponding to the welding object information;
A collation unit that compares the sequential welding condition information with the appropriate information;
a determination unit that determines whether the sequential welding condition information deviates from the appropriate information and generates a determination result;
and an output unit that outputs the judgment result. The welding judgment system according to claim 9 , wherein the resistance welding machine is a plurality of machines . A welding judgment method for a welding judgment device that judges a welding state in a resistance welding machine that welds a workpiece, comprising:
The computer of the welding judgment device
an object information acquisition step of acquiring welding object information relating to the workpiece itself, the shape of the workpiece, the material of the workpiece, and the number of welding points as welding object information;
an input step in which the acceptability of the welding quality is input;
a welding condition storage step of acquiring a plurality of types of welding conditions for each welding point, the plurality of types of welding conditions being controlled by the resistance welding machine when welding the workpieces in accordance with the welding object information, the plurality of types of welding conditions being a current and a voltage during operation of the resistance welding machine, a current application time, a welding pressure applied to the workpieces, and a resistance value, and storing information as accumulated information in which the acquired plurality of types of welding conditions are associated with the quality of the weld inputted through the input step ;
a suitability information generating step of generating and storing, as suitability information, a connection between a plurality of types of welding conditions corresponding to the welding object information, which is associated with the suitability of the welding quality, by a statistical method using the accumulated information;
a sequential acquisition step of sequentially acquiring, as welding condition information, the same types of welding conditions as the plurality of types of welding conditions controlled by the resistance welding machine when welding the workpiece corresponding to the welding object information;
A comparison step of comparing the sequential welding condition information with the appropriate information;
a determination step of determining whether the sequential welding condition information deviates from the appropriate information and generating a determination result;
and an output step of outputting the judgment result. A welding judgment program for a welding judgment device for judging a welding state in a resistance welding machine that welds a workpiece, comprising:
The welding judgment device's computer
An object information acquisition function for acquiring welding object information related to the workpiece itself, the shape of the workpiece, the material of the workpiece, and the number of welding points as welding object information;
An input function for inputting whether the welding quality is acceptable or not;
a welding condition storage function that acquires a plurality of types of welding conditions for each welding point when the plurality of types of welding conditions controlled by the resistance welding machine in welding the workpieces in response to the welding object information are defined as a current and a voltage, a current application time, a welding pressure applied to the workpieces, and a resistance value during operation of the resistance welding machine, and that stores information associated with each of the acquired plurality of types of welding conditions and the acceptability of the welding quality input via the input function as accumulated information;
a suitability information generating function that generates and accumulates a connection between the welding conditions corresponding to the welding object information, which is associated with the suitability of the welding quality, as suitability information by a statistical method using the accumulated information;
a sequential acquisition function for sequentially acquiring, as welding condition information, the same types of welding conditions as the plurality of types of welding conditions controlled by the resistance welding machine when welding the workpiece corresponding to the welding object information;
A collation function for collating the sequential welding condition information with the appropriate information;
a determination function for determining whether the sequential welding condition information deviates from the appropriate information and generating a determination result;
and an output function for outputting the judgment result.

Images