JP2023055168A - Ai evaluation method for spot welding part - Google Patents

Abstract

An AI evaluation method for non-destructively estimating a target variable such as the welding strength of a spot weld is provided.
A third step includes: a first step of generating an alternating magnetic field with a multi-frequency excitation signal; a second step of measuring the detection signal; and a third step of evaluating the state of the weld based on the detection signal; Based on the detection signal using the quantitative value of the first objective variable to be evaluated as comparison data in the stage, the quantitative value of the second objective variable for classification of the evaluation object, and the correlation data between the excitation signal and the detection signal An AI evaluation method for estimating first and second objective variables to be evaluated, wherein a complex amplitude ratio to an excitation signal is obtained from a detection signal as measurement data, and quantitative values of the first and second objective variables and corresponding comparison values are obtained. A first evaluation formula for obtaining a first objective variable from the complex amplitude ratio of the detection signal as data and a plurality of classifications related to the second objective variable are defined, and an evaluation formula for correlation data is defined for each classification and selected. A first objective variable to be evaluated is estimated by a first evaluation formula corresponding to the classification.
[Selection drawing] Fig. 1

Description

The present invention relates to spot welding in which two plate materials are superimposed, electrode rods are pressed and sandwiched from both sides, and a voltage is applied to perform welding. The present invention relates to an AI evaluation method for spot welds, which evaluates the state such as the bonding state.

2. Description of the Related Art Conventionally, spot welding has been widely used as a method for joining metal plates in the field of manufacturing metal products such as automobile manufacturing. For spot welding, for example, as shown in FIGS. 10 and 11, two plate materials 51 and 52 made of metal or the like to be joined are arranged with their joining regions 53 overlapping each other. A voltage is applied from a power source 56 between the pair of electrode rods 54 and 55 in a state in which the pair of electrode rods 54 and 55 are pressed from above and below and brought into contact with each other.
Alternatively, as shown in FIG. 12, the electrode rod 54 is brought into contact with the plate material 51 side, the electrode 55 is installed at a position sufficiently distant from the welded portion of the plate material 52, and the electrode rod 54 is pressurized so that the plate materials 51 and 52 are welded. The current may be applied while the portions 53 are in contact. As a result, a current I is caused to flow from one electrode rod 54 toward the other electrode rod 55 through the bonding region 53, thereby heating and melting both plate materials 51 and 52 in the bonding region 53, and then cooling. A spot weld is thereby formed. In this way, both plate members 51 and 52 are joined to each other at the joining region 53 by spot welding.

By the way, the welding strength of spot welds may be lowered due to various welding defects such as welding failure and insufficient welding. A decrease in the welding strength of such spot welds results in a decrease in the strength of the metal product as a whole.

On the other hand, Patent Document 1 discloses a transmission coil that generates a magnetic field, and a rod-shaped member that is used as a magnetic path for concentrating the magnetic field generated by the spot welding electrode and the transmission coil on the nugget to be inspected. and a plurality of receiving coils for generating an induced electromotive force corresponding to the magnetic field transmitted through the nugget, wherein the transmitting coil is wound around the rod-shaped member, and the plurality of receiving coils are wound around the outer periphery of the transmitting coil. An electromagnetic induction type sensor adapted to be rotated is disclosed.

However, the electromagnetic induction type sensor having such a configuration is for evaluating whether spot welds are good or bad, and cannot evaluate the welding strength of spot welds.

Japanese Patent No. 3317366

On the other hand, in order to evaluate the welding strength of spot welds, there is a method of preparing a test piece and conducting a destructive test using the test piece. However, in actual metal products, metal plate materials are often joined by a large number of spot welds, and it is difficult to evaluate the weld strength by destroying the actual metal products. Therefore, if the weld strength of each spot weld can be evaluated without destroying the metal product, it will be possible to calculate the strength of the entire metal product, which will contribute to the improvement of product safety. .

Further, regarding the metal plate materials to be joined by spot welding, depending on the appearance of the metal product, the plate material on the front side may be visible, but the plate material on the back side may not be visible. Since it is not possible to see the plate material on the back side of such a metal product, it is difficult to confirm whether the plate materials on both the front side and the back side are as designed. Furthermore, although it is possible to confirm the material quality of both plate materials to some extent just by visually recognizing them, it is impossible to evaluate the state of each plate material, that is, the exact material, composition, and bonding state.

In view of the above points, it is an object of the present invention to provide an AI evaluation method for spot welds that non-destructively estimates objective variables such as weld strength and the state of plate materials to be evaluated for spot welds. .

According to the present invention, a plurality of excitation signals having different frequencies are applied to an excitation coil positioned adjacent to a spot weld to be evaluated to generate an alternating magnetic field in the spot weld. a second step of measuring a detection signal generated in the detection coil by the magnetic flux disturbed according to the state of the spot weld; and a third step of evaluating the state of the spot weld based on the detection signal. and, in the third step, the quantitative value of the first target variable to be evaluated, which is set in advance as data for comparison of the spot welds, and the evaluation target of the spot welds. Based on the second-stage detection signal, using correlation data relating to the correlation between the quantified value of the second objective variable for classification and the first-stage excitation signal and the second-stage detection signal, A spot weld AI evaluation method for estimating a first objective variable and a second objective variable to be evaluated for a spot weld, wherein the complex amplitude for the first stage excitation signal is obtained from the second stage detection signal A ratio is obtained as measurement data, and from the quantitative values of the first objective variable and the second objective variable among the measurement data and the complex amplitude ratio of the detection signal as the corresponding comparison data, the first correlation data is obtained. A first evaluation formula for obtaining an objective variable and a plurality of types of classifications related to the second objective variable are defined, an evaluation formula as correlation data is defined for each classification, and then an unknown actual evaluation target is determined. A classification corresponding to the measurement data is selected for each measurement data relating to the correlation data, and a first objective variable to be evaluated is estimated from the measurement data by a first evaluation formula corresponding to the selected classification. , achieved by the AI evaluation method for spot welds.

In the above configuration, the first objective variable to be evaluated for the spot welds is preferably the welding strength of the spot welds.
The second objective variable to be evaluated for the spot-welded portion is preferably the state of each of the front and back plate materials or the welded portion related to the spot-welded portion.
The state of the plate materials on the front and back sides of the spot-welded portion is preferably the material, composition, structure, or bonding state of the plate materials.
The first objective variable may be estimated by selecting an evaluation formula for the first objective variable by classification using the second objective variable to be evaluated for the spot welds.
The first evaluation formula and the evaluation formula corresponding to the second classification are preferably linear analysis, PLS regression analysis, SVM (Support Vector Machine), neural network, or a supervised AI analysis method. is used to define the measured data as explanatory variables.
Preferably, the excitation coil and the detection coil are moved along the surface of the spot weld, and the measurements in the first step and the second step are performed at a plurality of measurement points in the measurement target area including the spot weld. work is done.
The excitation coil and the detection coil are preferably scanned along the surface of the spot weld in one-dimensional or two-dimensional directions so as to pass near the center of the area to be measured, and the measurement is performed.
Preferably, the evaluation in the third stage is performed based on the detection signal obtained by scanning a predetermined number of times near the center of the spot welded portion in two-dimensional scanning.
The excitation signals are preferably generated at five or more different frequencies. The excitation signal preferably has a frequency band of about 5 to 100 times from the lowest frequency to the highest frequency.

According to the present invention, it is possible to provide an AI evaluation method for spot welds that non-destructively estimates objective variables such as weld strength and the state of plate materials, which are objects to be evaluated for spot welds.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram showing the overall configuration of an example of an evaluation device for implementing an AI evaluation method for spot welds according to the present invention; It is a figure which shows another structural example of a sensor. 2 is a flowchart showing an evaluation method by the evaluation device of FIG. 1; FIG. FIG. 2 is an explanatory diagram showing an example of an evaluation method by the evaluation device of FIG. 1; FIG. 2 is a schematic explanatory diagram showing an example of a method for measuring a spot-welded portion by the evaluation device of FIG. 1; FIG. 2 is a schematic explanatory diagram showing another example of a method for measuring a spot welded portion by the evaluation device of FIG. 1; FIG. 3 is a schematic explanatory diagram showing still another example of a method for measuring a spot welded portion by the evaluation device of FIG. 1; 1 shows the relationship between predicted values and measured values of welding strength of spot welds in Example 1, where (A) shows the time of learning and (B) shows the time of verification. 2 shows the relationship between the predicted value and the measured value of the welding strength of the spot welds of Example 2, (A) at the time of learning and (B) at the time of verification. It is a schematic explanatory drawing which shows general spot welding. It is a schematic sectional drawing which shows general spot welding. It is a schematic explanatory drawing which shows the modification of general spot welding.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
FIG. 1 shows the configuration of an embodiment of an evaluation apparatus for implementing the AI evaluation method according to the present invention, and FIG. 2 is a diagram showing another configuration example of the sensor 10. As shown in FIG. As shown in FIG. 1, the evaluation device 1 includes a sensor 10, a measurement section 20, and a data processing section 30. As shown in FIG.

The sensor 10 includes an excitation coil 11 and a detection coil 12 as coils, and a magnetic path forming section 13 . The sensor 10 is housed in, for example, a metallic sensor holding portion 14 with an open upper end in order to block an external magnetic field, and is supported in the sensor holding portion 14 by a non-magnetic gap filler (not shown). . Note that the sensor 10 is arranged upside down so that the open surface 14a of the sensor holding portion 14 is in contact with the surface to be evaluated during measurement.

The magnetic path forming portion 13 includes, for example, a bottom portion 13a, a cylindrical portion 13b, and a shaft portion 13c. The bottom portion 13a supports the cylindrical portion 13b and the shaft portion 13c. An excitation coil 11 and a detection coil 12 are attached to the shaft portion 13c. The shape of the magnetic path forming portion 13 is not limited to that shown in the drawing, and the magnetic path forming portion 13 may be formed by combining the above-described respective parts. In the sensor 10, the detection coil 12 is arranged with respect to the magnetic path composed of the excitation coil 11 and the magnetic path forming portion 13, and the spot welded portion (described later) to be evaluated at the time of measurement is arranged in close proximity. Therefore, the signal detected by the detection coil 12 can be affected according to the magnetic permeability of the spot welded portion. The sensor 10 described above is only one configuration example, and may have the same effect. For example, in FIG. Alternatively, the excitation coil 11 may be arranged outside the measurement coil 12, that is, on the evaluation target 41 side by exchanging the positional relationship between the excitation coil 11 and the measurement coil 12. FIG. The positional relationship between the excitation coil 11 and the detection coil 12 can be freely set according to the object to be evaluated.

The measurement unit 20 includes an oscillation unit 21 , a signal processing unit 22 and a control unit 23 . The oscillator 21 repeatedly generates a signal of a certain frequency and increases or decreases the frequency of the signal step by step. A signal oscillated from the oscillation unit 21 is branched into an excitation signal and a reference signal, and the excitation signal is transmitted to the excitation coil 11 and output to the signal processing unit 22 as a reference signal. The signal processing unit 22 uses the reference signal from the oscillation unit 21 for the detection signal from the detection coil 12 to calculate the temporal change of the detection signal with respect to the excitation signal. At this time, the signal processing unit 22 has a Fourier transform function, and transforms the signal on the time axis into the signal on the frequency axis. The signal processing section 22 also digitizes the detection signal from the sensor 10 and outputs it to the data processing section 30 . The control unit 23 inputs and outputs data and various control signals to and from the data processing unit 30 and controls the oscillation unit 21 and the signal processing unit 22 .

In this case, multiple sensors 10 may be used as shown in FIG. By using a plurality of sensors 10, it is possible to reduce noise and improve the strength of the detection signal level by performing four arithmetic operations on the detection signals, for example, by obtaining a differential signal from two sensors 10. FIG.

In this case, the data processing unit 30 estimates the objective variable to be evaluated from the digital data of the detection signal processed by the signal processing unit 22 . In the following description, the object to be evaluated is the spot welded portion, the measurement data that is the detection signal from the sensor 10 is the explanatory variable, the first objective variable is the welding strength, and the second objective variable is the material of the spot welded portion. , welding strength as a first objective variable, and material quality as a second objective variable.
The data processing unit 30 includes an input/output interface unit 31 that interfaces with the control unit 23, a storage device 32 that includes a main storage device and an auxiliary storage device, an arithmetic device that performs arithmetic processing such as four arithmetic operations, a storage device and and a control device for controlling the arithmetic device, the data processing program is stored in the auxiliary storage device, and the data processing program is developed in the arithmetic device and executed, so that the data processing unit 30 , the storage unit 33 and the estimation unit 34 shown in FIG. 1 are functionally provided. Note that AI evaluation (called artificial intelligence) indicates estimation of objective variables from explanatory variables by calculation methods and algorithms executed by the data processing unit 30 .

The storage unit 33 stores data concerning the correlation between the quantitative value of the objective variable and the estimated value of the objective variable of the evaluation object obtained from the processed detection signal measured using the sensor 10 and the measurement unit 20. . The estimation unit 34 applies an alternating magnetic field to the evaluation object by the sensor 10, and uses the data stored in the storage unit 33 based on the detection signal processed using the measurement unit 20 to estimate the welding strength of the evaluation object. to estimate

The estimating unit 34 can evaluate the welding strength according to the following analysis method, for example, using a quantitative value such as the welding strength obtained in advance as an objective variable to be evaluated. As shown in FIG. 3, the evaluation of the welding strength of spot welds is performed as follows.

(Step ST1)
Acquisition of data for comparison using test pieces prepared in advance will be described. The specimen is prepared as follows. For example, plate materials 51 and 52 made of steel plates or aluminum plates are spot-welded under various conditions to prepare a plurality of test pieces having different materials and welding strengths.

(Step ST2)
In step ST2, a plurality of test pieces having different materials and welding strengths are used as comparison data 42 consisting of complex amplitudes and phases in a predetermined frequency range measured by the sensor 10 by the AI evaluation method described later. get. These multiple test pieces are measured by a known method such as a tensile test or a cross tensile test to obtain the weld strength for each test piece.
These comparison data 42 are the quantitative values obtained by measuring the welding strength of the spot welded portion related to the above-mentioned first objective variable by a conventional quantitative evaluation method or the like, and This is data related to the second objective variable having the material and characteristics of the plate material to be measured. The comparison data 42 is stored in the database of the storage unit 33 in association with the welding strength and material information of the spot welds.

(Step ST3 (pre-learning))
A first evaluation formula 45 for obtaining a first objective variable and a second evaluation formula for obtaining a second objective variable using the quantitative value of the test piece obtained in steps ST1 and ST2 and the obtained comparison data 43 Create 46. An evaluation formula 46 is created for each item of the classification 44 . Step ST3 is a step also called pre-learning.

Here, the first evaluation formula 45 related to the first objective variable for strength and the second evaluation formula 46 related to the second objective variable for classification for each material are linear analysis, SVM (Support Vector Machine), PLS regression analysis, neural network, and supervised AI analysis method. The same method or different methods may be used for the first objective variable and the second objective variable, but one data obtained in steps ST1 and ST2 is used. An evaluation formula can be easily and accurately defined based on these pre-learned comparison data 43 . Up to this point, preparations are made in advance.

(Step ST4)
Next, we move on to the actual evaluation process. An actual evaluation object 41 with unknown welding strength and material used for spot welds is obtained as measurement data 47 consisting of complex amplitudes and phases in a predetermined frequency range by the sensor 10 in the same manner as the test piece in step ST2.

(Step ST5 and Step ST6)
At step ST5, it is selected whether or not to classify the measured data 47. If the measured data 47 is to be classified, the classification is performed at step ST6. If the measurement data 47 are not to be classified (NO), the process proceeds to step ST7.

(Step ST6)
Specifically, in step ST6, the measurement data 47 is subjected to the selection of the classification of the material based on the second evaluation formula 46 for the second objective variable. For example, the classification by the first evaluation formula 46 determines the classification of the measurement data 47 obtained by spot-welding the plate materials 51 and 52 made of two steel plates. Based on the results of the classification, an actual evaluation object 41 that is only between steel plates may be selected from the measurement data 47, for example.

(Step ST7)
After the classification in step ST6, the first evaluation formula 45 for the first objective variable is selected to select whether or not to estimate the strength of the spot weld.

(Step ST8)
When estimating the strength of the spot welded portion in step ST7, in step ST8, using the result of classification by the first evaluation formula 46, for example, the evaluation target 41 of only the steel plate is used as the first objective variable Select the first evaluation formula 45 to perform calculation using the measurement data 47, estimate the welding strength of the spot welded portion, determine the classification and the estimated value of the welding strength in step ST9, and end. can.

(Step ST10)
In the above step ST5, if the measurement data 47 is selected not to be classified, that is, if the measurement data 47 is only the spot welded portion of the plate materials 51 and 52 made of two steel plates, that is, if the evaluation object 41 is iron, If there is, that is, if it is uniquely determined, the classification by the calculation of the first evaluation formula 46 is not necessary. good.

(Step ST11)
If it is not necessary to acquire the strength of the spot welded portion in step ST7, and only the classification of the material used is desired, estimation by calculation for acquiring the welding strength by the first evaluation formula 45 in step ST8 is unnecessary. Therefore, in step ST5, the measurement data 47 are classified into materials by the first evaluation formula 46, and in step ST7, the process proceeds to step ST9 to determine the estimated value of the classification and terminate the process.

The comparison data 43 in step ST2 and the evaluation object 41 in step ST4 are the spot welds measured by the sensor 10, and the objective variables are the welding strength, the state of the front and back plates joined by spot welding, such as material, composition or its binding state.

For comparison data 43 in step ST2, for example, when the objective variable is welding strength, tensile strength is a quantitative value. Estimation by the estimation unit 34 when the evaluation object 41 is the spot welded portion and the target variable is the weld strength will be described in detail below.

The estimating unit 34 can estimate not only the welding strength but also the state of each plate material on the front and back sides of the spot-welded portion. Here, as the state of the plate, it is possible to estimate, for example, the material and composition of the plate, the bonding state of the plate, and the like.

According to the above configuration, the welding strength of the spot-welded portion or the state of the front and back plates related to the spot-welded portion can be accurately predicted without destroying the product or the like including the spot-welded portion. In particular, regarding plate materials that cannot be seen from the outside on the back side, the material, composition or bonding state can be accurately predicted without destroying the product including the spot welded part, and for example, the plate material as designed can be correctly predicted. It can be determined whether it is used or not.

In the estimation unit 34, the real part and the imaginary part of the detection signal after processing output from the measurement unit 20, or the complex amplitude ratio, the first differential value or the second differential value of the amplitude or phase difference with respect to the frequency of the detection signal, etc. are parameters Perform regression analysis and estimation as In the following description, the real part and imaginary part of the detected signal after processing or the complex amplitude ratio are used as parameters.
The estimating unit 34 performs PLS regression analysis or AI analysis based on the quantitative value of the welding strength and the detection signal processed by the measuring unit 20 after applying an alternating magnetic field by the sensor 10 and the spot welded portion. 33 to generate data to be stored. For this reason, the estimation unit 34 may be called an AI analysis unit.

(AI evaluation method)
An AI evaluation method for spot welds using the evaluation device 1 will be described.
in this way,
a first step of applying excitation signals of different frequencies to an excitation coil 11 positioned adjacent to the spot weld to be evaluated to generate an alternating magnetic field in the spot weld;
a second step of measuring a detection signal generated in the detection coil 12 by the magnetic flux that is disturbed according to the state of the spot weld;
and a third step of evaluating the condition of the spot weld based on the detection signal.
In the third step, using correlation data on the correlation between the previously set quantitative value of the objective variable to be evaluated for the spot weld and the excitation signal of the first step and the detection signal of the second step, Based on the detection signal of the stage, the objective variable of the evaluation object of the spot weld is estimated.

Specifically, in the second stage, the complex amplitude ratio to the excitation signal in the first stage is obtained from the detection signal as measurement data, and the measurement data is detected as comparison data corresponding to the quantitative value of the objective variable. Define multiple types of classification from the complex amplitude ratio of the signal, define an evaluation formula as correlation data for each classification, further select the classification corresponding to the measurement data for each measurement data, and select the classification for the selected classification The target variable to be evaluated is estimated from the measured data using the corresponding evaluation formula.

According to the AI evaluation method of the present invention, regarding the measurement data, first, multiple types of classification are defined by comparing the quantitative values of the objective variables such as the welding strength to be evaluated and the state of the plate material with the corresponding comparison data, Define an evaluation formula for each category. Then, a corresponding classification is selected for each measurement data, and the welding strength and the state of the plate material as objective variables to be evaluated are evaluated from the measurement data using an evaluation formula corresponding to the selected classification.
Therefore, by evaluating the welding strength and the state of the plate material at the spot welded portion, especially the state of the plate material on the back side that cannot be seen, the welding strength of the spot welded portion can be increased without destroying the metal product including the spot welded portion. In addition to being able to predict with high accuracy, it is also possible to predict spot weld spatter, insufficient welding, and defective welding in accordance with the welding strength.

Furthermore, since the measurement data can be used to define the classification and generate the correlation data, there is no need to obtain the data for defining the classification and the data for creating the correlation data separately from the measurement data. processing can be simplified. In addition, before evaluating the measured data, the classification to which the measured data should be assigned is selected, and the measured data is evaluated by the evaluation formula corresponding to the selected classification, so that the measured data can be evaluated more accurately. becomes possible.

(Welding strength evaluation method)
Next, a welding strength evaluation method using the welding strength evaluation apparatus 1 will be described.
First, one or more spot welds are prepared. Then, for each spot-welded portion, a quantitative value of welding strength is obtained by, for example, a known method.

The sensor 10 of the evaluation device 1 is brought into contact with the spot welded portion to be evaluated, and while scanning along the surface of the spot welded portion, the following measurement work is performed at each measurement point. That is, under the control of the control unit 23, the frequency is stepped every arbitrary interval frequency (for example, several kHz) in a specified frequency range from the oscillation unit 21, for example, 3 to 100 times or about 5 to 100, for example, about 1 kHz to 100 kHz. A signal of each frequency is oscillated while increasing it exponentially. The estimating unit 34 obtains the correlation between the processed detection signal output from the signal processing unit 22 and the quantitative value of the welding strength, and performs PLS regression analysis. The result is stored in the storage unit 33 . At that time, regression analysis and estimation are performed using only the real part and imaginary part of the detection signal or the complex amplitude ratio as a parameter.

According to the above configuration, by evaluating an object to be evaluated using a plurality of, particularly five or more, excitation signals of a plurality of frequencies that are different from each other, the difference in propagation characteristics of the excitation signals of various frequencies causes the spot welds to The objective variable to be evaluated at various depths can be estimated. Therefore, regardless of the thickness and material of the plate material, the object to be evaluated can always be evaluated more accurately. That is, based on the position information of each measurement point and the internal information in the depth direction at each measurement point, it is possible to acquire the correlation between the predicted value and the actual measurement value by the learning effect of the neural network. At that time, since the measurement data for each frequency changes depending on the thickness and material of the plate material involved in spot welding, in order to reliably detect changes in the amplitude and phase of the measurement data, a plurality of, preferably five or more different types of using an excitation signal with a frequency of

Here, as shown in FIG. 5, the above-described scanning of the spot welded portion by the sensor 10 is performed in two-dimensional directions, that is, in the XY directions along the surface of the plate material 51 on the front side including the spot welded portion 57. . The scanning range is made larger than the size of the spot welded portion 57 (estimated nugget size). For example, when the diameter of the spot welded portion 57 is approximately 8 mm, the scanning range is approximately 10 mm in each of the X and Y directions. Then, while performing so-called uniaxial scanning in the X direction, measurement is performed at measurement points spaced at intervals smaller than the inner diameter of the sensor 10. For example, when the inner diameter of the sensor 10 is about 2 mm, measurement work is performed at measurement points spaced at intervals of about 1 mm. .
Here, after one scan in the X direction is completed, the sensor 10 is shifted in the Y direction by a predetermined interval and scanned again in the X direction, and the above-described measurement operation is performed for each measurement point at the predetermined interval.

Here, as shown in FIG. 5, the scanning of the sensor 10 with respect to the spot welded portion 57 is performed in the XY direction, that is, in a line along the X direction, and in the Y direction at a predetermined interval with respect to each other. In some cases, two scanning results passing through the vicinity of the center of the spot welded portion 57 are extracted from a plurality of scanning results, and measurement and evaluation are performed in a state in which the measurement points are spatially enlarged. However, without being limited to this, the spot welded portion 57 may be randomly measured and evaluated at a plurality of measurement points. As a result, among a plurality of one-dimensional scans obtained by two-dimensional scans of the excitation coil 11 and the detection coil 12, a predetermined number of scans, for example, one scan, of a portion closest to the center of the spot weld Alternatively, by performing the evaluation based on the detection signals obtained by the two scans, the evaluation of the vicinity of the center of the spot welded portion 57 can be reliably performed. Also, as shown in FIG. 6, evaluation is performed based on measurement data obtained by scanning one line in the X direction passing through the center of the spot welded portion 57, that is, scanning in one-dimensional direction. good too.
Furthermore, when evaluating the state of the plate material related to spot welding 57, that is, the material, composition, or bonding state thereof, as shown in FIG. You may make it As a result, measurement data regarding the plate members 51 and 52 can be obtained without being affected by the spot welded portion 57, and the state of the plate members can be evaluated more accurately.

For the spot welded portion 57 to be evaluated, similarly, the sensor 10 is brought into contact with the spot welded portion 57, and under the control of the control unit 23, the oscillation unit 21 emits light within a specified frequency range (for example, about 1 kHz to 100 kHz). A signal of each frequency is oscillated and output to the excitation coil 11 while increasing the frequency stepwise at arbitrary interval frequencies (for example, several kHz). For each signal of each frequency, the signal detected by the detection coil 12 is processed by the signal processing section 22 to be converted into a digital signal and output to the data processing section 30 . Based on the processed detection signal output from the signal processing unit 22 and the data stored in the storage unit 33, the estimating unit 34 estimates the welding strength of the spot welded portion 57 to be evaluated.
For example, when performing n-level evaluation, it is necessary to increase the number of data for the scanning range in accordance with the final target accuracy and dimensions such as the width of the spot welded portion 57 . Regarding the number of spatial data, it is preferable to perform measurement at at least n frequencies in consideration of the imaging of the scanning range, the skin depth of the plate material, and the like.
According to the above configuration, the spot welded portion 57 to be evaluated is evaluated at a plurality of measurement points, respectively, so that the welding strength of the entire spot welded portion 57 and the state of the plate material can be evaluated, resulting in higher accuracy. can evaluate the state of the spot welded portion 57. By scanning the excitation coil 11 and the detection coil 12 one-dimensionally or two-dimensionally, the spot welded portion 57 is efficiently measured at a plurality of measurement points.

The estimating unit 34 performs AI evaluation and estimation using only the real part and the imaginary part of the processed detection signal from the measuring unit 20 as parameters for the spot welded portion 57 to be evaluated. AI evaluation uses any analysis method such as the above-mentioned linear analysis, SVM, PLS regression analysis, neural network and supervised AI analysis method as machine learning, and welding stored in the storage device 32 in the data processing unit 30 It means that the estimating unit 34 executes a program for calculating the first evaluation formula 45 related to the first objective variable of strength and the second evaluation formula 46 related to the second objective variable of material. .
A method for estimating the weld strength will be described below.
That is, the estimated value (f') of the complex representation of the weld strength is expressed as follows.
Estimated weld strength = f'(Real( _eout / _ein ), Ima( _eout / _ein ))
where Real(e _out /e _in ) is the real part of the detection signal (e _out ) with respect to the excitation signal (e _in ), and Ima(e _out /e _{in )} is the detection signal ( e _out ) is the imaginary part. Here, the detection signal with respect to the excitation signal is represented by a complex number represented by a real part and an imaginary part, but it may be represented by absolute amplitude and phase as long as the same complex quantity is represented.
The estimating unit 34 performs AI evaluation such as PLA regression based on the quantitative value of the welding strength of each plate material, the detection signal processed by the measuring unit 20 after applying an alternating magnetic field by the sensor 10, and stores it in the storage unit 33. Generate data to store.

An evaluation example will be described with reference to FIG.
In this evaluation example, the evaluation apparatus 1 is used to evaluate the welding strength of spot welds.
First, in the case of spot welding of the same kind of metal (iron-iron) and in the case of spot welding of dissimilar metals (iron-aluminum), spot welding is performed at one point for each of two plate materials to create a test piece. . Then, learning data is acquired for these spot welds. In this evaluation example, the materials iron and aluminum are classified items.
That is, the sensor 10 is scanned in the XY direction over a range of 10 mm with respect to the spot welded portion with an inner diameter of 8 mm. At this time, at each measurement point in 1 mm steps, the measurement is performed at 6 frequencies, for example, by changing the frequency from 5 kHz to 30 kHz by 5 kHz. Alternatively, at each measurement point, the frequency is changed from 5 kHz to 20 kHz by 1 kHz, and measurement is performed at 16 frequencies.
Then, each test piece is subjected to a tensile test to actually measure the welding strength of the spot-welded portion of the test piece.

As a result, the measured welding strength (actual value) and part of the measurement data are used as learning data, and the correlation between the actual measurement value of welding strength and the measurement data is learned by, for example, a neural network, and the rest of the measurement data is used. Verification of neural network learning results is performed as verification data.

The measurement data measured in the measurement work are processed as follows.
That is, the complex amplitude ratio of the detection signal to the excitation signal is obtained as measurement data D, these measurement data and comparison data are used, and a neural network or the like is implemented as an analysis method. Here, the complex amplitude ratio means the ratio represented by the (absolute) amplitude ratio and the phase difference or the real part and the imaginary part of two signals having mutually different amplitudes and phases.

In addition, the spot-welded portions are classified, for example, into two according to the material of the plate material.
That is, focusing on the correlation between the measured value and the predicted value, the classification of similar measurement data is examined, and two categories of similar metals (iron-iron) and dissimilar metals (iron-aluminum) are created. Thus, the preparatory work for welding strength evaluation is completed.

Next, after preparatory work for welding strength evaluation, that is, after classifying each measurement data into corresponding categories, calculation formulas for estimation for analysis evaluation from measurement data and comparison data for each category, That is, it defines an evaluation formula. Then, after defining the classification by classification and defining the evaluation formula for each classification, analysis using a neural network is performed using the measurement data and comparison data classified for each classification to determine the welding strength. Estimate (see measurements on the right side of FIG. 4). The calculation formula for estimation created at this time becomes an evaluation formula for estimating the welding strength for each of the classified categories.

Subsequently, as another analysis work, the original measurement data is analyzed for classification by a neural network using the above-described categories as explanatory variables. At that time, the neural network in this example randomly divides the measured data into three, uses two-thirds of the measured data for learning, and uses the remaining one-third of the measured data for estimation. This will result in the reclassification of the measurement data, possibly moving some of the measurement data to a different classification. The calculation formula for estimation obtained here becomes the "calculation formula for classification." In the analysis work, these classification formulas are used to classify the measured data into categories.

Finally, the welding strength is estimated by a neural network based on the measurement data reclassified into each category in this way. In this case, a newly created evaluation formula is used as the evaluation formula.

In this way, in the actual analysis work, spot welds are measured first, and the acquired measurement data are classified using the calculation formula for classification. Then, by reviewing the classification, the calculation formula for classification is updated, and the weld strength is obtained by the evaluation formula based on the measurement data after reclassification for each classification.

Next, specific examples will be described.
(Example)
In the examples, the welding strength of the spot-welded portion is evaluated in the case of spot-welding plate materials (iron-iron) made of the same kind of metal and the case of spot-welding the plate materials (iron-aluminum) made of dissimilar metals. The thicknesses of the plate members 51 and 52 of iron and aluminum were both set to 1 mm.
After the preparatory work for evaluating the welding strength of the spot weld described in FIG. The measurement work was carried out at frequencies ranging from 5 kHz to 20 kHz four times.

Based on the obtained measurement data, analysis using a neural network was performed using the classified measurement data and comparison data to estimate the welding strength. At that time, as the number of data, 91 sample data were obtained for iron-iron and 360 sample data for iron-aluminum. Data for verification.
Then, with a neural network, when defining the classification of both samples with iron as the plate material on the front side, the measurement data of each classification in the case of iron-iron and iron-aluminum was reliably obtained with 100% accuracy. I was able to do the classification. After classifying the measured data for each material of the plate material to be spot-welded in this way, the welding strength was evaluated for each classification by a neural network.

As a result, in the case of iron-iron, as shown in FIG. 8(A), the correlation coefficient was 0.999 during learning, and as shown in FIG. It became 0.918. In the case of iron-aluminum, as shown in FIG. 9A, the correlation coefficient is 0.905 during learning and 0 during verification as shown in FIG. 9B. 0.868. As a result, it is possible to predict the welding strength of spot welds for each material of plate materials to be spot welded, and by increasing the measurement data, further learning can be performed by the neural network as the measurement work progresses. It is expected that the accuracy can be improved by

The present invention can be embodied in various forms without departing from its gist. For example, in the above-described embodiment, both the definition of the classification and the definition of the evaluation formula for inferring the weld strength are defined using neural analysis, but this is not the only method of analysis. , for example, linear analysis, SVM (Support Vector Machine), PLS regression analysis and supervised AI analysis techniques may be used for the definition.

Further, in the above-described embodiment, regarding the state of the plate material related to the spot welded portion, the cases of iron-iron and iron-aluminum were described, but the combination of the plate material of iron and other metals is not limited to this, and the spot It is obvious that the method can be applied to the estimation of the weld strength of the weld zone, as well as the estimation of the composition of the plate material or the bonding state thereof.

In addition, when the material used for the spot welded part is determined, only the strength is estimated, and when it is desired to use it for classifying the material, it is clear that only the material classification can be applied alone as necessary.

1 evaluation device 10 sensor 11 excitation coil 12 detection coil 13 magnetic path forming unit 14 sensor holding unit 14a open surface 20 measurement unit 21 oscillation unit 22 signal processing unit 23 control unit 30 data processing unit 31 input/output interface unit 32 storage device 33 storage Part 34 Estimating part 41 Evaluation target (spot welded part)
42 Measurement data 43 Comparison data 44 Classification 51, 52 Plate material 53 Joining area 54, 55 Electrode rod 56 Spot welding part

Claims (11)

a first step of applying a plurality of different frequency excitation signals to an excitation coil positioned adjacent to the spot weld to be evaluated to generate an alternating magnetic field in the spot weld;
a second step of measuring a detection signal generated in the detection coil by the magnetic flux that is disturbed according to the state of the spot weld;
a third step of evaluating the condition of the spot weld based on the detection signal;
contains
In the third step, the quantitative value of the first objective variable to be evaluated, which is set in advance as data for comparison of the spot welded part, and the second objective variable for classifying the spot welded part to be evaluated. Using the correlation data on the correlation between the quantitative value of the first stage excitation signal and the second stage detection signal, the evaluation target of the spot weld based on the second stage detection signal A spot weld AI evaluation method for estimating a first objective variable and / or a second objective variable of
A complex amplitude ratio to the excitation signal in the first stage is obtained as measurement data from the detection signal in the second stage, and the measurement data corresponds to quantitative values of the first objective variable and the second objective variable. A first evaluation formula for obtaining the first objective variable as the correlation data from the complex amplitude ratio of the detection signal as comparison data, and a plurality of types of classifications related to the second objective variable are defined. defining a second evaluation formula for obtaining the second objective variable as the correlation data for each classification,
Next, for each piece of measurement data relating to the unknown correlation data to be actually evaluated, a classification corresponding to the measurement data is selected, and the evaluation is performed from the measurement data using the first evaluation formula corresponding to the selected classification. An AI evaluation method for spot welds, characterized by estimating the first objective variable of a target. 2. The AI evaluation method for spot welds according to claim 1, wherein the first objective variable to be evaluated for the spot welds is the welding strength of the spot welds. 2. The spot welded portion according to claim 1, wherein the second objective variable to be evaluated for the spot welded portion is the front and back plate materials related to the spot welded portion or the state of the spot welded portion. AI evaluation method. AI evaluation of spot welds according to claim 3, wherein the front and back plate materials or the state of the spot welds related to the spot welds are the material, composition, structure, or combination state thereof of the plate materials. Method. The first objective variable is estimated by selecting the evaluation formula for the first objective variable by the classification using the second objective variable to be evaluated for the spot welded portion in the first stage. The AI evaluation method for spot welds according to any one of claims 1 to 4. The first evaluation formula and the second evaluation formula corresponding to the classification use any analysis method of linear analysis, PLS regression analysis, SVM (Support Vector Machine), neural network, and supervised AI analysis method. 6. The AI evaluation method for spot welds according to any one of claims 1 to 5, wherein the measured data are defined as explanatory variables. The excitation coil and the detection coil are moved along the surface of the spot weld, and the measurement operations in the first step and the second step are performed at a plurality of measurement points in a measurement target area including the spot weld. The AI evaluation method for spot welds according to any one of claims 1 to 6, wherein the AI evaluation method is performed. The excitation coil and the detection coil are scanned along the surface of the spot weld in a one-dimensional direction or a two-dimensional direction so as to pass near the center of the measurement target area, and the measurement is performed. The AI evaluation method for spot welds according to claim 7. 9. The evaluation according to the third stage is performed based on the detection signal obtained by scanning a predetermined number of times near the center of the spot welded portion of the two-dimensional scanning, according to claim 8. AI evaluation method for spot welds. The AI evaluation method for spot welds according to any one of claims 1 to 9, wherein the excitation signals are generated at five or more different frequencies. 11. The AI evaluation method for spot welds according to claim 10, wherein the excitation signal has a frequency band of about 5 times to 100 times from the lowest frequency to the highest frequency.

Images