阅读说明：本技术 焊接状态判定装置、焊接状态判定方法和具有程序的介质 (Welding state determination device, welding state determination method, and medium having program ) 是由 大根努 日高一辉 藤井达也 于 2019-05-17 设计创作，主要内容包括：提供一种容易判定焊接状态的焊接状态判定装置。焊接状态判定装置具备如下：取得机构,其取得被供给到脉冲电弧焊接的电极的脉冲电流或脉冲电压的含有下降部、上升部和其间的平坦部的脉冲波形；预处理机构,其使所述平坦部成为指定宽度而对所述脉冲波形进行；判定机构,其根据经过整形的所述脉冲波形,与基于过去的多个经过整形的脉冲波形而作成的正常模式的差异,判定脉冲电弧焊接的状态。(Provided is a welding state determination device which can easily determine a welding state. The welding state determination device is provided with the following components: an acquisition means for acquiring a pulse waveform containing a falling portion, a rising portion, and a flat portion therebetween, of a pulse current or a pulse voltage supplied to an electrode for pulse arc welding; a preprocessing mechanism that performs the pulse waveform so that the flat portion has a predetermined width; and a determination means for determining the state of pulse arc welding based on a difference between the shaped pulse waveform and a normal pattern created based on a plurality of past shaped pulse waveforms.)

1. A welding state determination device is provided with:

An acquisition means for acquiring a pulse waveform including a falling portion, a rising portion, and a flat portion therebetween of a pulse current or a pulse voltage supplied to an electrode for pulse arc welding;

a preprocessing mechanism that shapes the pulse waveform so that the flat portion has a prescribed width;

and a determination means for determining the state of pulse arc welding based on the difference between the shaped pulse waveform and a normal pattern created based on a plurality of past shaped pulse waveforms.

2. The welding state determination device according to claim 1, wherein the preprocessing mechanism expands a width of a portion having an inclination of a prescribed value or less among a plurality of portions included in the pulse waveform.

3. A welding state determination device is provided with:

An acquisition means for acquiring a probability density of a pulse current or a pulse voltage supplied to an electrode for pulse arc welding;

And a determination means for determining the state of pulse arc welding based on the difference between the probability density and a normal pattern created based on a plurality of probability densities in the past.

4. The welding state determination device according to claim 3, wherein the acquisition means acquires the probability density for each period during which the oscillating electrode moves from one end to the other end.

5. A welding state determination device is provided with:

An acquisition means for acquiring a value of a specified position of a pulse current or a pulse voltage supplied to an electrode for pulse arc welding;

And a determination means for determining the state of pulse arc welding based on the difference between the value of the designated position and a normal pattern created based on the values of a plurality of past designated positions.

6. The welding state determination device according to any one of claims 1 to 5, further comprising a creation means for creating the normal mode.

7. A method for determining a welding state, wherein,

Obtaining a pulse waveform including a falling portion and a rising portion of a pulse current or a pulse voltage supplied to an electrode for pulse arc welding and a flat portion therebetween,

shaping the pulse waveform so that the flat portion has a predetermined width,

The state of pulse arc welding is determined based on the difference between the shaped pulse waveform and a normal pattern created based on a plurality of past shaped pulse waveforms.

8. a method for determining a welding state, wherein,

The probability density of the pulse current or pulse voltage supplied to the electrode of the pulse arc welding is obtained,

The state of pulse arc welding is determined based on the difference between the probability density and a normal pattern created based on a plurality of probability densities in the past.

9. A method for determining a welding state, wherein,

The value of a specified position of a pulse current or a pulse voltage supplied to an electrode for pulse arc welding is acquired,

The state of pulse arc welding is determined based on the difference between the value of the designated position and a normal pattern created based on the values of a plurality of designated positions in the past.

10. A medium having a program for causing a computer to function as means for:

An acquisition means for acquiring a pulse waveform including a falling portion, a rising portion, and a flat portion therebetween of a pulse current or a pulse voltage supplied to an electrode for pulse arc welding;

A preprocessing mechanism that shapes the pulse waveform so that the flat portion has a prescribed width;

And a determination means for determining the state of pulse arc welding based on the difference between the shaped pulse waveform and a normal pattern created based on a plurality of past shaped pulse waveforms.

11. A medium having a program for causing a computer to function as means for:

An acquisition means for acquiring a probability density of a pulse current or a pulse voltage supplied to an electrode for pulse arc welding; and

And a determination means for determining the state of pulse arc welding based on the difference between the probability density and a normal pattern created based on a plurality of probability densities in the past.

12. A medium having a program for causing a computer to function as means for:

An acquisition means for acquiring a value of a specified position of a pulse current or a pulse voltage supplied to an electrode for pulse arc welding; and

and a determination means for determining the state of pulse arc welding based on the difference between the value of the designated position and a normal pattern created based on the values of a plurality of past designated positions.

Technical Field

The present invention relates to a welding state determination device, a welding state determination method, and a medium having a program.

Background

Patent document 1 discloses that a neural network is made to learn the difference between normal and abnormal states by analyzing in advance a power spectrum based on at least 1 frequency among a welding current, a welding voltage, and a welding arc sound in each of the normal and abnormal states of welding, and at least one power spectrum among the welding current, the welding voltage, and the welding arc sound during welding is evaluated using the learned neural network to determine whether the normal or abnormal state is present, and at the same time, to determine the abnormal state.

[ Prior Art document ]

[ patent document ]

[ patent document 1] Japanese patent application laid-open No. H10-235490

However, the method disclosed in the above document requires a large amount of data of abnormal patterns under various welding conditions to be prepared in advance by an experiment for reproducing the abnormal patterns, and is difficult to implement.

Disclosure of Invention

The present invention has been made in view of the above problems, and a main object thereof is to provide a welding state determination device, a welding state determination method, and a medium having a program, which facilitate determination of a welding state.

In order to solve the above problem, a welding state determination device according to an aspect of the present invention includes: an acquisition means for acquiring a pulse waveform including a falling portion, a rising portion, and a flat portion therebetween of a pulse current or a pulse voltage supplied to an electrode for pulse arc welding; a preprocessing mechanism that shapes the pulse waveform so that the flat portion has a prescribed width; and a determination means for determining the state of pulse arc welding based on the difference between the shaped pulse waveform and a normal pattern created based on a plurality of past shaped pulse waveforms.

In addition, a welding state determination device according to another aspect of the present invention includes: an acquisition means for acquiring a probability density of a pulse current or a pulse voltage supplied to an electrode for pulse arc welding; and a determination means for determining the state of pulse arc welding based on the difference between the probability density and a normal pattern created based on a plurality of probability densities in the past.

in addition, a welding state determination device according to another aspect of the present invention includes: an acquisition means for acquiring a value of a specified position of a pulse current or a pulse voltage supplied to an electrode for pulse arc welding; and a determination means for determining the state of pulse arc welding based on the difference between the value of the designated position and a normal pattern created based on the values of a plurality of past designated positions.

In addition, a welding state determination method according to another aspect of the present invention is a welding state determination method for determining a state of pulse arc welding based on a difference between a pulse waveform obtained by obtaining a pulse waveform including a falling portion and a rising portion of a pulse current or a pulse voltage supplied to an electrode for pulse arc welding and a flat portion therebetween, shaping the pulse waveform so that the flat portion has a predetermined width, and determining a state of pulse arc welding based on a difference between the shaped pulse waveform and a normal pattern created based on a plurality of past shaped pulse waveforms.

In addition, a welding state determination method according to another aspect of the present invention is a welding state determination method in which a probability density of a pulse current or a pulse voltage supplied to an electrode for pulse arc welding is acquired, and a state of pulse arc welding is determined based on a difference between the probability density and a normal pattern created based on a plurality of probability densities in the past.

In addition, a welding state determination method according to another aspect of the present invention is a welding state determination method for determining a state of pulse arc welding based on a difference between a value of a specified position of a pulse current or a pulse voltage supplied to an electrode for pulse arc welding and a normal pattern created based on values of a plurality of specified positions in the past.

Another aspect of the present invention provides a medium having a program for causing a computer to function as: an acquisition means for acquiring a pulse waveform including a falling portion, a rising portion, and a flat portion therebetween of a pulse current or a pulse voltage supplied to an electrode for pulse arc welding; a preprocessing mechanism that shapes the pulse waveform so that the flat portion has a specified width; and a determination means for determining the state of pulse arc welding based on the difference between the shaped pulse waveform and a normal pattern created based on a plurality of past shaped pulse waveforms.

Another aspect of the present invention provides a medium having a program for causing a computer to function as: an acquisition means for acquiring a probability density of a pulse current or a pulse voltage supplied to an electrode for pulse arc welding; and a determination means for determining the state of pulse arc welding based on the difference between the probability density and a normal pattern created based on a plurality of probability densities in the past.

another aspect of the present invention provides a medium having a program for causing a computer to function as: an acquisition means for acquiring a value of a specified position of a pulse current or a pulse voltage supplied to an electrode for pulse arc welding; and a determination means for determining the state of pulse arc welding based on the difference between the value of the designated position and a normal pattern created based on the values of a plurality of past designated positions.

According to the present invention, the welding state can be easily determined.

Drawings

Fig. 1 is a block diagram showing a configuration example of a system including a welding state determination device according to an embodiment.

Fig. 2 is a block diagram showing a functional configuration example of the welding state determination device.

fig. 3 is a flowchart showing an example of the procedure of the normal mode creation process executed by the welding state determination device.

Fig. 4 is a flowchart showing an example of the procedure of the welding state determination process executed by the welding state determination device.

Fig. 5A is a diagram showing an example of a pulse waveform.

Fig. 5B is a diagram showing an example of a pulse waveform.

Fig. 6A is a diagram showing an example of a pulse waveform before shaping.

Fig. 6B is a diagram showing an example of a pulse waveform before shaping.

Fig. 7A is a diagram showing an example of a pulse waveform after shaping.

fig. 7B is a diagram showing an example of the pulse waveform after shaping.

fig. 8 is a diagram showing an example of the contribution ratio of the principal component.

Fig. 9A is a diagram showing an example of the calculation result of the abnormality degree.

Fig. 9B is a diagram showing an example of the calculation result of the abnormality degree.

Fig. 10 is a diagram showing an example of the calculation result of the gravity center vector of each cluster.

Fig. 11A is a diagram showing an example of the calculation result of the cluster to which the cluster belongs.

fig. 11B is a diagram showing an example of the calculation result of the cluster to which the cluster belongs.

Fig. 12 is a diagram showing an example of a pulse waveform.

fig. 13 is a diagram showing an example of a pulse waveform.

Fig. 14A is a diagram showing an example of the estimation result of the current probability density.

Fig. 14B is a diagram showing an example of the estimation result of the current probability density.

Fig. 15 is a diagram showing an example of the contribution ratio of the principal component.

Fig. 16A is a diagram showing an example of the calculation result of the abnormality degree.

Fig. 16B is a diagram showing an example of the calculation result of the abnormality degree.

Fig. 17 is a diagram showing an example of a pulse waveform.

Fig. 18 is a diagram showing an example of sampling points at a predetermined position.

Fig. 19A is a diagram showing an example of the calculation result of the abnormality degree.

Fig. 19B is a diagram showing an example of the calculation result of the abnormality degree.

[ notation ] to show

1 welding state determination device, 11 data acquisition unit, 13 preprocessing unit, 15 welding state determination unit, 17 normal mode creation unit, 2 database, 8 pulse arc welding device, 81 robot, 83 welding torch, 85 electrode, 9 power supply unit, 100 welding system

Detailed Description

hereinafter, preferred embodiments of the present invention will be described with reference to the drawings. The embodiments described below are exemplified by a method and an apparatus for embodying the technical idea of the present invention, and the technical idea of the present invention is not limited to the following description. The technical idea of the present invention can be variously modified within the technical scope described in the claims.

Fig. 1 is a block diagram showing an example of the configuration of a welding system 100 including a welding state determination device 1 according to the embodiment. Welding system 100 includes pulse arc welding device 8, power supply device 9, and welding state determination device 1.

The pulse arc welding apparatus 8 includes a welding torch 83 supported by a robot arm 81. The welding torch 83 has an electrode 85 for generating an arc, for example, arc welding that realizes mig (metal insert gas) welding, mag (metal Active gas) welding, or the like.

The pulse arc welding apparatus 8 realizes pulse arc welding by using a pulse current and a pulse voltage supplied from the power supply apparatus 9. The power supply device 9 has an ammeter or voltmeter and outputs a detection signal of the pulse current or the pulse voltage to the welding state determination device 1.

The welding state determination device 1 is a computer including a CPU, a RAM, a ROM, a nonvolatile memory, an input/output interface, and the like. The CPU executes information processing in accordance with a program loaded from the ROM or the nonvolatile memory to the RAM. The program may be supplied via an information storage medium such as an optical disk or a memory card, or may be supplied via a communication network such as the internet.

Fig. 2 is a block diagram showing a functional configuration example of the welding state determination device 1. The welding state determination device 1 includes a data acquisition unit 11, a preprocessing unit 13, a welding state determination unit 15, and a normal mode creation unit 17. These functional units are realized by the CPU of the welding state determination device 1 executing information processing in accordance with a program. The database 2 may be provided inside the welding state determination device 1 or may be provided outside.

The data acquisition unit 11 is an example of an acquisition mechanism, the preprocessing unit 13 is an example of a preprocessing mechanism, the welding state determination unit 15 is an example of a determination mechanism, and the normal mode creation unit 17 is an example of a creation mechanism.

Fig. 3 is a flowchart showing an example of the procedure of the normal mode creation process executed by the CPU of the welding state determination device 1. The same processing is performed to create a normal mode for the welding state determination processing described later in advance.

first, the CPU acquires a pulse waveform from a detection signal of a pulse current or a pulse voltage supplied from the power supply device 9 to the pulse arc welding device 8 (S11, which is a process of the data acquisition unit 11). The pulse waveform is divided into units including a falling portion, a rising portion, and a flat portion therebetween. The flat portion may be a base portion or a peak portion.

Next, the CPU shapes the pulse waveform so that the flat portion has a predetermined width (S12, which is a process of the preprocessing unit 13), and stores the shaped pulse waveform in the database 2 (S13). The width of the flat portion of the pulse waveform varies depending on welding conditions, power control, and the like, and therefore, the width of the flat portion is made uniform here for easy comparison between the pulse waveforms.

Next, the CPU creates a normal mode based on the plurality of shaped pulse waveforms stored in the database 2 (S14, which is the processing of the normal mode creating unit 17), and stores the normal mode in the database 2 (S15). In the present embodiment, since the normal mode is created, it is not necessary to prepare a large number of abnormal pulse waveforms less than the number of normal pulse waveforms.

Fig. 4 is a flowchart showing an example of the steps of the welding state determination process executed by the CPU of the welding state determination device 1. The same processing is performed to determine the welding state during welding of the pulse arc welding apparatus 8.

First, the CPU acquires a pulse waveform from a detection signal of a pulse current or a pulse voltage supplied from the power supply device 9 to the pulse arc welding device 8 (S21, which is a process of the data acquisition unit 11). Here, the pulse waveform is cut in the same units as S11 of the normal mode creation process shown in fig. 3.

Next, the CPU shapes the pulse waveform so that the flat portion has a predetermined width (S22 as processing by the preprocessing unit 13). Here, the pulse waveform is shaped so that the flat portion has the same width as S12 in the normal mode creation process shown in fig. 3.

Next, the CPU reads the normal pattern stored in the database 2, calculates the difference between the pulse waveform shaped at S22 and the normal pattern (S23), and determines the welding state based on the calculated difference (S24, which is a process of the welding state determination unit 15). In the present embodiment, since the normal mode is used, the welding state can be easily determined.

A more specific example of the normal mode creation process and the welding state determination process will be described below.

[ pulse waveform shaping ]

A method of detecting an abnormality by preprocessing a pulse waveform, shaping the pulse waveform into a normalized waveform, and then performing pattern matching will be described. Fig. 5A is a diagram showing an example of a normal pulse waveform. Fig. 5B is a diagram showing an example of an abnormal pulse waveform. In this example, the base of the pulse waveform is disturbed.

here, it is considered to intercept the pulse waveform at each pulse. Fig. 6A is a diagram showing a result of overlapping a plurality of pulse amounts by dividing a range from, for example, about 400A which is a fall to about 400A which is a subsequent rise to about 400A with respect to a normal pulse waveform. On the other hand, fig. 6B is a diagram showing a result of dividing the same range for a pulse waveform in which an abnormality occurs in which disturbance occurs at the base portion and superimposing a plurality of pulse amounts. The pulse waveform is extracted in correspondence with the data acquisition units 11, S11, and S21.

In either case, under the influence of the control on the power supply apparatus side relating to the pulse width, there is a mixture of waveforms having different pulse widths, and in this state, it is difficult to directly extract the disturbance of the pulse waveform. Therefore, as shown in fig. 7A and 7B, a process of matching the widths of the pulse waveforms is performed. That is, the steep gradient descending portion and the steep gradient ascending portion of the pulse waveform are made to coincide with each other, and the flat portion having a gentle gradient is expanded to make the pulse waveform substantially uniform in shape.

Specifically, the pulse waveform is subjected to a process of moving average at several points, then calculating the gradient by taking the difference between the preceding and following sampling points, and expanding the width of a portion where the absolute value of the gradient is equal to or less than a predetermined value. In the expansion, linear interpolation or the like is performed between the sampling point and the sampling point. Thereafter, if only the flat portion with a gentle inclination is expanded, the number of sampling points of the lateral width may not exactly match if the steep-inclination falling portion and the steep-inclination rising portion are included, and the entire width of the combination of the steep-inclination falling portion, the portion expanded by the flat portion (base portion) with a gentle inclination, and the steep-inclination rising portion is processed to approximately 100 points.

Fig. 7A shows a calculation result of thinning out the entire pulse waveform to about 100 points after the interpolation process is performed and the interpolation is performed to expand the entire pulse waveform to thousands to tens of thousands of points. Fig. 7B is a diagram showing the calculation result of the same processing performed on the pulse waveform in which the abnormality in the base portion has occurred. From this, it is found that the difference between the normal pulse waveform and the abnormal pulse waveform is emphasized. The above-described shaping of the pulse waveform corresponds to the preprocessing units 13, S12, and S22.

Next, an example of detecting abnormal data by performing principal component analysis using only a normal pulse waveform or using a plurality of pulse waveforms which are mostly normal pulse waveforms will be described. In general welding, it is considered that most of the welding is a normal pulse waveform, and a part of the welding is an abnormal pulse waveform. The following calculations are not problematic as long as the majority of the pulse waveforms are normal.

When the shaped pulse waveform is N (x1, x2, …, xN), 1 pulse x1 ═ x11, x21, …, xp1] T is a vector of p dimensions and about 100 dimensions. That is, when X ═ X1, X2, …, xN ] is expressed in a matrix form, the following expression 1 is given.

[ equation 1]

Here, the average μ and the standard deviation σ of each line are calculated. Which is also a p-dimensional vector. Next, N pulse waveforms x1, x2, …, and xN are normalized by subtracting the average μ and dividing by the standard deviation σ so as to be 0 on average and 1 on standard.

Thereafter, principal component analysis is performed to calculate a reconstruction error, thereby obtaining an abnormality degree. Specifically, as shown in the following expression 2, assuming that m upper-order components of the contribution rate of the obtained principal component vector are u1, u2, …, um, and Ip is a p-row and p-column unit matrix, the abnormality degree α 1(x ') of x ' can be calculated by the following expression 3 using x to (x plus the wavy line sign) obtained by dividing the average μ by the standard deviation σ for each shaped pulse waveform x ' for which the abnormality degree is to be calculated.

[ equation 2 ]

U＝[u，u…，u]

[ equation 3 ]

Fig. 8 is a diagram showing a situation where the contribution rate exceeds 90% when the contribution rate of the principal component using the shaped pulse waveform is calculated and integrated from the upper level and m is 10. This accumulation can determine m.

And displaying the result of calculating the degree of abnormality. Fig. 9A is a diagram showing a result of calculating the degree of abnormality for a normal pulse waveform. On the other hand, fig. 9B is a diagram showing a result of calculating the degree of abnormality with respect to the pulse waveform in which abnormality in which the base portion is disturbed occurs.

Calculation of a principal component vector based on principal component analysis is an example of normal mode generation, and corresponds to the above-described normal mode generation unit 17 and S14. The calculation of the reconstruction error, that is, the calculation of the degree of abnormality is an example of the welding state determination, and corresponds to the welding state determination unit 15 and S24 described above.

The degree of abnormality is calculated as described above, and the abnormality of the welding can be detected by extracting the degree of abnormality. In the above example, the normalization process is performed, but calculation is possible without performing the normalization process.

Even if the pulse waveform is normal, the width and shape of the pulse change depending on the set current, the set voltage, and the power supply control, but the accuracy of the abnormality detection can be improved by shaping the pulse waveform in a standardized manner as in the present embodiment. That is, even under various welding conditions, since the pulse waveform is normalized and shaped, abnormality detection can be performed. In a situation where the average current and the average voltage, which vary depending on the relative position to the workpiece such as the height and the left-right position of the welding torch, have a macroscopic influence, the pulse waveform is normalized and shaped, and therefore, abnormality detection is possible.

In the above example, the welding abnormality is detected by the principal component analysis, but it is considered that the abnormality can be detected by various methods without being limited to the principal component analysis in order to make it possible to clarify the difference from the normal mode by shaping the pulse waveform.

For example, the degree of abnormality may be calculated by shaping only a normal pulse waveform or shaping a plurality of pulse waveforms which are mostly normal, averaging the shaped pulse waveforms, setting the averaged pulse waveform as a normal mode, and calculating the distance between the averaged pulse waveform and the pulse waveform for which the welding state is to be determined. By averaging, even if a small number of abnormal pulse waveforms are included, the influence thereof can be suppressed. Further, the degree of abnormality may be calculated by calculating the mahalanobis distance (マ ハ ラ ノ ビ ス distance) between the plurality of shaped pulse waveforms and the pulse waveform of which the welding state is to be determined.

Next, a method of performing k-means-based clustering after pulse waveform shaping to detect an abnormality will be described. Fig. 10 is a diagram showing the results of calculation of the gravity center vector when clustering is performed in a situation where a normal pulse waveform and an abnormal pulse waveform are mixed, but the normal pulse waveform is more frequent, and the clustering is performed to divide the pulse waveforms into 3 clusters.

Fig. 11A is a diagram showing a calculation of the distance from the centroid vector for a normal pulse waveform, and a calculation of which cluster the closest cluster belongs to. On the other hand, fig. 11B is a diagram showing a distance between the abnormal pulse waveform and the gravity center vector, and which cluster the closest cluster belongs to is calculated.

That is, the welding abnormality can be detected by determining that the cluster assigned to the lower cluster is abnormal. Further, a support vector machine (1 ク ラ ス サ ポ ー ト ベ ク タ マ シ ン) of one kind may be used, or when there are a large number of abnormal data, it is needless to say that a supervised learning method such as a normal support vector machine and a decision tree can be applied.

In the above-described example, the lower base of the pulse waveform is focused on, but the present invention is not limited to this, and for example, the entire pulse waveform including both the peak and the base may be preprocessed from the rise of the pulse waveform to the next rise. Fig. 12 is a result of preprocessing for a normal pulse waveform. In addition to the pretreatment for making the widths of both the crests and the bases uniform, the heights of the crests and the bases were normalized by averaging them to 0.5 for the height of the crest and 0.1 for the height of the base, respectively.

In the normalization for matching the heights, assuming that the average of the peaks is μ P, the average of the peaks is μ B, and the current value at each time is a, the normalized value a ^ (a is given a cabin (hut) symbol) is obtained by the following equation 4.

[ equation 4 ]

In the above example, the pulse current is subjected to the preprocessing, but the preprocessing may be performed for both the pulse current and the pulse voltage. Fig. 13 is a graph showing the result of preprocessing both the pulse current and the pulse voltage.

The left half represents a vector of about 100 dimensions of the current value to be subjected to the preprocessing, and the right half represents a vector of about 100 dimensions of the voltage value to be subjected to the preprocessing, and these vectors are arranged only in the left-right direction and are about 200 dimensions. The voltage value fluctuates up and down due to the influence of the yaw rate and the like, but when the abnormality degree is calculated, the average μ is subtracted so as to be 0 on the average and 1 on the standard deviation, and the result is divided by the standard deviation σ, so that it is considered that the deviation of the voltage value of such a degree does not become a problem.

As described above, pretreatment may be applied not only to the base but also to the peak, and pretreatment may be applied not only to the pulse current but also to the pulse voltage. Here, although the pulse voltage may be shaped simply in the same manner as the pulse current described above, since the fluctuation of the voltage value is large, the sampling point used in the processing of matching the width of the flat portion of the pulse voltage may be a sampling point where the absolute value of the inclination of the current value used in matching the width of the flat portion of the pulse current is small, that is, the sampling point used in shaping the pulse current is the first point from the front, and the processing of matching the width of the flat portion of the pulse voltage may be performed using the sampling point.

[ estimation of probability Density ]

Next, a method of extracting an output pattern of a current value by performing probability density estimation and detecting an abnormality by performing pattern matching will be described. The probability density is estimated, for example, by the following equation 5. n is the sample size, K is the kernel smoothing function, and h is the bandwidth.

[ equation 5 ]

Fig. 14A is a diagram showing the calculation results of overlapping the amount of the multiple yaw strips after probability density estimation for each yaw strip (i.e., for each period during which the electrode that swings moves from one end to the other end) for a normal pulse current. On the other hand, fig. 14B is a diagram showing the calculation results of overlapping the yaw rate amounts a plurality of times after performing probability density estimation for each yaw rate for the pulse current in which the abnormality of the disturbance occurs. The probability density estimation is an example of processing performed by the acquisition means.

By performing the preprocessing of estimating the probability density for each yaw rate, it is found that the difference between the normal pulse current and the abnormal pulse current is emphasized after averaging the influence of the change in the current value of the yaw rate.

Next, an example of detecting abnormal data by performing principal component analysis using the probability density estimation result will be described. Fig. 15 is a diagram showing a situation where the contribution rate of the principal component using the probability density estimation result is calculated and accumulated from the upper level, and when m is 2, the contribution rate exceeds 90%. By integrating in this way, m can be determined.

And displaying the calculation result of the degree of abnormality. Fig. 16A is a diagram showing a result of calculating the degree of abnormality for a normal pulse current. On the other hand, fig. 16B is a diagram showing a result of calculating the degree of abnormality for the pulse current in which the abnormality of the disturbance has occurred.

The degree of abnormality is calculated as described above, and the abnormality of the welding can be detected by extracting the degree of abnormality having a high degree of abnormality. The calculation of the principal component vector based on the principal component analysis is an example of the processing performed by the creation means, and the calculation of the reconstruction error, that is, the calculation of the degree of abnormality is an example of the processing performed by the determination means.

Since the probability density is calculated in a period including one yaw strip, even in a situation where the relative position of the welding torch and the workpiece changes due to the yaw strip, the changes in the current and voltage in one yaw strip can be obtained in a group, whereby it is considered that the probability density pattern can be stably obtained. That is, even in a situation where the relative position of the welding torch and the workpiece changes due to the yaw movement, since the waveform is normalized and shaped, it is possible to detect an abnormality of welding.

[ sampling Point extraction ]

next, a method of detecting an abnormality by extracting an output pattern of a current value using 1 point at the same position of the base of each pulse of a set of pulse waveforms repeated in a pulse current as a sampling point and performing pattern matching will be described.

Fig. 17 is a diagram showing a result of overlapping a plurality of pulse amounts in a normal pulse waveform, for example, in a range from about 400A of a fall to about 400A of a subsequent rise.

here, it is considered to sample a point near the center of the base. Fig. 18 is a graph showing sampling points of about 10000 pulses per pulse, for example, from the 15 th point from the head in fig. 17. The sampling point extraction is an example of processing performed by the acquisition means.

Here, assuming that the sampling points are x1, x2, …, and xN, the average thereof is μ, and the standard deviation is σ, the degree of abnormality with respect to the newly acquired sampling point x' is represented by the following equation 6.

[ equation 6 ]

And displaying the calculation result of the degree of abnormality. Fig. 19A is a diagram showing a result of calculating the degree of abnormality with respect to the extraction point of the normal pulse waveform. On the other hand, fig. 19B is a diagram showing a result of calculating the degree of abnormality for the extracted point of the pulse waveform in which the abnormality of the disturbance has occurred.

by using 1 point per pulse as described above, it is possible to detect welding abnormality while suppressing the processing time. The sampling point extraction is an example of processing performed by the creation means, and the abnormality degree calculation is an example of processing performed by the determination means. Instead of the 1-point, a plurality of points may be extracted, and a pulse voltage may be used instead of the pulse current.