JP5044806B2 - Real-time welding quality judgment device and judgment method - Google Patents

Description

  The present invention relates to a real-time welding quality determination apparatus and determination method, and more particularly to a real-time welding quality determination apparatus and a determination method for performing digital signal processing on a measured value of welding energy during welding and comparing it with a reference value.

  In chemical reactions such as chemical reactions, sintering, and vapor deposition, and physicochemical reactions (including processing), various conditions related to reactions such as pressure, temperature, atmosphere, current, and voltage in order to obtain high-quality products. It is important to control properly. For this reason, with the recent development of microprocessors, personal computers, etc., attempts are being made to digitally process and appropriately control the various conditions.

  Also in welding, for example, in arc welding, analog signals for detected currents and voltages are A / D converted and digitized, processed by digital filters, etc., and inverters, and thus currents and voltages are appropriately controlled. (Patent Document 1, Non-Patent Document 1).

  In addition, in resistance welding, an appropriate welding time is set based on the data on the relationship between the welding time and strength, and the result of the sampling inspection for welding performed at the set welding time is reflected to reflect the welding strength. There is also an invention for judging the above in real time (Patent Document 2).

Japanese Patent Laid-Open No. 7-136776 JP 7-136777 A

“Development of digitally controlled arc welding equipment” by Toshiro Uenono, 100th Anniversary Seminar of the Welding Data System Research Committee sponsored by the Japan Welding Society Welding Data System Research Committee (January 14, 2005)

  However, in recent years, user requirements for welding accuracy, efficiency, and cost have become increasingly severe, and the above-described technology is not necessarily sufficient to meet the strict requirements of users.

  That is, the techniques disclosed in Patent Document 1 and Non-Patent Document 1 process the digitized values of the measured voltage and current, and based on the processing result, the inverter and the like, and thus the voltage and current, etc. This is a technique for controlling the control to obtain good welding.

  However, there are many parameters related to the quality of welding, such as the welding speed, the composition of the shielding gas, the tip angle of the cathode and the arc length in the case of TIG welding, and other non-uniformities depending on the location of the plate or welding rod to be welded. Simply monitoring only the voltage and current is not sufficient to determine the quality of the weld.

  Moreover, the technique currently disclosed by patent document 2 manages the intensity | strength of welding only for welding time for resistance welding. However, such management based on welding time alone is difficult to apply to spot welding, etc., and it is also difficult to detect quality defects due to factors other than welding time, such as defects due to electrode tip deterioration. It is.

  Therefore, for quality control of welding at the actual welding construction site, for example, in the case of arc welding, if the current and voltage are within a predetermined range, it is assumed that there is no problem with the construction, and the welding is finished. After that, non-destructive inspection and destructive inspection of the sampled part are performed. For this reason, it is only after the welding has been completed that some defects have been found in the welding. Then, if it is found that a welding defect has occurred, the defect part is reworked at that time, the defective product is screened and repaired, the electrode tip is polished or replaced, etc. This process is necessary, and there are difficulties in terms of welding accuracy, work efficiency, production cost, and the like.

  Furthermore, depending on the welding system and scale, such as an unmanned welding line, once a defect has occurred, if welding continues as it is, it will be necessary to investigate and rework all subsequent welding locations and products. There is. In this case, the maintenance of welding accuracy, work efficiency, production cost, etc. will be greatly affected.

  For this reason, development of a technique capable of determining quality in real time during welding, such as discovery of occurrence of defects in real time, has been desired.

  The present invention has been made for the purpose of solving the above problems, and measures physical quantities related to welding energy such as current and voltage during welding in each processing unit, the measured physical quantities are digital values, and Signal processing is performed so as to accurately reflect the quality of the welding, and the obtained result is compared with a reference value to determine the welding quality in the processing unit in real time. The invention of each claim will be described below.

The invention described in claim 1
A physical quantity measuring unit that measures a plurality of types of physical quantities related to the welding energy of the welding location during welding;
A plurality of kinds of physical quantities the physical quantity measuring section is measured for each unit processing time if continuous welding, if spot welding for each spot, real-time and in a predetermined procedure, for each type, a large number of digitized Physical quantity digitizing section for sampling data,
A sampling data one-dimensionalization unit that converts each type of digitized sampling data into one-dimensional data in a predetermined procedure including real-time and normalization according to the type and condition of the welding,
A vector expression unit that expresses multidimensional sampling data composed of each type of sampling data, which is made one-dimensional data by the sampling data one-dimensionalization unit, as a one-dimensional sampling vector in a predetermined procedure including real-time and dimensional reduction. When,
A quality determination unit that compares the one-dimensional sampling vector expressed by the vector expression unit with a vector serving as an evaluation reference in real time in a predetermined procedure, and determines the unit processing time or the welding quality of the spot;
It is the real-time welding quality judgment apparatus characterized by having.

The invention of this claim is basically a digital signal processing for measuring a plurality of physical quantities that have a significant effect on welding, such as current and voltage, and characterizing and extracting the quality of the measured values. And determining the quality of the weld by evaluating the processed signal.

Specifically, first, during welding, a plurality of types of physical quantities such as current and voltage related to the welding energy of the welded part are measured, and then, for each type , if these physical quantities are continuous welding, each unit processing time. , if spot welding for each spot, a large number of digitized sampling data. Next, each type of digitized sampling data is converted into one-dimensional data by a procedure including normalization according to the type and condition of welding, and further, multidimensional sampling data consisting of each type of one-dimensional data is dimensionally reduced. A one-dimensional sampling vector is obtained by a procedure including Thereafter, the obtained one-dimensional sampling vector is compared with a vector serving as an evaluation criterion to determine the unit processing time or the welding quality of the spot.

As described above, the real-time welding quality determination device according to the present invention measures and digitizes a plurality of types of physical quantities related to welding energy in real time so that changes, variations, or occurrences of abnormalities can be easily detected. A process including normalization is performed to obtain one-dimensional data, and multidimensional data composed of one-dimensional data of various types of physical quantities is converted to a one-dimensional vector by dimensional reduction. For this reason, since a lot of information is finally made into a one-dimensional vector, it is very easy to compare with a preliminarily created evaluation criterion vector.

  The above processing is not limited as long as it is a physical quantity that can be measured, and thus can be widely applied to various types of welding.

  Therefore, it is possible to determine the quality with high accuracy in real time during welding regardless of the type of welding, for example, if a defect occurs during welding, the occurrence of the defect can be easily known in real time. And the quality control of welding is greatly improved, and the accuracy, efficiency and cost of welding are greatly improved.

  The details will be described later in “DETAILED DESCRIPTION OF THE INVENTION”, but the main terms used in the claims will be described in advance.

  First, the term “real time” as used in this claim includes a case where it is considered to be real time in practice, and the convenience of other processing in the computer, for example, other welding locations in the case where a large number of weldings are performed simultaneously in parallel. Including cases where there is a delay of about 2 or 3 seconds due to restrictions such as checking, the length of unit processing time, and other restrictions determined by the convenience of other machinery and processes in the factory and welding equipment as a whole. It is.

Next, “measuring multiple types of physical quantities related to welding energy” includes not only measuring physical quantities related to electrical energy such as current and voltage, but also light energy such as irradiation energy in laser welding. Measurement of related physical quantities is also included. For this reason, for example, in the case of laser welding, the intensity of a predetermined part of the laser beam is measured with a photocoupler using a light monitor coaxial with the laser irradiator, and the light (reflected light, etc.) from the welding location is measured. It also includes doing.

  The “unit processing time” refers to a time that is a basic unit of measurement that is appropriately set according to the type of welding, the state of signal change, and the like. The number of data per unit time is constant, and in AC welding and pulse welding such as coated arc welding where the signal change is relatively large, the number of data per unit time (1 second) is set to 8000, and the unit processing time. The (basic unit) is set to 0.5 seconds, for example. On the other hand, when the change in the signal is small, the unit processing time is set to a large value, for example, 1 second (that is, the number of data is increased), thereby facilitating the quality determination and improving the accuracy of learning identification described later. Become.

  Further, the specific number of “a large number of digitized sampling data” is preferably 60 or more, and more preferably 90 or more in the case of spot welding. If it is continuous welding, it is 2000 or more per 0.5 second, Preferably it is 4000 or more, More preferably, it is 8000 or more.

  “Measurement of physical quantity” is usually performed continuously and in an analog manner, and an analog value is digitized by a physical quantity digitizing unit. By digitizing, it is possible to easily extract features according to the welding method in the processing by a computer to be described later. If possible, the physical quantity may be acquired directly as a digital value.

  “A large number of digitized sampling data” refers to data sampled in the basic unit, but is not only data with measured values as digital values, but measurements with positive and negative values such as the voltage of an AC power supply. Also included is data that has been processed to take absolute values or remove specific noise.

  “The sampling data is converted into one-dimensional data by the one-dimensionalization unit” means that a large number of digitized sampling data for a plurality of types of physical quantities created by the physical quantity digitizing unit is converted into one-dimensional for each type of physical quantity. To be done.

  The term “normalization” as used in this claim means that when a certain kind of physical quantity, for example, current (S1) is converted into a total of N digitized sampling data from x1 to xN in time series, S1 ( i) It means that N pieces of data represented by = xi / s (where i = 1 to N) are used. Note that s is a standard deviation of digitized sampling data composed of x1 to xN.

  “Multi-dimensional sampling data composed of each type of sampling data made into one-dimensional data by the sampling data one-dimensionalization unit” means other types of physical quantities different from the current, such as resistance (S2). Similarly, the data is made one-dimensional and each sampling point indicates data indicated as a plurality of types of sampling data such as S1 and S2, that is, multidimensional sampling data.

  “Represent as a one-dimensional sampling vector” means that each sampling point indicated by multi-dimensional sampling data is expressed (expressed) as one type of numerical value indicating the quality of welding by the vector expression unit. ).

  “Comparing the expressed one-dimensional sampling vector with a vector serving as an evaluation criterion in real time in a predetermined procedure” means, for example, taking an average value of the one-dimensional sampling vector for each sampling point as a predetermined procedure, etc. This means that the numerical value obtained by performing the process is compared with a vector serving as an evaluation criterion.

  Of course, the real-time welding quality determination apparatus of the present invention includes a recording unit, an input unit, and the like for predetermined determination criteria.

The invention described in claim 2
The physical quantity measuring unit is
The real-time welding quality determination device according to claim 1, wherein the real-time welding quality determination device is a physical quantity measuring unit that measures current and voltage as a plurality of types of physical quantities related to the welding energy of a welding location to be measured.

  In the invention of this claim, it is easy to measure during welding, and it is usually possible to determine the quality of welding in real time by measuring current and voltage controlled to a constant value or a certain range. Therefore, the quality of welding can be managed without installing a special measuring device, and the accuracy, efficiency, and cost of welding are improved.

The invention according to claim 3
The physical quantity digitizing unit
As one of the predetermined procedures, the corresponding resistance is calculated from the current and voltage measured by the physical quantity measurement unit,
The real-time welding quality determination device according to claim 2, wherein the current and the calculated resistance are physical quantity digitizing units that use a large number of digitized sampling data.

  In the invention of this claim, the resistance is obtained from the current and voltage measured during welding and is monitored by the resistance and current, so that it is suitable for quality judgment such as resistance welding.

The invention according to claim 4
The sampling data one-dimensionalization unit
When each type of digitized sampling data is converted into one-dimensional data in a predetermined procedure including normalization according to the type and condition of the welding, (1) any type of sampling data 2. A sampling data one-dimensionalization unit that performs at least one of a weighting process and (2) a process of discarding sampling data at a measurement location where a digital value is known to change suddenly in advance. A real-time welding quality determination apparatus according to claim 3.

  In the present invention, when the one-dimensional data is obtained by a predetermined procedure including normalization, (1) a process of weighting any kind of sampling data according to the type of welding or the like, (2) Since at least one process of discarding the sampling data of the measurement location where the digital value has been known to change suddenly in advance is performed, it is easy to extract the characteristics when a problem occurs in quality in the welding.

  Note that how much weighting is performed on which sampling data is performed based on each type of welding. For example, in the case of resistance welding, the resistance is weighted about three times.

  In addition, as the measurement locations where the digital value is known to change suddenly in advance, for example, in the case of spot welding, the first and last locations can be cited.

The invention described in claim 5
The vector expression unit is:
When expressing multi-dimensional sampling data composed of each type of sampling data that is made one-dimensional data by the sampling data one-dimensionalization unit as a one-dimensional sampling vector in a predetermined procedure including real-time and dimensional reduction, Prior to performing dimension reduction, (1) a process for weighting any kind of sampling data, (2) a process for discarding sampling data at a measurement location where a digital value is known to change suddenly in advance, (3) 5. The vector expression unit for performing at least one of processing using sampling data of a measurement location that has been proved to be preferable for use in quality determination in advance. This is a real-time welding quality judgment device.

  In the present invention, prior to dimensional reduction, (1) a process for weighting any kind of sampling data, and (2) sampling data at a measurement location that has been previously known to have a sudden change in digital value. (3) Since at least one of the processes using the sampling data of the measurement location that has been found to be preferable for quality determination in advance is performed, if there is a quality problem in the welding, Features can be extracted more easily.

  In addition, “(1) a process for weighting any kind of sampling data and (2) a process for discarding sampling data at a measurement location where a digital value is known to change suddenly” are converted into one-dimensional sampling data. If it is done in the part, it is not done in the vector expression part.

  In addition, “processing using sampling data of a measurement location that has been found to be preferable for quality determination in advance” means that most of each type of sampling data that is one-dimensional data, for example, 90% or more, This refers to processing such as adopting the same value when taking the same value, or adopting the center value of the range when distributed in an extremely narrow range.

The invention described in claim 6
The quality determination unit holds a vector that is an evaluation standard according to the type and condition of welding,
The physical quantity measurement unit, the physical quantity digitization unit, the sampling data one-dimensionalization unit, and the vector expression unit are each configured to perform processing according to the type and condition of the welding,
The welding input part to be judged for inputting the type and conditions of the welding for which the user performs quality judgment,
A control unit that causes the physical quantity measurement unit, the physical quantity digitization unit, the sampling data one-dimensionalization unit, the vector expression unit, and the quality determination unit to perform processing according to the type and condition of the welding input to the determination target welding input unit When,
6. The real-time welding quality determination device according to claim 1, wherein the real-time welding quality determination device is provided.

  In the invention of this claim, the quality determination unit holds in advance a vector as an evaluation standard according to the type and condition of welding, and it is not necessary for the user to newly create an evaluation standard. Practicality will be greatly increased. That is, it can be provided to the user as a convenient device in which the evaluation reference data is input in advance.

The invention described in claim 7
The welding energy of the welding location is electrical energy or light energy,
The electrical energy is used in arc welding, resistance welding, electroslag welding, electron beam welding, upset welding, flash welding,
The said light energy is used in laser welding, It is a real-time welding quality determination apparatus in any one of Claims 1-6 characterized by the above-mentioned.

  The welding energy used in the present invention typically includes electrical energy and light energy. Examples of the welding in which electrical energy is adopted include arc welding, resistance welding, electroslag welding, electron beam welding, upset welding, flash welding, and the like, and welding in which light energy is adopted includes laser welding. However, the present invention is not limited to these weldings, and electrical energy and light energy are appropriately selected and adopted according to the type of each welding.

The invention according to claim 8 provides:
A physical quantity measuring unit that measures current and voltage in the spot welding as a plurality of kinds of physical quantities related to the welding energy of the welding spot during spot welding,
A physical quantity digitizing unit that calculates a corresponding resistance from the current and voltage measured by the physical quantity measuring unit, and further converts the current and the calculated resistance into a large number of digitized sampling data in a predetermined procedure; ,
A sampling data one-dimensionalization unit that normalizes the two types of digitized sampling data and normalizes them into one-dimensional data;
Two-dimensional sampling data composed of the sampling data of the current made one-dimensional data by the sampling data one-dimensionalization unit and the sampling data of the resistor weighted 2 to 4 times is subjected to dimensional reduction in real time. Processing, or in addition to this, (1) processing of discarding sampling data at a measurement location that has been previously known to have a sudden change in digital value, and (2) sampling at a measurement location that has been previously determined to be preferable for quality judgment. A vector expression unit that represents a one-dimensional sampling vector in a predetermined procedure including any one of processes using data;
A quality determination unit that compares the expressed one-dimensional sampling vector with a vector serving as an evaluation criterion in real time and in a predetermined procedure, and determines the quality of welding of the spot;
It is the real-time welding quality judgment apparatus characterized by having.

The invention of this claim defines a particularly preferred form of apparatus in the case of spot welding.
In the present invention, the current and voltage are measured during spot welding, the corresponding resistance is calculated from the measured current and voltage, and the current and the calculated resistance are targeted in real time and normalized. Further, since processing such as weighting and dimensional reduction is made into a one-dimensional sampling vector and compared with a vector serving as an evaluation reference, the quality management of spot welding in real time can be accurately performed.

The invention according to claim 9 is:
As a plurality of types of physical quantities related to the welding energy of the welding location during continuous welding, a physical quantity measuring unit that measures current and voltage every unit processing time;
For the current and voltage measured by the physical quantity measuring unit, or the current calculated by the physical quantity measuring unit and the resistance calculated from the current and voltage, a large number of digitized sampling data is obtained in a predetermined procedure in real time. Physical quantity digitization department,
A sampling data one-dimensionalization unit that normalizes the two types of digitized sampling data and normalizes them into one-dimensional data;
A process for performing two-dimensional sampling data composed of the sampling data of the current made one-dimensional data by the sampling data one-dimensionalization unit and the sampling data of the voltage or the resistance in real time and dimensional degeneration; In addition, a vector expression unit that represents a one-dimensional sampling vector in a predetermined procedure including a process of performing weighting of 2 to 4 times on one side prior to dimensional reduction,
A quality determination unit that compares the expressed one-dimensional sampling vector with a vector serving as an evaluation criterion in real time and in a predetermined procedure, and determines the quality of welding in the unit processing time;
It is the real-time welding quality judgment apparatus characterized by having.

The claimed invention defines a particularly preferred form of apparatus in the case of continuous welding.
In the present invention, the current and voltage are measured every unit processing time during continuous welding, and the measured current and the voltage or the resistance calculated from the current and voltage are targeted in real time, and Normalization, weighting, dimensional reduction, etc. are processed into a one-dimensional sampling vector and compared with the evaluation standard vector, so the quality control in real time of spot welding per unit processing time during continuous welding Can be achieved accurately.

The invention according to claim 10 is:
The quality judgment unit, as a vector as an evaluation criterion,
A one-dimensional sampling vector related to a constant number of welds of the same type and the same condition expressed by the vector expression unit of the real-time welding quality determination device according to any one of claims 1 to 9 input from a user; The use of a vector calculated based on a predetermined learning rule from the data relating to the quality of welding corresponding to the one-dimensional sampling vector as a vector serving as an evaluation criterion for welding in the same type and under the same conditions. The real-time welding quality judgment device according to any one of claims 1 to 9, wherein the real-time welding quality judgment device is characterized.

  The invention of this claim uses an evaluation standard created by inputting data relating to welding that the user has in advance. Specifically, it is calculated based on a predetermined learning rule from a one-dimensional sampling vector relating to a certain number of weldings of the same type and under the same conditions and data relating to the quality of welding corresponding to the one-dimensional sampling vector. These vectors are used as the evaluation criteria for welding under the same type and conditions. By using the evaluation standard calculated based on the learning rule using such user data, it is possible to perform more accurate determination reflecting the data and knowledge possessed by the user.

The invention according to claim 11
A real-time welding quality judgment method characterized in that the quality of welding is judged in real time using the real-time welding quality judgment device according to any one of claims 1 to 10.

  The invention of this claim is the invention of the device according to any one of claims 1 to 10 which is an invention of a device, and is regarded as a method invention.

  In the present invention, it is possible to provide a technology capable of determining the quality in real time during welding, such as finding the occurrence of a defect in real time if a defect occurs during welding, so that the quality control of the welding can be provided. This greatly improves welding accuracy, work efficiency, production costs, and the like.

It is a figure which shows the basic process of the real-time welding quality determination of this invention. In the 1st Embodiment of this invention, it is a figure which shows the analog data at the time of making the vertical axis | shaft the current and voltage in spot welding, and making time into a horizontal axis. FIG. 5 is a diagram in which each sampling point is plotted by taking the resistance converted into one-dimensional data on the vertical axis and taking the current converted into one-dimensional data on the horizontal axis. It is a figure which shows the result of having performed the dimension reduction | restoration to one dimension, in order to extract the characteristic relevant to the quality of welding from the two-dimensional information of each sample point shown in FIG. It is a figure which shows the mismatch degree in many spot welding points. It is a figure which shows the mode of the change of the voltage at the time of welding, and the penetration of a welding cross section when changing the speed of TIG welding with 4.0 mm / sec, 5.0 mm / sec, and 6.0 mm / sec. It is the figure which made each the current value and voltage value which were sampled and measured at 8000 Hz at the time of the said TIG welding one-dimensional, and displayed each sampling point on the two-dimensional coordinate. It is a figure which shows the result of having performed the dimension reduction | restoration to one dimension, in order to extract the characteristic relevant to the quality of welding from the two-dimensional information of each sample point shown in FIG. It is a figure which shows the outline of the process of this invention.

  Hereinafter, the present invention will be described based on embodiments. Note that the present invention is not limited to the following embodiments. Various modifications can be made to the following embodiments within the same and equivalent scope as the present invention.

  First, FIG. 1 shows a basic process of real-time welding quality determination according to the present invention. As shown in FIG. 1, in the present invention, while welding is continuously performed, items related to welding energy such as current and voltage are measured, and the measured values are reduced in dimension every processing unit time. Including digital signal processing, extracting as a vector expressing the features associated with the quality of the welding, and comparing the extracted result with the welding standard in real time to determine the quality of the welding in each processing unit time.

(First embodiment)
The first embodiment relates to quality control of welding in real time in continuous spot welding in which a number of spot weldings using resistance welding performed in an automobile production line or the like are continuously performed. For this reason, the quality of welding is judged in units of each spot, unlike welding continuously performed on a line.

1. Digital signal processing:
First, in each spot welding, digital signal processing (DSP: Digital Signal Processing) of a digital value related to current and voltage during welding measured by sampling at high speed will be described.

  FIG. 2 shows an example of analog data when the vertical axis represents current or voltage in spot welding and the horizontal axis represents time. The thin line in FIG. 2 indicates current (A), the thick line indicates voltage (mV), and the horizontal axis indicates the number of alternating cycles of 60 cycles.

  The digital value is, for example, from the analog data shown in FIG. 2, the current and voltage values on the vertical axis at equal intervals with respect to the horizontal axis (time axis) at 90 locations, specifically, 0.001 second. It is obtained by sampling at intervals and digitizing the resulting analog values.

  Further, in spot welding, resistance and voltage have substantially the same waveform, so resistance and current are employed as digital signal processing targets. For this reason, the resistance corresponding to the sampling point (time point) is calculated based on the relationship of (R = V / I) from the digital value of each sampled current value and voltage value.

  Then, each digitized numerical value (sampling data) of the resistance and current at each sampling point is converted into one-dimensional data. Specifically, first, resistance and current dispersion are normalized, and resistance welding is performed in the present embodiment. Therefore, as a process of extracting quality characteristics according to the type of welding, the resistance is weighted, for example, three times. do.

The specific method for normalization is as follows.
That is, S1 (i) = xi / s (where i = 1 to N) when the current is a total of N digitized sampling data (digital values) from x1 to xN in time series. N data S1 indicated by Note that s is a standard deviation of digitized sampling data composed of x1 to xN. The same applies to the resistor (S2).

  As a result, each sampling point is two-dimensionally represented by each digitized one-dimensional data of resistance and current at the sampling point. This is shown in FIG. FIG. 3 is a diagram in which each sampling point is plotted by taking the resistance converted into one-dimensional data on the vertical axis and the current converted into one-dimensional data on the horizontal axis. Although only about 30 sampling points are displayed in FIG. 3 in spite of 90 sampling points, many sampling points are concentrated around the vertical axis being slightly less than 5 and the horizontal axis being slightly more than 4. This is because the coordinate values are the same.

  (1) in FIG. 3 is spot welding data to be determined, (2) is good spot welding data, and (3) is spot welding data in which a defect has occurred.

  Since welding in the present embodiment is spot spot welding, unlike the linear continuous welding shown in the following embodiment, the points plotted particularly near the vertical axis are greatly expanded vertically.

2.1 dimensional sampling vector representation:
Next, a description will be given of a process of expressing the welding characteristics as a one-dimensional sampling vector for each sampling point displayed as two-dimensional data as shown in FIG. Since this process is a process for expressing digitalized (DSP) data as a vector, in this specification, this process is referred to as a “DSPV process (Digital Signal Processing Vector)”.

(1) One-dimensional data covariance between S1 signal (current) and S2 signal (resistance) in order to extract features related to the quality of welding from the two-dimensional sample point information shown in FIG. Using a matrix composed of eigenvectors of a matrix, linear transformation is performed to obtain a one-dimensional vector, thereby performing dimensional reduction to one dimension.

  The result is shown in FIG. In FIG. 4, the horizontal axis indicates the sample numbers in time order, that is, the sampling points during spot welding are arranged in time order, and the vertical axis indicates the origin of the one-dimensional axis when the dimension of the sampling point is reduced. It is the distance from. In addition, the solid line is a good weld, the broken line is a weld subjected to quality determination, and the one-dot chain line is a weld in which a defect has occurred. In FIG. 4, at the sampling point at the start of welding on the horizontal axis, both the target welding and good welding have large values on the vertical axis, and the values on the vertical axis are also small at the end of welding on the horizontal axis. This is because the current and resistance change suddenly at the start and end of spot welding as shown in FIG.

(2) Vector representation (DSPV processing)
In FIG. 4, the difference between the values on the vertical axis between the target weld and the reference good weld or the weld with a defect is defined as the mismatch degree. The degree of inconsistency of the spot welding as a whole is, for example, the average of the sampling points excluding the sampling points up to the fifteenth and the sampling points after the 80th, which average the values of the vertical axes of the sampling points shown in FIG. It is determined as appropriate in consideration of the characteristics of welding, such as taking a sampling point at the center of a stable portion with little change. In the present embodiment, the value of the 41st sampling point is adopted as the degree of inconsistency of the entire spot welding (because it is a result of the DSPV process, hereinafter also referred to as “DSPV result”).

  FIG. 5 shows the results when spot welding near 9000 locations is continuously performed and the degree of inconsistency of the entire spot welding is obtained by the above method for each spot welding. In FIG. 5, the horizontal axis represents spot welding locations arranged in the order of welding, that is, time (shown as hitting point numbers in the figure), and the vertical axis represents the corresponding inconsistency. In FIG. 5, solid spot ellipses from the left and center right are good spot welds, and the right dashed ellipse is a weld spot where good spot welding was not performed.

3. Next, the determination of the quality of welding at each spot of spot welding near 9000 locations shown in FIG. 5 will be described. In this embodiment, the perceptron learning process is used based on the degree of inconsistency (DSPV result) of the entire spot welding shown on the vertical axis in FIG.

The learning process of the perceptron in this embodiment is
WVA = ω1 × DSPV result + ω0
WVA defined as (an index for analyzing the quality of welding, and WVA (abbreviated as “Welding Voice Analyzer”) in this specification) can be judged as good quality when positive, and negative when negative. The quality identification functions ω0 and ω1 are determined using the data shown in FIG. 5 so that the quality can be determined to be defective. Then, the obtained ω0 and ω1 are substituted into the above equation, and the quality of welding is immediately determined from the DSPV result.

A specific procedure will be described below.
First, the data shown in FIG. 5 is appropriately extracted in the first and second columns of Table 1, and the DSPV result and quality determination result of each data are entered. In this table, 10 data are used, and the first to third rows correspond to the three black dots on the left side of FIG. 5, and the fourth to sixth rows are from the center right. Correspond to the three black circle points in the ellipse indicated by the broken line, and the seventh to twelfth rows also correspond to the six black circle points in the ellipse indicated by the right broken line.

  Next, arbitrary numerical values are initially set as ω0 and ω1 in Table 1. And WVA is calculated | required based on a previous formula in an order from the 1st line. If the judgment is negative despite the good quality, or if the judgment is positive despite the occurrence of a defect, positive if it is good, if the defect occurs The next row is calculated by changing ω0 and ω1 in accordance with the learning rules of the perceptron so as to be negative. When the process reaches the last line, the first process is terminated. Next, as the second process, the same process as the first process is performed using the final ω0 and ω1 of the first time.

This process is repeated until all the lines are processed correctly without changing ω0 and ω1, that is, until all the WVA positive and negative and welding quality can be correctly handled (Tables 1 to 6), and ω0 and ω1 are constant. Converge to the value of. In the present embodiment, this state is obtained by the sixth processing, and ω0 = 57.1675 and ω1 = −0.95 are obtained. And using this value,
WVA = −0.95 × DSPV result + 57.1675
To get the formula As a result, if the degree of inconsistency (DSPV result) is known, the welding quality can be determined by the sign of WVA.

  In this embodiment, the perceptron learning process is used. However, other learning processes may be used. For example, if the perceptron learning process does not work, another learning rule such as a Widrow-Hoff learning rule may be used. In addition, when this inventor performed the process using the learning rule of Widrow-Hoff using the same sample as the above, the quality discriminant function was able to be converged in the 7.61 millionth time.

  Thus, according to the present embodiment, welding can be determined in real time based on a predetermined standard. If it is determined that there is a welding defect, the welding is stopped, the welded part is repaired, the cause of the defect is investigated, and a countermeasure is taken according to the investigation result. Specifically, in the first embodiment, as shown in FIG. 5, defects have occurred after the 8000th, and for example, deterioration of the electrode chip is estimated as the cause of the defects. Then, the electrode tip is polished or replaced to remove the cause of the defect.

(Second Embodiment)
The second embodiment relates to determining the quality of welding in continuous welding. Specifically, in TIG welding, the present invention relates to detection of defects when the welding speed changes for some reason.

  FIG. 6 shows changes in voltage during welding and penetration of the weld cross section when the TIG welding speed is changed to 4.0 mm / sec, 5.0 mm / sec, and 6.0 mm / sec under predetermined conditions. Show. (1) in FIG. 6 shows how the voltage changes over time when the welding speed is 4.0 mm / sec, 5.0 mm / sec, and 6.0 mm / sec in order from the top. The vertical axis represents voltage, and the horizontal axis represents elapsed time. (2) in FIG. 6 shows the difference in penetration depth when welding speeds are 4.0 mm / sec, 5.0 mm / sec, and 6.0 mm / sec in order from the left, 90 is a base material, 91 , 92 and 93 indicate the penetration at the above speed.

  In the case of the present embodiment, it can be seen from (1) of FIG. 6 that even if the welding speed is changed, the voltage values look almost the same both in terms of absolute values and changes in values at each time. Note that a constant current power source of 110 A is used for the current. For this reason, although not shown, the change in the current value is the same as (1) in FIG. However, as shown in FIG. 6 (2), it can be seen that the state of penetration greatly varies depending on the welding speed.

  In FIG. 7, in order from the left, (1) 4.0 mm / sec, (2) 5.0 mm / sec, (3) TIG welding at a speed of 6.0 mm / sec, at intervals of 0.5 seconds and 8000 Hz. The current value and voltage value are sampled and measured, and the measured current value and voltage value are digitized and normalized. Further, the voltage is weighted three times, and then the vertical axis represents the voltage converted into one-dimensional data. The horizontal axis shows the result of plotting each sampling point by taking the current converted into one-dimensional data. That is, FIG. 7 corresponds to FIG. 3 of the first embodiment.

  As shown in FIG. 7, at the two-dimensional sampling points, the difference in coordinates due to the difference in welding speed that could not be determined in FIG. In the second embodiment, the number of sampling points is large. However, since this is continuous welding, it is considered that there are few changes in current and voltage. However, since there are many sampling data with the same current and voltage, the number of sampling points displayed in FIG. 7 is much smaller than the actual number.

  Next, FIG. 8 shows a result of performing a process of reducing the dimension and the current value and the voltage value at the sampling point to one dimension for extracting features related to the quality of welding. FIG. 8 corresponds to FIG. 4 in the first embodiment, where the horizontal axis indicates sampling points in sampling order, that is, time, and the vertical axis indicates the origin of the one-dimensional axis when the dimension of the sampling point is reduced. The distance from is shown. In FIG. 8, the solid line of (1), the broken line of (2), and (3) the alternate long and short dash line are in this order when TIG welding is performed at the speeds of 4.0 mm / sec, 5.0 mm / sec, and 6.0 mm / sec. Shows the case.

  In FIG. 8, the upper limit value, the lower limit value, and the center value of each welding speed vibrate up and down violently at a short frequency. In the second embodiment, the center value of the belt-like region at the center of fluctuation is shown. Is adopted as the degree of inconsistency of the entire unit processing time. Assuming that 4.0 mm / sec is the standard for good quality (zero mismatch), the mismatch is 41 for 5.0 mm / sec, and the mismatch is 6.0 mm / sec. For example, if the degree of mismatch in a certain processing unit time is around 41 or around 119, it is determined that the penetration depth is insufficient, that is, a welding defect has occurred. Furthermore, the degree of lack of the penetration depth can be estimated from the absolute value of the mismatch degree.

  As described above, the quality of welding that cannot be grasped only by monitoring a normal current value or voltage value can be clearly grasped by performing digital signal processing based on the present invention.

(Other)
Note that even if the composition or flow rate of the atmospheric gas changes for some reason in TIG welding, or if the wire feed rate also changes, the degree of inconsistency changes, so the presence or absence of welding defects caused by these changes in real time. I can know.

  In this way, the quality of the weld is determined by multiple calculations to determine what kind of mismatch causes a change in the type of physical quantity to be measured and the weighting factor. The cause of the occurrence of the welding defect may be investigated.

  In addition, during the determination of the quality of the welding in real time using the real-time welding quality determination apparatus of the present invention, or newly obtained data in actual use is accumulated as a database, and the real-time of the subsequent welding You may make it use for the quality judgment in.

  Finally, the basic processing flow of the present invention described above will be described with reference to FIG. As shown on the right side in FIG. 9, WVA is determined from data prepared in advance by the user. Next, as shown in the upper center of FIG. 9, a welding process to be actually performed, for example, spot welding, arc welding, or the like is selected.

  Then, the DSPV process is performed in real time for the voltage and current under construction. Thereafter, the obtained DSPV result is compared with the predetermined WVA to determine the welding quality in real time.

  In some cases, as shown on the left side of FIG. 9, the quality can be determined in comparison with the result of DSPV processing in real time using the default WVA.

  The present invention greatly improves the quality control of welding, which greatly improves the accuracy, efficiency and cost of welding, and thus has a very large industrial applicability.

90 Base material 91, 92, 93

Claims (11)

A physical quantity measuring unit that measures a plurality of types of physical quantities related to the welding energy of the welding location during welding;
A plurality of kinds of physical quantities the physical quantity measuring section is measured for each unit processing time if continuous welding, if spot welding for each spot, real-time and in a predetermined procedure, for each type, a large number of digitized Physical quantity digitizing section for sampling data,
A sampling data one-dimensionalization unit that converts each type of digitized sampling data into one-dimensional data in a predetermined procedure including real-time and normalization according to the type and condition of the welding,
A vector expression unit that expresses multidimensional sampling data composed of each type of sampling data, which is made one-dimensional data by the sampling data one-dimensionalization unit, as a one-dimensional sampling vector in a predetermined procedure including real-time and dimensional reduction. When,
A quality determination unit that compares the one-dimensional sampling vector expressed by the vector expression unit with a vector serving as an evaluation reference in real time in a predetermined procedure, and determines the unit processing time or the welding quality of the spot;
A real-time welding quality judgment device characterized by comprising: The physical quantity measuring unit is
The real-time welding quality determination apparatus according to claim 1, wherein the real-time welding quality determination unit is a physical quantity measuring unit that measures current and voltage as a plurality of types of physical quantities related to the welding energy of a welding location to be measured. The physical quantity digitizing unit
As one of the predetermined procedures, the corresponding resistance is calculated from the current and voltage measured by the physical quantity measurement unit,
The real-time welding quality determination device according to claim 2, further comprising: a physical quantity digitizing unit that uses a plurality of digitized sampling data for each of the current and the calculated resistance. The sampling data one-dimensionalization unit
When each type of digitized sampling data is converted into one-dimensional data in a predetermined procedure including normalization according to the type and condition of the welding, (1) any type of sampling data 2. A sampling data one-dimensionalization unit that performs at least one of a weighting process and (2) a process of discarding sampling data at a measurement location where a digital value is known to change suddenly in advance. The real-time welding quality determination apparatus according to claim 3. The vector expression unit is:
When expressing multi-dimensional sampling data composed of each type of sampling data that is made one-dimensional data by the sampling data one-dimensionalization unit as a one-dimensional sampling vector in a predetermined procedure including real-time and dimensional reduction, Prior to performing dimension reduction, (1) a process for weighting any kind of sampling data, (2) a process for discarding sampling data at a measurement location where a digital value is known to change suddenly in advance, (3) 5. The vector expression unit for performing at least one of processing using sampling data of a measurement location that has been proved to be preferable for use in quality determination in advance. Real-time welding quality judgment device. The quality determination unit holds a vector that is an evaluation standard according to the type and condition of welding,
The physical quantity measurement unit, the physical quantity digitization unit, the sampling data one-dimensionalization unit, and the vector expression unit are each configured to perform processing according to the type and condition of the welding,
The welding input part to be judged for inputting the type and conditions of the welding for which the user performs quality judgment,
A control unit that causes the physical quantity measurement unit, the physical quantity digitization unit, the sampling data one-dimensionalization unit, the vector expression unit, and the quality determination unit to perform processing according to the type and condition of the welding input to the determination target welding input unit When,
6. The real-time welding quality determination device according to claim 1, wherein the real-time welding quality determination device is provided. The welding energy of the welding location is electrical energy or light energy,
The electrical energy is used in arc welding, resistance welding, electroslag welding, electron beam welding, upset welding, flash welding,
The real-time welding quality determination apparatus according to claim 1, wherein the light energy is used in laser welding. A physical quantity measuring unit that measures current and voltage in the spot welding as a plurality of kinds of physical quantities related to the welding energy of the welding spot during spot welding,
A physical quantity digitizing unit that calculates a corresponding resistance from the current and voltage measured by the physical quantity measuring unit, and further converts the current and the calculated resistance into a large number of digitized sampling data in a predetermined procedure; ,
A sampling data one-dimensionalization unit that normalizes the two types of digitized sampling data and normalizes them into one-dimensional data;
Two-dimensional sampling data composed of the sampling data of the current made one-dimensional data by the sampling data one-dimensionalization unit and the sampling data of the resistor weighted 2 to 4 times is subjected to dimensional reduction in real time. Processing, or in addition to this, (1) processing of discarding sampling data at a measurement location that has been previously known to have a sudden change in digital value, and (2) sampling at a measurement location that has been previously determined to be preferable for quality judgment. A vector expression unit that represents a one-dimensional sampling vector in a predetermined procedure including any one of processes using data;
A quality determination unit that compares the expressed one-dimensional sampling vector with a vector serving as an evaluation criterion in real time and in a predetermined procedure, and determines the quality of welding of the spot;
A real-time welding quality judgment device characterized by comprising: As a plurality of types of physical quantities related to the welding energy of the welding location during continuous welding, a physical quantity measuring unit that measures current and voltage every unit processing time;
For the current and voltage measured by the physical quantity measuring unit, or the current calculated by the physical quantity measuring unit and the resistance calculated from the current and voltage, a large number of digitized sampling data is obtained in a predetermined procedure in real time. Physical quantity digitization department,
A sampling data one-dimensionalization unit that normalizes the two types of digitized sampling data and normalizes them into one-dimensional data;
A process for performing two-dimensional sampling data composed of the sampling data of the current made one-dimensional data by the sampling data one-dimensionalization unit and the sampling data of the voltage or the resistance in real time and dimensional degeneration; In addition, a vector expression unit that represents a one-dimensional sampling vector in a predetermined procedure including a process of performing weighting of 2 to 4 times on one side prior to dimensional reduction,
A quality determination unit that compares the expressed one-dimensional sampling vector with a vector serving as an evaluation criterion in real time and in a predetermined procedure, and determines the quality of welding in the unit processing time;
A real-time welding quality judgment device characterized by comprising: The quality judgment unit, as a vector as an evaluation criterion,
A one-dimensional sampling vector related to a constant number of welds of the same type and the same condition expressed by the vector expression unit of the real-time welding quality determination device according to any one of claims 1 to 9 input from a user; The use of a vector calculated based on a predetermined learning rule from the data relating to the quality of welding corresponding to the one-dimensional sampling vector as a vector serving as an evaluation criterion for welding in the same type and under the same conditions. The real-time welding quality determination apparatus according to claim 1, wherein the real-time welding quality determination apparatus is characterized.   11. A real-time welding quality judgment method, wherein the quality of welding is judged in real time using the real-time welding quality judgment device according to claim 1.