WO2012050108A1 - 溶接品質判別装置 - Google Patents
溶接品質判別装置 Download PDFInfo
- Publication number
- WO2012050108A1 WO2012050108A1 PCT/JP2011/073379 JP2011073379W WO2012050108A1 WO 2012050108 A1 WO2012050108 A1 WO 2012050108A1 JP 2011073379 W JP2011073379 W JP 2011073379W WO 2012050108 A1 WO2012050108 A1 WO 2012050108A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- welding
- welding quality
- value
- function
- quality
- Prior art date
Links
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B23—MACHINE TOOLS; METAL-WORKING NOT OTHERWISE PROVIDED FOR
- B23K—SOLDERING OR UNSOLDERING; WELDING; CLADDING OR PLATING BY SOLDERING OR WELDING; CUTTING BY APPLYING HEAT LOCALLY, e.g. FLAME CUTTING; WORKING BY LASER BEAM
- B23K9/00—Arc welding or cutting
- B23K9/095—Monitoring or automatic control of welding parameters
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B23—MACHINE TOOLS; METAL-WORKING NOT OTHERWISE PROVIDED FOR
- B23K—SOLDERING OR UNSOLDERING; WELDING; CLADDING OR PLATING BY SOLDERING OR WELDING; CUTTING BY APPLYING HEAT LOCALLY, e.g. FLAME CUTTING; WORKING BY LASER BEAM
- B23K11/00—Resistance welding; Severing by resistance heating
- B23K11/10—Spot welding; Stitch welding
- B23K11/11—Spot welding
- B23K11/115—Spot welding by means of two electrodes placed opposite one another on both sides of the welded parts
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B23—MACHINE TOOLS; METAL-WORKING NOT OTHERWISE PROVIDED FOR
- B23K—SOLDERING OR UNSOLDERING; WELDING; CLADDING OR PLATING BY SOLDERING OR WELDING; CUTTING BY APPLYING HEAT LOCALLY, e.g. FLAME CUTTING; WORKING BY LASER BEAM
- B23K31/00—Processes relevant to this subclass, specially adapted for particular articles or purposes, but not covered by only one of the preceding main groups
- B23K31/12—Processes relevant to this subclass, specially adapted for particular articles or purposes, but not covered by only one of the preceding main groups relating to investigating the properties, e.g. the weldability, of materials
- B23K31/125—Weld quality monitoring
Definitions
- the present invention relates to a welding quality discrimination device that discriminates the welding quality of a welded portion.
- the present invention relates to a welding quality discriminating apparatus suitably used for discriminating welding quality such as the presence or absence of welding defects occurring in spot welding of metal materials.
- time series changes in the interelectrode voltage (welding voltage) and interelectrode current (welding current) of a spot welding machine arranged in the production line can be measured by various measuring instruments. Since the nugget (ellipsoidal melted and solidified portion) of the welded part is formed by heat generation due to electric resistance between the electrodes, when the nugget formation failure occurs, the above welding current and welding Minor changes in voltage occur. In particular, in spot welding, the welding current and welding voltage show a unique transition phenomenon from the initial contact resistance in the initial energization state to the nugget formation / growth process in the later energization period, so these signals are monitored. Therefore, it is thought that the change that leads to the deterioration of the welding quality can be read.
- Patent Documents 1 to 4 have been proposed as devices for evaluating the welding quality of a welded portion using changes in welding current and welding voltage.
- Patent Document 1 a welding current and an interelectrode voltage that change from moment to moment during energization are detected at least every half cycle, and a predetermined same current value for each of the increasing and decreasing processes of the welding current for each half cycle.
- a resistance spot welding quality monitoring device has been proposed which monitors in time series the power difference obtained by subtracting the power applied to the work in the increasing process from the power applied to the work in the process of reducing the inter-electrode power corresponding to the above.
- the apparatus described in Patent Document 1 represents a curve indicating the behavior of the power difference on a cycle diagram in which the power difference is taken on the vertical axis and the number of cycles is taken on the horizontal axis, and an arbitrarily designated cycle on the horizontal axis. Welding quality is evaluated during the welding nugget growth process based on the change in power difference at several measurement reference points.
- Patent Document 2 a means for inputting the shape and material of the material to be welded, a means for detecting the welding current and the voltage between the chips, and calculating the temperature of the material to be welded based on the heat conduction model from both detection values, A means for estimating the estimated nugget diameter A from the calculated temperature distribution, a means for inputting a reference nugget diameter A necessary for ensuring the welding strength of the workpiece, the estimated nugget diameter A and the reference nugget diameter A And a welding quality monitoring device for resistance welding provided with means for comparing the two and outputting the comparison result.
- the temperature of the weld is calculated based on the heat conduction model from the welding current and the welding voltage (voltage between the chips), and the nugget diameter of the weld is estimated. Evaluation results depend on the accuracy of the heat conduction model. Moreover, in order to perform calculation with high accuracy based on this heat conduction model, it is necessary to acquire enormous data such as specific heat and resistance of various materials to be welded up to a high temperature. Also, the more accurate the calculation, the longer the calculation time, and it is not suitable for monitoring the welding quality online.
- Patent Document 3 discloses a welding current measuring means for detecting a welding current, an interchip voltage detecting means for detecting an interchip voltage, and an interchip movement for calculating an apparent interchip dynamic resistance from detection values of both the detecting means.
- the apparatus described in Patent Document 3 does not perform learning (creation of a discrimination boundary for discriminating the welding quality) using data of a welded portion whose actual welding quality is known by a fracture test or the like, the welding quality Cannot be evaluated accurately.
- the criteria for determining the nugget formation time is clear, industrially, the finally obtained nugget diameter or the nugget diameter has a necessary size. It is necessary to determine whether or not Patent Document 3 describes that the nugget size may be determined using the absolute value of the dynamic resistance instantaneous value change rate and its elapsed time (right column, page 43, page 3 of Patent Document 3). (Line 44), but the specific method is not clearly described.
- a first sampling means for sampling a welding voltage or a welding current and supplying a first signal value sequence, and a second voltage value sequence for sampling the welding voltage or welding current are supplied.
- a second sampling means, and one or more artificial second dependent on values of the first signal and the second signal by a generalized discrete point convolution operation from the first signal and the second signal are supplied.
- An operation welding evaluation apparatus has been proposed that includes three means for collecting three sets of values useful for quality monitoring and collecting them into a plurality of groups or regions.
- the apparatus described in Patent Document 4 stores a welding current / welding voltage signal when good welding quality is obtained as a sample signal.
- the apparatus described in Patent Literature 4 calculates an artificial signal generated by multiplying the stored sample signal by a specified coefficient.
- the apparatus described in Patent Document 4 plots the artificial signal and the sample signal three-dimensionally, divides the plotted area into small areas, and counts plot points existing in each small area.
- the apparatus described in Patent Document 4 uses the count number multiplied by a weight set for each small region to obtain a reference signal.
- Patent Document 4 calculates the average and variance of the reference signal, and based on these, calculates the probability density function of the welding current / welding voltage signal input during the online inspection, and the probability density is low (10 -4) determines that welding failures occurred in the case.
- the tip of the electrode wears as the number of hits increases.
- the electrodes wear, the welding current and welding voltage change slightly. Therefore, even if the welding quality is good, the welding current / welding voltage signal varies to some extent.
- the criterion for judging whether or not the welding quality is good in the apparatus described in Patent Document 4 depends on the reference signal under ideal welding conditions, and as described above, there is a slight change in the welding current / welding voltage signal. When it occurs, it is out of the range of the reference signal, and there is a high possibility that the welding quality is erroneously determined.
- the apparatus described in Patent Document 4 calculates an artificial signal according to a specific coefficient from a sample signal under ideal welding conditions.
- the specific coefficient is a value empirically set based on a number of experiments. The coefficient must be set for each welding condition, and the involvement of an expert in statistical analysis is indispensable.
- Patent Documents 5 and 6 have been proposed as devices and methods for evaluating the welding quality of a welded portion using changes in electrode pressure and electrode displacement, for example.
- Patent Document 5 a member to be welded is sandwiched between a pair of electrodes, and in a resistance welding machine that performs welding by energizing the electrodes, a pressure sensor that is provided on one of the electrodes and detects a pressure; Means for calculating a difference area between a set pressure corresponding to a pressure at the time of energization and a time change locus of output of the pressure sensor, and means for outputting and displaying a nugget diameter corresponding to the calculated difference area.
- a characteristic resistance welding machine monitoring device has been proposed.
- Patent Document 6 discloses a welding quality data acquisition step for acquiring correlation data between the displacement amount of the electrode and welding strength using the number of weldings based on the time of a new electrode as a parameter, and a change in the characteristics of the correlation data. And a correlation data change analysis step for analyzing the relationship between the number of weldings, and a welding step for performing welding by estimating welding strength based on the displacement amount of the electrodes and the number of welding times obtained by the correlation data change analysis step And a welding quality estimation method in resistance welding characterized in that the welding strength obtained by the welding process is managed.
- Japanese Unexamined Patent Publication No. 2006-110554 Japanese Unexamined Patent Publication No. Hei 6-170552 Japanese Unexamined Patent Publication No. 10-314956 Japanese National Table 2003-516863 Publication Japanese Patent Laid-Open No. 1-271078 Japanese Unexamined Patent Publication No. 7-290254
- an object of the present invention is to provide a welding quality discriminating apparatus capable of discriminating welding quality relatively easily and accurately.
- the present invention provides at least one of welding current, welding voltage, welding electrode pressure, and welding electrode displacement when welding a discrimination target welding portion whose welding quality is unknown.
- a mapping point obtained by mapping a data point indicating feature information whose components are a plurality of feature amounts obtained based on a physical quantity into a mapping space having a number of dimensions higher than the number of feature amounts constituting the feature information is the mapping It is determined which of the two welding quality regions formed by dividing the space is located, and the welding quality corresponding to the region where it is determined that the mapping point is the welding quality of the discrimination target welding portion
- the acquisition unit for acquiring the feature value, the determination unit that determines the discriminant function that divides the mapping space into two, and the discriminant function determined by the determination unit A discriminating unit that discriminates the welding quality of the discriminating target welded part based on an output value of the discriminant function when the characteristic information of the discriminating target welded part is input, and the acquisition unit
- the discriminant function is determined using the characteristic information of the learning weld that is known, and the discriminant function is characteristic information of the learning weld having one or the other of the two welding qualities.
- the kernel function k (x, x ′) is a function composed of the weight of each feature quantity constituting information, and the matrix K whose elements are given by k (x, x ′) is a semi-definite value.
- x is characteristic information of the learning welding part having the one welding quality
- x ′ is characteristic information of the learning welding part having the other welding quality
- the determination part is The difference between the output value of the discriminant function and the value corresponding to the one welding quality when the characteristic information of the learning welded portion having the one welding quality is input to the kernel function k (x, x ′).
- the feature information of the learning weld having the other welding quality It is defined by the difference between the output value of the discriminant function when input to the kernel function k (x, x ′) and the value corresponding to the other welding quality, and the absolute value of either of the two differences is small.
- the discriminant error becomes smaller and becomes larger as the discriminant error has a positive correlation with the number of dimensions of the discriminant function, and varies according to the weight of each feature quantity constituting the feature information.
- the weight of each feature quantity constituting the feature information is specified for a predetermined regularization parameter so as to minimize the value of the error function consisting of the sum with the regularization term multiplied by the value of the error function
- the learning welded portion having the one welding quality Before the feature information
- the learning welded portion having the one welding quality than the absolute value of the difference between the output value of the discriminant function when inputted to the kernel function k (x, x ′) and the value corresponding to the one welding quality.
- the difference between the output value and the value corresponding to the one welding quality If the misclassification number obtained by adding the number of the welds for learning having the other welding quality that has a smaller absolute value is equal to or greater than a predetermined value, the regularization parameter is adjusted and the error function is again set.
- the weight of each feature amount constituting the feature information is specified so as to minimize the value of the feature information, and the error function value is specified to be minimized when the misclassification number is less than a predetermined value.
- a welding quality discriminating apparatus characterized by determining that the discriminant function is determined by determining that the weight of each feature amount constituting the feature information is adopted as the weight of each feature amount constituting the discriminant function. .
- the welding quality discrimination device provides a welding point for which a welding point of unknown welding quality is mapped to a high-dimensional number mapping space, which is one of the two welding quality regions formed by bisecting the mapping space. It is determined whether it is located in.
- determination apparatus which concerns on this invention is the welding quality corresponding to the area
- the welding quality discriminating apparatus determines a discriminant function indicating a discriminant boundary that bisects the mapping space as follows.
- the welding quality discriminating apparatus is characterized in that, first, feature information on a predetermined regularization parameter is set so as to minimize the value of an error function consisting of a sum of a discriminating error and a regularization term multiplied by the regularization parameter.
- the weight of each feature quantity that constitutes is specified.
- the welding quality discrimination device tentatively employs the weight of each feature amount constituting the feature information specified to minimize the value of the error function as the weight of each feature amount constituting the discrimination function. If the number of misclassifications is less than a predetermined value, it is determined that the specified weight is adopted as the weight of each feature quantity constituting the discrimination function, and the discrimination function is determined.
- the regularization term In order to reduce the value of the error function consisting of the sum of the discrimination error and the regularization term multiplied by the regularization parameter, at least one of the discrimination error and the regularization term needs to be reduced.
- the regularization term when the regularization parameter is large, the regularization term has a large influence on the value of the error function.
- the regularization term varies according to the weight of each feature quantity constituting the feature information. For this reason, when the regularization parameter is large, the weight that makes the regularization term sufficiently small is specified as the weight that minimizes the value of the error function.
- the number of dimensions of the discriminant function and the regularization term have a positive correlation.
- a weight that makes the regularization term sufficiently small is specified as a weight that minimizes the value of the error function, and that the weight is adopted as the weight of each feature quantity constituting the discriminant function.
- the welding according to the present invention adjusts the regularization parameter and again specifies the weight that minimizes the value of the error function. If the regularization parameter is reduced by adjusting the regularization parameter described above, the influence of the regularization term on the value of the error function is reduced, while the influence of the discrimination error on the value of the error function is increased. For this reason, when the regularization parameter is adjusted to be small, the weight that makes the discrimination error smaller than before the adjustment can be specified as the weight that minimizes the value of the error function.
- the discriminant error is the discriminant function output value (hereinafter referred to as “discriminant function corresponding to one weld quality”) when the characteristic information of the welded part for learning having one weld quality is input to the kernel function k (x, x ′). ) And the value corresponding to one welding quality, and the characteristic information of the learning weld having the other welding quality is input to the kernel function k (x, x ′). It is defined by the difference between the output value of the discriminant function (hereinafter referred to as “the output value of the discriminant function corresponding to the other welding quality”) and the value corresponding to the other welding quality.
- the discrimination error includes the difference between the output value of the discriminant function corresponding to one welding quality and the value corresponding to the one weld quality, and the output value of the discriminant function corresponding to the other weld quality and the other weld.
- the absolute value of any difference from the value corresponding to the quality decreases, the absolute value decreases, and when the absolute value increases, the absolute value increases. That is, when the discriminant error decreases, the absolute value of the difference between the output value of the discriminant function corresponding to one welding quality and the value corresponding to one weld quality, or the output value of the discriminant function corresponding to the other weld quality And the absolute value of the difference between the value corresponding to the other welding quality is small.
- the output value of the discriminant function corresponding to one weld quality is smaller than the absolute value.
- the absolute value of the difference from the value corresponding to the other welding quality increases, and the number of learning welds having one welding quality decreases.
- the discriminant function corresponding to the other weld quality is more than the absolute value.
- the absolute value of the difference between the output value and the value corresponding to one welding quality increases, and the number of learning welds having the other welding quality decreases. Therefore, when the determination error is reduced, the number of erroneous determinations is reduced. Therefore, even if the number of misclassifications when the weight specified before adjusting the regularization parameter is adopted as the weight of each feature quantity constituting the discrimination function is a predetermined value or more, the regularization parameter is small. By adjusting so that the weight that the number of misclassifications is less than the predetermined value can be identified, it is determined that the identified weight is adopted as the weight of each feature quantity constituting the discriminant function, and the discriminant function is determined. it can.
- the welding quality determination apparatus determines that the specified weight is adopted as the weight of each feature amount constituting the determination function, and determines the determination function. For this reason, the welding quality discrimination
- the possibility of overlearning increases as the regularization term increases.
- the regularization parameter is increased in the initial stage, and when the weight at which the number of misclassifications is less than a predetermined value cannot be specified, the number of misclassifications can be reduced by adjusting the regularization parameter to be gradually reduced. It is preferable to specify a weight that is less than a predetermined value.
- the above-mentioned discrimination error includes, for example, the difference between the output value of the discriminant function corresponding to one welding quality and the value corresponding to one weld quality, and the output value of the discriminant function corresponding to the other weld quality.
- the value has a positive correlation with the sum of squares of the difference from the value corresponding to the other welding quality.
- the value having a positive correlation with the square sum is, for example, the square root of the square sum.
- the discriminant function is composed of a kernel function k (x, x ') and the weight of each feature quantity, and has no mapping function. For this reason, it is not necessary to calculate a mapping function in order to determine a discriminant function. The calculation amount of the mapping function is enormous. For this reason, the welding quality discriminating apparatus according to the present invention which does not need to calculate the mapping function for determining the discriminant function can determine the discriminant function with a small amount of calculation.
- Patent Document 4 discloses Unlike the described apparatus, it is not necessary to involve an expert in statistical analysis, and the welding quality can be determined relatively easily.
- the concept of “two welding qualities” in the present invention includes, for example, a state where welding is good and a state where welding is poor (for example, when the nugget diameter of the welded portion is larger than a predetermined reference value). In addition to a state in which welding is good and the welding is poor if it is below the reference value), a state where the electrode needs to be replaced and a state where the electrode does not need to be replaced are also included. In addition, the concept of “two welding qualities” includes the same state in terms of poor welding, but also includes different factors of poor welding. In addition, the “value corresponding to the welding quality” in the present invention is a value determined in advance so that one welding quality and the other welding quality can be distinguished, and one welding quality and the other welding quality. And different values.
- the “characteristic information of the learning welded portion” in the present invention means at least one of a welding current, a welding voltage, a welding electrode pressure, and a welding electrode displacement when the learning welded portion is welded. It means feature information having a plurality of feature amounts obtained based on one physical quantity as components. Furthermore, the “plurality of characteristic quantities obtained based on physical quantities” in the present invention is obtained from at least one physical quantity itself among welding current, welding voltage, welding electrode pressure, and welding electrode displacement. The feature amount obtained from the calculation result using the plurality of physical quantities is also included. For example, “a plurality of feature amounts obtained based on physical quantities” in the present invention includes a feature amount obtained from an energization resistance that is a result of dividing a welding voltage by a welding current.
- the discriminating unit bisects the mapping space and a mapping point obtained by mapping the data point indicating the characteristic information of the discriminating target welding portion together with the welding quality of the discriminating target welding portion and the mapping space.
- a certainty factor of the discrimination result of the discrimination target welded portion expressed by a linear distance from the discrimination boundary is calculated.
- the reliability of the discrimination result of the discrimination target welded portion is calculated together with the welding quality of the discriminated discrimination target welded portion. For this reason, for example, when it is determined that the welding is in a good state but the certainty level is low, a re-inspection of the welding quality by another device is performed, or the welding is in a poor state uniformly. It is possible to prevent the possibility of defective welds flowing out.
- the welding quality determination device According to the welding quality determination device according to the present invention, the welding quality can be determined relatively easily and accurately.
- FIG. 1 It is a schematic block diagram of the welding quality discrimination
- FIG. 1 It is a figure which shows an example of the evaluation result of the discrimination test 1. It is a figure which shows the example which further divided what the welding quality is good welding into several welding quality. An example in which the welding quality is poorly welded is divided into a plurality of welding qualities. It is a schematic block diagram which shows the modification of the welding quality discrimination
- FIG. 1 It is explanatory drawing explaining the method of determining the actual welding quality in the discrimination
- FIG. 1 is a schematic configuration diagram of a welding quality discrimination device 100 according to the present embodiment.
- the welding quality determination device 100 includes an acquisition unit 1 for acquiring a feature value, a determination unit 2 that determines a discrimination function, and a determination unit 3 that determines a welding quality.
- the acquisition unit 1 includes a current / voltage measuring device 11 and a coil (toroidal coil) as detection means for detecting a welding current and a welding voltage when spot welding the welded parts W of the metal workpieces M1 and M2. 12.
- the current / voltage measuring device 11 is electrically connected to each of the welding electrode E1 and the welding electrode E2 constituting the spot welder, thereby measuring a change in welding voltage with time. Further, the current / voltage measuring device 11 is connected to the coil 12 arranged so as to surround the shank S2 connected to one electrode E2, thereby measuring a change in welding current with time.
- the acquisition unit 1 also includes a feature amount extraction unit 13 that extracts a feature amount based on the welding current and / or welding voltage detected by the detection unit (current / voltage measuring device 11 and coil 12).
- the feature amount extraction unit 13 of the present embodiment extracts feature amounts based on both the welding current and the welding voltage.
- the feature quantity extracted by the feature quantity extraction unit 13 includes, for example, fractal dimension analysis or Fourier analysis in order to express the characteristics of the time series change corresponding to the welding quality of the welding current and welding voltage detected by the detection unit.
- the result of applying signal processing such as wavelet analysis to the signal waveforms of the welding current and welding voltage can be used.
- the number of feature amounts constituting the feature information is not limited as long as it is plural.
- the feature quantity extraction means 13 of this embodiment applies a fractal dimension analysis to the signal waveforms of the welding current and welding voltage, and extracts the feature quantity.
- the fractal dimension indicates the degree of geometric complexity of a geometric structure when the time series signal waveform is regarded as a geometric structure. A larger fractal dimension means that the time-series signal waveform is more complicated.
- Equation (A) The fractal dimension d of a certain time series signal waveform is given by the following equation (A).
- S means a data string of a signal waveform
- ⁇ means a box size
- N ⁇ (S) means the number of boxes necessary for covering the signal waveform.
- First step An appropriate box size ⁇ larger than 0 is set.
- Second step As shown in FIG. 2A, the number of boxes N ⁇ (S) necessary for covering the signal waveform S is counted (in FIG. 2A, the box size ⁇ 4 An example of covering a signal waveform is shown).
- Third step As shown in FIG. 2A, the box size ⁇ is changed (in the example shown in FIG. 2A, the step size is changed step by step to ⁇ 1 , ⁇ 2 , ⁇ 3 , ⁇ 4 ).
- the number of boxes N ⁇ (S) corresponding to each box size ⁇ is calculated by repeating the first step and the second step.
- Fourth step As shown in FIG. 2B, a graph is plotted in which ln ⁇ is plotted on the horizontal axis and lnN ⁇ (S) is plotted on the vertical axis.
- Fifth step As shown in FIG. 2B, an approximate straight line L is applied to the graph drawn in the fourth step by the least square method or the like, and the inclination is calculated as the fractal dimension d.
- the feature amount extracted by the feature amount extraction unit 13 will be described below in order to more fully express the characteristics of the time series change corresponding to the welding quality of the welding current and welding voltage detected by the detection unit. It is also possible to employ such feature quantities.
- the signal waveforms of the welding current and the welding voltage are shown in FIG. 12A from a state where the initial contact resistance in the initial energization occurs to the nugget formation / growth process in the later energization when viewed macroscopically. Shows a unique transition phenomenon. When a nugget formation failure occurs, a large change in the signal waveform of the welding current and welding voltage is not always seen. It is thought that has occurred. Therefore, among the changes in the signal waveform, macroscopic (basic) changes are expressed by fitting an approximate curve (higher order curve, spline curve, etc.), and minute changes are expressed by the error of the approximate curve with respect to the signal waveform. That is, it is considered that the information of the signal waveform can be expressed in more detail by decomposing the change of the signal waveform into two elements.
- feature amounts are extracted as follows.
- an approximate curve is applied to the actual data of the signal waveform (in the example shown in FIG. 12 (a), the welding voltage).
- a seventh-order curve is applied.
- the error of the approximate curve for the signal waveform is defined as a population, and the cumulative density distribution of this population is expressed by the following equation (B).
- the maximum likelihood estimation method is applied to the density function (likelihood function) represented by the following equation (C) to determine the degree of freedom ⁇ that is an unknown parameter.
- the likelihood function represented by the equation (C) indicates how much the Student's t distribution represented by the equation (B) corresponds to the cumulative density distribution of the real data population when a value at a degree of freedom ⁇ is designated. It is a function to measure whether the fit is good. Specifically, in the above-described maximum likelihood estimation method, the degree of freedom ⁇ is determined so as to maximize the log likelihood function obtained by taking the logarithm of the likelihood function represented by the formula (C). The degree of freedom ⁇ determined in this way is that the Student's t distribution represented by the formula (B) is an estimated value of the degree of freedom ⁇ that best fits the cumulative density distribution of the population of actual data. Conceivable.
- z represents actual data of feature information (welding current and welding voltage), and ⁇ (•) represents a gamma distribution.
- ⁇ means a position parameter, and ⁇ means a scale parameter ( ⁇ > 0).
- ⁇ and ⁇ are values that can be estimated from the cumulative density distribution of the population.
- the coefficient of the approximated curve obtained in the above (1) parameter eg, n following approximate curve P 0 t + P 1 t 2 + ⁇ P n-1 t each coefficient n P 0 ⁇ P n-1
- the determining unit 2 determines a discriminant function indicating a discriminant boundary for discriminating the welding quality of the discriminating target weld with unknown welding quality.
- This discrimination boundary is a feature having as a component a plurality of feature amounts obtained based on the welding current and / or welding voltage (both welding current and welding voltage in the present embodiment) when welding the discrimination target weld.
- a mapping space having a number of dimensions higher than the number of features constituting the information (vector) is represented by two welding qualities (hereinafter, one of the two welding qualities is “welding quality A” and the other welding quality is “welding quality”. B ”).
- the welding quality A and the welding quality B are different welding qualities set in advance by a user of the welding quality discrimination device 100 or the like.
- the welding quality A and the welding quality B can be, for example, a state where welding is good and a state where welding is poor.
- the determining unit 2 determines the discriminant function using the characteristic information of the learning welded part that is known to have either the welding quality A or the welding quality B.
- the feature information of the learning welding part input to the determination unit 2 is obtained by using, for example, the feature amount extracted by the feature amount extraction unit 13 of the acquisition unit 1 described above.
- the characteristic information of the learning welded portion having the welding quality A, the characteristic information of the learning welded portion having the welding quality B, and the characteristic information of the above-described discrimination target welded portion are composed of the same type of characteristic amount. Whether the learning welded portion has the welding quality A or the welding quality B is known by, for example, extracting the feature amount from the learning welded portion and then breaking the learning welded portion and evaluating the nugget diameter. be able to.
- each feature amount constituting the feature information of each learning weld portion having the welding quality A and each feature amount constituting the feature information of each learning weld portion having the welding quality B are:
- the determination unit 2 stores the identifier of the learning welded part and the welding quality of the learning welded part.
- the value of each feature quantity constituting the feature information of each learning weld is normalized so as to fall within a range of 0 to 1.
- the discriminant function f (x) determined by the determining unit 2 is expressed by the following formula (1).
- “w” indicates weight information (vector) whose component is the weight of each feature quantity constituting the feature information.
- X in the equation (1) indicates characteristic information (vector) of the learning welded portion having the welding quality A or the welding quality B.
- ⁇ ( ⁇ ) represents a mapping function that maps data points (points at the tips of vectors) indicating feature information in the mapping space and has positive definiteness.
- a mapping function having positive definiteness includes a Gaussian distribution function.
- the discriminant function f (x) expressed by the following equation (2) is used as the discriminant function so that the discriminant function f (x) can be determined with a small amount of calculation.
- the discriminant function f (x) means the discriminant function f (x) expressed by the following formula (2).
- ⁇ indicates the weight of each feature amount constituting the feature information.
- k (x, x ′) represents a kernel function in which a matrix K whose elements are given by k (x, x ′) is a semi-definite value.
- X described in the equation (2) and thereafter indicates the feature information (vector) of the learning welded portion having the welding quality A.
- x ′ represents the feature information (vector) of the weld for learning having the welding quality B.
- the matrix K whose element is given by k (x, x ′) is the kernel function obtained when the feature information x of the learning weld having the welding quality A is input to the kernel function k (x, x ′).
- This is a matrix having the output value and the output value of the kernel function obtained when the learning weld feature information x ′ having the welding quality B is input to the kernel function k (x, x ′) as elements.
- kernel functions k (x, x ′) are examples of kernel functions k (x, x ′) whose elements are given by kernel functions k (x, x ′) and are positive semidefinite. . Further, as other examples of the kernel function k (x, x ′) in which the matrix K given by the kernel function k (x, x ′) is a semi-definite value, there are a sigmoid function and a Gaussian function represented by the following equations. .
- f (•) indicates an arbitrary function
- q (•) indicates a non-negative coefficient polynomial
- k a ( ⁇ , ⁇ ) and k b ( ⁇ , ⁇ ) denote arbitrary kernel functions
- subscripts a and b denote identifiers of learning welds
- ⁇ denotes a gain of the sigmoid function
- the above equation (2) is derived as follows.
- the following formula (4) is derived.
- d indicates the number of feature quantities constituting the feature information. If the number of feature amounts constituting the feature information is sufficiently increased, the following equation (5) is derived from the above equation (1).
- the weight information w is expressed by the following equation (6).
- the above formula (2) is derived from the above formula (1) using the above formula (6).
- the discriminant function f (x) in the above equation (2) is a function in which the number of dimensions is affected by the number of feature amounts constituting the feature information of the learning weld.
- the determination unit 2 first determines whether or not the number of misidentifications is less than a predetermined value (step S1 in FIG. 4).
- the misidentification number is the output of the discriminant function f (x) when the feature information x of the learning welded portion having the welding quality A is input to the kernel function k (x, x ′) of the discriminant function f (x).
- the output value of discriminant function corresponding to welding quality A is used to determine the number of learning defects having the welding quality A having a smaller absolute value of the difference from the value corresponding to the quality B and the characteristic information x ′ of the learning welded portion having the welding quality B.
- the output value of the discriminant function f (x) corresponds to the output value of the discriminant function f (x) and the welding quality B.
- the feature information x of the learning welded portion having the welding quality A or the feature information x ′ of the learning welded portion having the welding quality B is used as the kernel function k (x, x ′) of the discriminant function f (x).
- the weight ⁇ of each feature quantity constituting the feature information of the discriminant function f (x) is set to an arbitrary value (for example, 1).
- the value corresponding to the welding quality A is 1 and the value corresponding to the welding quality B is -1.
- the value corresponding to the welding quality A and the value corresponding to the welding quality B are determined in advance by a user or the like of the welding quality discriminating apparatus 100 and are different from each other so that the welding quality A and the welding quality B can be discriminated. Is done.
- the determination unit 2 determines that the number of erroneous determinations is equal to or greater than a predetermined value, the determination unit 2 minimizes the value of an error function that is a sum of the determination error and the regularization term ⁇ T K ⁇ multiplied by the regularization parameter ⁇ .
- the weight ⁇ of each feature amount constituting the feature information to be specified is specified (step S2 in FIG. 4).
- the minimum value of the error function is expressed by the following formula (7).
- Superscript (i) indicates the identifier of the learning weld.
- the regularization parameter ⁇ takes a value in the range of 0 to 1.
- the discriminant error is an output value of the discriminant function f (x) when the feature information of the learning welded portion having the weld quality A is input to the kernel function k (x, x ′), and a value corresponding to the weld quality A
- the output value of the discriminant function f (x) when the characteristic information of the learning welded portion having the welding quality B is input to the kernel function k (x, x ′), and the value corresponding to the welding quality B The absolute value of any one of the two differences becomes smaller and becomes larger when the absolute value becomes larger.
- the ⁇ cost of the equation (7) is expressed by the following equation (8).
- the above equation (8) is a convex function approximating the following equation (9).
- y represents a vector whose component is the weight of each feature amount.
- each component of the vector y is set to 1 and the learning having the welding quality B
- Sgn [f (x)] in the above formula (9) is expressed by the following formula (10).
- the regularization term ⁇ T K ⁇ is expressed by the following equation (11).
- the subscript i indicates an identifier representing the type of feature amount constituting the feature information of the learning welded portion having the welding quality A.
- the subscript j indicates an identifier representing the type of feature quantity constituting the feature information of the learning welded portion having the welding quality B.
- the regularization term ⁇ T K ⁇ has a positive correlation with the weight ⁇ of each feature quantity.
- the regularization term ⁇ T K ⁇ is derived as follows.
- Linear sum w 0 of the feature quantity of the learning welds with weld quality A is represented by the following formula (12).
- the weight information w is obtained by adding the ⁇ component orthogonal to the mapping point ⁇ (x (i) ) obtained by mapping the data point indicating the characteristic information of the learning weld to the linear sum w 0. It is expressed by equation (13).
- Equation (5) f (x) in the above-described equation (5) is It is expressed by the following formula (14). That is, it can be seen that ⁇ cost on the left side of Equation (8) does not depend on the value of the ⁇ component. Further, the following equation (15) can be derived from the orthogonality between the linear sum w 0 and the ⁇ component. From equation (15), it is clear that ⁇
- the equation (11) can be derived from the equation (15).
- step S2 the determination unit 2 inputs an arbitrary value (for example, 1) to the weight ⁇ i and the weight ⁇ j of the above equation (11), and for learning having the welding quality A in x (i) of the above equation (11).
- the feature information of the weld is input, the feature information of the learning weld having the welding quality B is input to x (j) , and the regularization term ⁇ T K ⁇ is calculated (step S21 in FIG. 4).
- ⁇ cost of the left side of the above equation (8) is input to y (i) of the above equation (7), and each learning having the welding quality A or the welding quality B to x (i) of the above equation (7).
- enter the feature information of use weld enter a value for the regularization term alpha T K [alpha calculated at step S21 to the regularization term alpha T K [alpha in the formula (7) (step S22 in FIG. 4).
- the regularization parameter of the above equation (7) at this time is an initial value, and the initial value is 1 here.
- Characteristic information x (i) if the output of the determination error for the values y (i) xi] i, the minimum value of the output xi] i is two inequalities (17), the minimum value defined by (18) Become.
- the output ⁇ i at the time of the minimum value is called a slack variable, and by introducing the output ⁇ i at the time of the minimum value into the above equation (7), the above equations (17) and (18) are used as constraints.
- the above equation (7) is converted into the above equation (16).
- the above equation (16) is in the form of a convex quadratic programming problem relating to the output ⁇ and the weight ⁇ of each feature quantity constituting the feature information.
- the solution of the convex quadratic programming problem of the above equation (16) will be shown.
- Equation (16) is solved using Lagrange's undetermined multiplier method.
- the following formula (19) is defined as Lagrangian. Domain: ⁇ R n R n represents the entire real number.
- the quadratic programming problem can be converted into a dual problem with simpler constraints.
- the determining unit 2 uses the weight ⁇ of each feature amount constituting the feature information of the learning weld portion specified as described above as the weight of each feature amount constituting the feature information of the learning weld portion of the discriminant function f (x). Temporarily adopted for ⁇ . Then, in the same manner as in step S1 of FIG. 4, the determination unit 2 tentatively adopts the weight ⁇ of each identified feature quantity as the weight ⁇ of each feature quantity constituting the discrimination function f (x). Is calculated. If the calculated number of misclassifications is equal to or greater than a predetermined value, the determination unit 2 adjusts the regularization parameter ⁇ so as to decrease, and again, as described above, each feature constituting the feature information that minimizes the error function.
- the quantity weight ⁇ is specified (step S2 in FIG. 4).
- the weight ⁇ i and the weight ⁇ j in the above-described equation (11) are set to the respective values specified in the previous step S25.
- the feature amount weight ⁇ is input.
- the weight ⁇ of each specified feature quantity is adopted as the weight ⁇ of each feature quantity constituting the discrimination function f (x).
- the function f (x) is determined (step S3 in FIG. 4).
- the determining unit 2 of the present embodiment sets the initial value of the regularization parameter ⁇ as the maximum value of the regularization parameter ⁇ , and reduces the regularization parameter ⁇ when the number of misclassifications is less than a predetermined value. adjust.
- the regularization parameter ⁇ is large, the regularization term ⁇ T K ⁇ has a large influence on the value of the error function. Therefore, when the regularization parameter ⁇ is large, the weight ⁇ of each feature amount that makes the regularization term ⁇ T K ⁇ sufficiently small is specified as the weight ⁇ of each feature amount that minimizes the value of the error function.
- the number of dimensions of the discriminant function f (x) and the regularization term ⁇ T K ⁇ have a positive correlation.
- the weight ⁇ of each feature amount that makes the regularization term ⁇ T K ⁇ sufficiently small is specified as the weight ⁇ of each feature amount that minimizes the value of the error function, and the weight ⁇ of each feature amount is determined.
- the discriminant function is determined by determining that it is adopted as the weight ⁇ of each feature quantity constituting the function f (x)
- the higher-order discriminant function discriminant boundary
- the weight ⁇ of each feature amount is used as the weight ⁇ of each feature amount constituting the discriminant function f (x).
- the regularization parameter ⁇ is adjusted to be small, and the weight ⁇ of each feature amount that minimizes the value of the error function is specified again.
- the regularization parameter ⁇ is reduced, the influence of the regularization term ⁇ T K ⁇ on the value of the error function is reduced, while the influence of the discrimination error on the value of the error function is increased.
- the discriminant error is the difference between the output value of the discriminant function corresponding to the welding quality A and the value corresponding to the weld quality A, and the output value of the discriminant function corresponding to the weld quality B and the value corresponding to the weld quality B. It is defined by the difference.
- the discriminant error corresponds to the difference between the output value of the discriminant function corresponding to the welding quality A and the value corresponding to the weld quality A, and the output value of the discriminant function corresponding to the weld quality B and the weld quality B.
- the absolute value of any of the differences from the value becomes smaller, it becomes smaller, and when any of the differences becomes larger, it becomes larger. That is, when the discrimination error is reduced, the absolute value of the difference between the output value of the discriminant function corresponding to the welding quality A and the value corresponding to the weld quality A, or the output value of the discriminant function corresponding to the weld quality B and the weld quality
- the absolute value of the difference from the value corresponding to B becomes smaller.
- the number of learning welds having the welding quality B in which the absolute value of the difference from the value corresponding to the welding quality B increases is reduced. Therefore, when the determination error is reduced, the number of erroneous determinations is reduced. Therefore, when the number of misclassifications when the weight ⁇ of each feature amount specified before adjusting the regularization parameter is temporarily adopted as the weight ⁇ of each feature amount constituting the discrimination function f (x) is greater than or equal to a predetermined value. Even so, by adjusting the regularization parameter ⁇ to be small, it is possible to specify the weight ⁇ of each feature quantity in which the number of misclassifications is less than a predetermined value. It is possible to determine the discriminant function f (x) by confirming that it is adopted as the weight ⁇ of each feature quantity constituting the.
- the discriminant function f (x) expressed by the above-described equation (2) has a kernel function k (x, x ′) and a weight ⁇ of each feature quantity constituting the feature information, and has a mapping function. Not done. For this reason, it is not necessary to calculate a mapping function when calculating the number of misclassifications. In other words, it is not necessary to calculate a mapping function to determine the discriminant function. The calculation amount of the mapping function is enormous. Therefore, the welding quality discriminating apparatus 100 that does not need to calculate a mapping function for determining the discriminant function f (x) can determine the discriminant function f (x) with a small amount of calculation.
- the discriminating unit 3 discriminates whether the welding quality of the discrimination target welding portion is the welding quality A or the welding quality B.
- the discriminating unit 3 inputs the feature information composed of the feature amount acquired by the acquiring unit 1 to the kernel function k (x, x ′) of the discriminant function f (x) determined by the determining unit 2, and the discriminant function An output value of f (x), that is, a mapping point obtained by mapping a data point indicating the feature information in a mapping space is calculated.
- determination part 3 discriminate
- determination part 3 will be the absolute value of the difference of the output value of the discrimination function f (x) when the characteristic information of the discrimination
- the determination unit 2 specifies the weight ⁇ of each feature amount constituting the feature information that minimizes the error function.
- the error function is a function composed of the sum of the discrimination error and the regularization term ⁇ T K ⁇ .
- the discriminant error is the difference between the output value of the discriminant function corresponding to the welding quality A and the value corresponding to the weld quality A, and the output value of the discriminant function corresponding to the weld quality B and the value corresponding to the weld quality B.
- the discrimination error varies depending on the output value of the discrimination function f (x).
- the discriminant function f (x) varies according to the weight of the feature amount from the above-described equation (2). For this reason, the variation amount of the discrimination error when the weight of the feature amount that hardly affects the output value of the discrimination function f (x) or does not affect the output value is small.
- the regularization term ⁇ T K ⁇ has a positive correlation with the weight ⁇ of each feature quantity.
- the value of the error function consisting of the sum of the discrimination error and the regularization term is likely to be reduced by minimizing the weight of the feature quantity having a small change amount of the discrimination error (that is, 0). Therefore, the weight of the feature quantity that hardly affects the output value of the discriminant function f (x) or does not affect the output value is highly likely to be specified as 0 by the determination unit 2.
- the feature information input to the kernel function k (x, x ′) of the discriminant function f (x) determined by the deciding unit 2 in order for the discriminating unit 3 to discriminate the welding quality of the discrimination target welding portion is as described above.
- the feature information may be composed of feature amounts other than the feature amount identified with a weight of 0.
- the feature amount identified as having a weight of 0 is the discriminant function. It is not input to the kernel function k (x, x ′) of f (x), and accordingly, the amount of calculation required for determining the welding quality of the determination target weld is reduced.
- the weld quality of the discrimination target weld can be determined at high speed.
- the weight of the feature amount that hardly affects the output value of the discriminant function f (x) or does not affect it at all is specified as 0. Therefore, even if the feature quantity identified as having a weight of 0 is not input to the kernel function k (x, x ′) of the discriminant function f (x), the discrimination target to be performed using the output value of the discriminant function f (x)
- the determination of the welding quality of the welded portion can be performed with a certain accuracy or more.
- the discrimination error is expressed using ⁇ cost of the above-described equation (8) that is a convex function. Since ⁇ cost in Equation (8) described above is a convex function, the weight ⁇ that minimizes the value of the discrimination error can be obtained without falling into a local solution. Therefore, it is possible to efficiently specify the weight ⁇ that makes the discrimination error less than a predetermined value.
- this embodiment demonstrated the example which discriminate
- the discriminating unit 3 uses the discriminant function determined by the determining unit 2 to discriminate whether the welding quality of the discrimination target welded portion is good welding (welding quality A) or poor welding (welding quality B).
- a discrimination boundary for discriminating whether the welding quality of the discrimination target welding portion that has been discriminated as a welding failure is the welding failure B1 or the welding failure B2. Is determined by the same procedure as described above.
- the discriminating unit 3 uses the discriminant function determined by the determining unit 2 to determine whether the weld quality of the discriminating target weld that has been discriminated as a weld failure (weld quality B) is either the weld failure B1 or the weld failure B2. Is determined. Thereby, it is discriminate
- the welding failure is further classified (for example, classified for each cause of welding failure) and discriminated, for example, the spot welding machine according to the discrimination result (according to the factor of welding failure)
- the contact area between the electrode E1 (or E2) and the material to be welded M1 (or M2) and the contact area between the material to be welded M1 and the material to be welded M2 are increased.
- a decrease in energization resistance due to expansion and (2) an increase in energization resistance due to a rise in the temperature of the welded parts of the welded materials M1 and M2 occur at the same time.
- Signal waveform changes. Specifically, when the welding quality is good, such as a characteristic signal waveform change that suddenly decreases energization resistance when dust occurs, or an increase in energization resistance is suppressed when the contact area expands rapidly.
- the weld quality discrimination device 100 is further divided into good welds (for example, dust is generated). It can be expected that the discrimination accuracy of the welding quality is improved by using a configuration in which the discrimination is performed by dividing the case into the case where it does not occur and the case where it does not occur.
- the welding quality can be classified into one of good welding A1, good welding A2, poor welding B1, and poor welding B2.
- the variation in the characteristic information of the learning welded portion having the welding quality of good welding A1 and the variation in the characteristic information of the learning welded portion having the welding quality of good welding A2 are summarized as welding good A1 and good welding A2 as good welding. It can be expected to be suppressed as compared with the variation in characteristic information in the case of And in the determination part 2, first, the welding quality of the discrimination
- a discriminant function indicating a discriminant boundary for discriminating whether the welding is good A1 or other welding qualities is determined by the procedure described above. At this time, as described above, it is expected that the discriminant function can be determined with high accuracy if the variation in the characteristic information of the welded part for learning having the welding quality of good welding A1 is small.
- the discriminating unit 3 uses the discriminant function determined by the determining unit 2 to determine whether the welding quality of the discrimination target welded portion is good welding A1 or other welding quality (welding good A2, welding failure B1, welding failure B2). Determine if it exists.
- the determination unit 2 determines whether the welding quality of the determination target welded portion determined to be other welding quality is good welding A2 or poor welding (welding failure B1 or welding failure B2).
- the discriminant function indicating the discriminant boundary is determined by the same procedure as described above. As described above, it is expected that the discriminant function can be determined with high accuracy if the variation in the characteristic information of the welded part for learning having the welding quality of good welding A2 is small.
- the discriminating unit 3 uses the discriminant function determined by the determining unit 2 to determine that the welding quality of the discriminating target welded part that has been determined to be other welding quality is good welding A1 and poor welding (welding failure B1, welding failure B2). ).
- the determination target welding portion is: One of the four welding qualities A1, A2, B1, and B2 is determined.
- FIG. 9 shows an example in which the welding quality is good and further divided into a plurality of welding qualities (welding good 1 to 5).
- FIG. 10 shows an example in which the welding quality is poor and further divided into a plurality of welding qualities (welding defects 1 to 4).
- the nugget diameter of the welded portion is larger than a predetermined reference value regardless of the presence or absence of dust generation, the presence or absence of electrode fouling, and the presence or absence of other disturbance factors It is assumed that the welding is good and the welding is poor if it is below the reference value. However, for example, by learning that the welded part in which the signal waveform of “welding good 5” shown in FIG. It is possible to arbitrarily classify the welding quality as required.
- the present invention is not limited to this, and as shown in FIG. It is also possible to adopt a configuration in which the welding quality is discriminated using the applied pressure of the electrode or the displacement of the welding electrode.
- FIG. 11 is a schematic configuration diagram showing a modification of the welding quality discrimination device according to the present invention.
- a welding quality determination device 100A according to this modification example, similarly to the above-described welding quality determination device 100, an acquisition unit 1A for acquiring a feature amount, and a determination unit that determines a determination function ( And a determination unit (not shown) for determining the welding quality. Since the function of the determination part and the discrimination
- the acquisition unit 1A spot welds the load cell 11A as detection means for detecting the pressure applied to the electrode E2 when spot welding the welds W of the workpieces M1 and M2, and the welds W of the workpieces M1 and M2. And a displacement meter 12A as means for detecting the displacement of the electrode E1.
- the acquisition unit 1 ⁇ / b> A includes a feature amount extraction unit 13 and a data logger 14.
- the load cell 11A is disposed at a position to receive a load load of the electrode E2, and thereby, a change with time of the applied pressure of the electrode E2 during welding is measured.
- the displacement meter 12A includes a contact or non-contact displacement sensor 121A and a measurement target part 122A that is a displacement measurement target by the displacement sensor 121A.
- One of the displacement sensor 121A and the measurement target part 122A is attached to a movable part (a part that moves integrally with the electrode) of the spot welder, and the other is attached to a stationary part of the spot welder.
- the displacement sensor 121A is attached to the non-moving part
- the measurement target part 122A is attached to the movable part.
- the displacement of the measurement target portion 122A is measured by the displacement sensor 121A (the distance between the displacement sensor 121A and the measurement target portion 122A is measured), whereby the displacement of the electrode E1 during welding is measured.
- the measured applied pressure of the electrode E2 and the displacement of the electrode E1 are input to the feature quantity extraction means 13 directly or via the data logger 14, and the feature quantity extraction means 13 receives the applied pressure of the electrode E2 and the displacement of the electrode E1.
- feature values are extracted.
- this modification demonstrated the structure which discriminate
- dome radius type electrodes having a tip radius of curvature of 40 mm and a tip diameter of 6 mm were used.
- the current / voltage measuring device 11 a weld checker manufactured by Miyachi Technos was used, and the sampling speed was set to 0.1 msec.
- the welded portions of the welded materials M1 and M2 after energizing the electrodes E1 and E2 have an ellipsoidal shape centering on the interface between the welded material M1 and the welded material M2, as shown in FIG. A melted and solidified portion N is formed, and this portion is referred to as a nugget as described above.
- the diameter of the nugget N was measured according to the procedure shown below, and the actual welding quality (welding good / bad) was determined based on the magnitude of the value.
- FIG. 6 is an explanatory diagram for explaining a method of determining actual welding quality (good / bad welding).
- FIG. 6A of the welded workpieces M1 and M2, as shown in FIG. 6B, one welded material M1 is twisted with respect to the other welded material M2.
- the welded portion W is broken.
- FIG.6 (c) the fracture
- the welded portion W is broken, as shown in (i) of FIG. 6 (c), when the fracture occurs at the interface between the welded material M1 and the welded material M2, As shown in ii), breakage may occur in the base material of the workpiece M1 or M2.
- the dimension of the nugget N exposed to the outside was defined as the fracture diameter (nugget diameter) d.
- the dimension of the breakage point of the base material was defined as a fracture diameter (nugget diameter) d.
- any of the above fractured forms was determined as “good” when the nugget diameter d exceeded 2.5 mm, and “bad” when it was 2.5 mm or less.
- FIG. 7 is a diagram illustrating an example of electrode wear and spot change nugget diameter change in spot welding.
- FIG. 7A shows a change in the contact state between the tip of the electrode and the material to be welded
- FIG. 7B shows a change in the nugget diameter of the welded portion.
- the contact state between the tip of the electrode and the material to be welded was observed by placing a pressure sensitive paper between the electrode and the material to be welded every time spot welding with a predetermined number of hit points was completed.
- FIG. 7A the contact area between the tips of the electrodes E1 and E2 and the workpieces M1 and M2 increases as the number of hit points increases.
- the contact state becomes unstable, for example, the contact portion has a donut shape.
- the nugget diameter d obtained under the same welding conditions also changes and eventually the nugget N is not formed.
- the technique of determining the welding quality (good / bad welding) based on the nugget diameter measurement result by the fracture test described above is generally used.
- the welding of the welded part for learning is also used. This was used for quality determination and evaluation of the discrimination result of the welding quality discrimination device 100.
- the welding quality discrimination device 100 (discrimination unit 3) according to the present embodiment is configured to calculate the certainty factor of the discrimination result in addition to the welding quality (good / bad welding) of the discrimination target welding part. ing.
- This certainty factor is represented by a linear distance between a mapping point obtained by mapping the data point indicating the characteristic information of the welding target to be discriminated in the mapping space, and a discrimination boundary that bisects the mapping space. represented by D l (38).
- the subscript l indicates the identifier of the discrimination target weld.
- y indicates a vector of 1 when the discriminant function f (x) ⁇ 0 when the feature information of the discriminating target weld is input, and each component when the discriminant function f (x) ⁇ 0. Indicates a vector of -1.
- the certainty factor D 1 represented by the above equation (38) can be a value from 0 to 1.
- a region where the discriminant function f (x) ⁇ 0 is well welded and a region where the discriminant function f (x) ⁇ 0 is poor weld.
- the certainty (%) of the discrimination result is represented by (0.5 ⁇ D 1 +0.5) ⁇ 100.
- the welding quality of the discriminating target weld is good. It means that the certainty of the determination result (good welding) is 50%.
- the reliability of the discrimination result of the discrimination target welded portion is calculated together with the welding quality of the discriminated discrimination target welded portion, for example, when it is determined that the welding is good but the certainty level is low (If it is below the predetermined threshold value), re-inspection of the welding quality by another device or treat it as being in a poorly welded state, possibly causing a defective welded product to flow out. It is possible to prevent it in advance.
- FIG. 8 shows the evaluation results of this discrimination test.
- the “discrimination result” shown in FIG. 8 indicates the discrimination result by the welding quality discrimination device 100
- the “nugget diameter” indicates the measurement result of the nugget diameter by the fracture test.
- the welded material No. 13
- the threshold 2.5 mm
- a servo pressurization DC power spot welding machine For spot welding of the workpieces M1, M2, and M3, a servo pressurization DC power spot welding machine is used, the pressure applied by the electrodes E1 and E2 is set to 3.2 kN, and the welding current (setting value) is set to 7.8 kA. Then, energization was performed with an energization time of 417 msec per hit point.
- the electrodes E1 and E2 dome radius type electrodes having a tip radius of curvature of 40 mm and a tip diameter of 6 mm were used.
- the current / voltage measuring device 11 As the current / voltage measuring device 11, a weld checker manufactured by Miyachi Technos was used, and the sampling speed was set to about 0.38 msec.
- an ellipsoidal melted and solidified portion (nugget) N is formed in the welded portions of the materials to be welded M1, M2, and M3 after the electrodes E1 and E2 are energized. .
- the diameter (nugget diameter) of each part of the nugget N was measured according to the following procedure, and the actual welding quality (good / bad welding) was determined based on the magnitude of the value. *
- FIG. 15 is an explanatory diagram for explaining a method of determining actual welding quality (good / bad welding).
- the welded materials M1, M2, and M3 that were welded were cut so that a cross section passing through the center of the welded portion W could be observed, and polishing and corrosion were performed.
- An example of a cross-sectional photograph after corrosion is shown in FIG. 15 (b), and a schematic diagram thereof is shown in FIG. 15 (c).
- the diameter of the nugget N at the interface between the material to be welded M1 and the material to be welded M2 is the nugget diameter D1
- Each nugget diameter was defined as D2 and measured.
- the nugget diameter D1 exceeds 5.5 mm (corresponding to four times the square root of the plate thickness of the workpieces M1 and M2), and the nugget diameter D2 is 3.3 mm (the plate of the workpiece M3). When it exceeded 4 times the square root of the thickness, it was determined as “good”, and when either of the nugget diameters D1 and D2 did not satisfy the above condition, it was determined as “bad”.
- FIG. 16 is a diagram showing an example of electrode wear in spot welding and a change in the nugget formation state of the welded portion.
- FIG. 16A shows a change in the contact state between the tip of the electrode and the material to be welded
- FIG. 16B shows a change in the nugget diameters D1 and D2 of the welded portion.
- the contact area between the tips of the electrodes E1 and E2 and the workpieces M1 and M3 increases as the number of hit points increases.
- the nugget diameters D1 and D2 obtained under the same welding conditions change, and the nugget diameter D2 becomes less than 3.3 mm described above.
- the measurement result of the nugget diameter by the cross-sectional observation described above was used for the determination of the welding quality of the learning welded part and the evaluation of the determination result of the welding quality determination apparatus 100.
- the weld quality of the welded parts of 65 sets of materials to be welded was evaluated by the cross-validation method. Specifically, 65 sets of materials to be welded are divided into a first group (22 sets), a second group (22 sets), and a third group (21 sets), and the welds of the materials to be welded in the first group are identified. When setting it as the object welding part, the welding part of the to-be-welded material of the 2nd and 3rd group was used as a welding part for learning.
- the welded portion of the welded material of the second group is used as the discrimination target welded portion
- the welded portion of the welded material of the first and third groups is used as the learning welded portion
- the welded portion of the third group is used.
- the welded portion of the welded material is set as the discrimination target welded portion
- the welded portions of the welded materials of the first and second groups are used as learning welded portions.
- the welding voltage signal waveform The discrimination was performed both in the case of using the fractal dimension as a feature amount (the result is shown in FIG. 18 described later). Further, the actual welding quality of the learning welded portion was evaluated by the cross-sectional observation described above.
- FIG. 17 shows the evaluation results of this discrimination test when the coefficient parameters of the approximate curve of the signal waveform of the welding voltage and the above-mentioned parameters ⁇ , ⁇ , and ⁇ are used as feature amounts.
- the “discrimination result” shown in FIG. 17 indicates the discrimination result by the welding quality discrimination device 100, and the “actual quality” indicates the actual welding quality evaluation result by cross-sectional observation.
- “Probability” shown in FIG. 17 is synonymous with that shown in FIG. As shown in FIG. 17, the actual “good / good” is excluded except for seven workpieces (No. 53, No. 57, No. 60, No. 67, No. 69, No. 71, No. 81). It can be seen that a discrimination result that matches the “bad” judgment is obtained.
- FIG. 18 shows the evaluation result of this discrimination test when the fractal dimension of the signal waveform of the welding voltage is used as the feature quantity.
- the “discrimination result”, “actual quality”, and “probability” shown in FIG. 18 are synonymous with those shown in FIG.
- 14 workpieces No. 53, No. 55, No. 57, No. 60, No. 63, No. 67, No. 70 to No. 72, No. 76, Except for No. 78, No. 80, No. 82, No. 83), it can be seen that a discrimination result that matches the actual “good / bad” judgment is obtained.
- a welding quality discrimination device 100A having the configuration shown in FIG. 11 was used. However, the displacement meter 12A was not used.
- a welded material M1 made of a steel plate having a tensile strength of 590 MPa and a plate thickness of 2.0 mm, and a welded material M2 made of a steel plate having a tensile strength of 270 MPa and a plate thickness of 0.7 mm Were overlapped, sandwiched between the electrodes E1 and E2, and a test for determining the welding quality when spot welding was performed.
- a servo pressurizing DC power spot welding machine For spot welding of the workpieces M1 and M2, a servo pressurizing DC power spot welding machine is used, the pressure applied by the electrodes E1 and E2 is set to 2.5 kN, and the welding current (set value) is set to 7.0 kA. Then, energization was performed with an energization time of 417 msec per hit point.
- the electrodes E1 and E2 dome radius type electrodes having a tip radius of curvature of 40 mm and a tip diameter of 6 mm were used.
- the signal waveform of the applied pressure output as a voltage from the load cell 11 ⁇ / b> A was collected by the data logger 14 and input to the feature amount extraction means 13.
- an ellipsoidal nugget N is formed in the welded portions of the materials to be welded M1 and M2 after the electrodes E1 and E2 are energized.
- the diameter (nugget diameter) DI of the nugget N at the interface between the material to be welded M1 and the material to be welded M2 was measured in the same procedure as described with reference to FIG.
- the nugget diameter DI at the interface exceeds 3.3 mm (corresponding to four times the square root of the thickness of the workpiece M2), the welding quality is determined as “good” and 3.3 mm. It was determined as “bad” in the following cases.
- FIG. 20 is a diagram illustrating an example of a change in the nugget diameter of a welded portion in spot welding.
- the nugget diameter DI obtained under the same welding conditions changes, and the nugget diameter DI becomes less than 3.3 mm described above.
- a nugget N is formed inside one of the materials to be welded as shown in FIG. Even if it is done, the nugget diameter DI may not satisfy 3.3 mm or may not be joined at all at the interface between the material to be welded M1 and the material to be welded M2.
- the quality of the welding quality is evaluated based on the nugget diameter DI of the interface. That is common.
- the measurement result of the interface nugget diameter by the cross-sectional observation described above was used for the determination of the welding quality of the learning welded part and the evaluation of the determination result of the welding quality determination apparatus 100A.
- the weld quality of the welded part of 46 sets of welded materials was evaluated by a cross-validation method. Specifically, 46 sets of materials to be welded are divided into a first group (15 sets), a second group (15 sets), and a third group (16 sets), and the welds of the materials to be welded in the first group are identified. When setting it as the object welding part, the welding part of the to-be-welded material of the 2nd and 3rd group was used as a welding part for learning.
- the welded portion of the welded material of the second group is used as the discrimination target welded portion
- the welded portion of the welded material of the first and third groups is used as the learning welded portion
- the welded portion of the third group is used.
- the welded portion of the welded material is set as the discrimination target welded portion
- the welded portions of the welded materials of the first and second groups are used as learning welded portions.
- the welding voltage signal waveform The discrimination was performed both in the case of using the fractal dimension as a feature amount (the result is shown in FIG. 22 described later). Further, the actual welding quality of the learning welded portion was evaluated by the cross-sectional observation described above.
- FIG. 21 shows the evaluation results of this discrimination test when the coefficient parameters of the approximate curve of the signal waveform of the welding voltage and the above-mentioned parameters ⁇ , ⁇ , and ⁇ are used as feature amounts.
- the “discrimination result”, “actual quality”, and “probability” shown in FIG. 21 are synonymous with those shown in FIG. 17 and FIG. As shown in FIG. 21, it is understood that a discrimination result that matches the actual “good / bad” judgment is obtained except for the three workpieces (No. 101, No. 110, No. 113). .
- FIG. 22 shows the evaluation results of this discrimination test when the fractal dimension of the signal waveform of the welding voltage is used as the feature quantity.
- the “discrimination result”, “actual quality”, and “certainty” shown in FIG. 22 are synonymous with those shown in FIG. As shown in FIG. 22, except for the four workpieces (No. 108, No. 111, No. 112, No. 114), a discrimination result that matches the actual “good / bad” judgment is obtained. I understand that.
Landscapes
- Engineering & Computer Science (AREA)
- Mechanical Engineering (AREA)
- Quality & Reliability (AREA)
- Physics & Mathematics (AREA)
- Plasma & Fusion (AREA)
- Investigating And Analyzing Materials By Characteristic Methods (AREA)
- General Factory Administration (AREA)
Abstract
Description
具体的には、特許文献4に記載の装置は、良好な溶接品質が得られる際の溶接電流・溶接電圧信号をサンプル信号として記憶する。そして、特許文献4に記載の装置は、その記憶されたサンプル信号に指定の係数を乗じて生成される人工信号を算出する。特許文献4に記載の装置は、その人工信号とサンプル信号とを3次元的にプロットし、そのプロットした領域を小領域に分割し、各小領域に存在するプロット点をカウントする。特許文献4に記載の装置は、そのカウント数に、各小領域毎に設定された重みをかけて、基準信号とする。特許文献4に記載の装置は、基準信号の平均と分散を算出し、これらに基づき、オンライン検査中に入力された溶接電流・溶接電圧信号の確率密度関数を計算し、確率密度が低い(10-4以下)場合に溶接不良が発生していると判定する。
また、本発明における「溶接品質に対応する値」とは、一方の溶接品質と他方の溶接品質とを判別し得るように予め決定された値であって、一方の溶接品質と他方の溶接品質とで異なる値とされている。
また、本発明における「学習用溶接部の前記特徴情報」とは、学習用溶接部を溶接した際の、溶接電流、溶接電圧、溶接用電極の加圧力及び溶接用電極の変位のうちの少なくとも一つの物理量に基づいて得られた複数の特徴量を成分とする特徴情報を意味する。
さらに、本発明における「物理量に基づいて得られた複数の特徴量」には、溶接電流、溶接電圧、溶接用電極の加圧力及び溶接用電極の変位のうちの少なくとも一つの物理量そのものから得られた特徴量に限られず、上記の複数の物理量を用いた演算結果から得られた特徴量も含まれる。例えば、本発明における「物理量に基づいて得られた複数の特徴量」には、溶接電圧を溶接電流で除算した結果である通電抵抗から得られた特徴量も含まれる。
フラクタル次元とは、時系列信号波形を幾何構造とみなした場合に、その幾何構造の形状的な複雑さの程度を示す。フラクタル次元が大きいほど、時系列信号波形が複雑であることを意味する。溶接電流及び溶接電圧の信号波形に対してフラクタル次元解析を適用する際には、1回の溶接打点に対応する信号波形全てを1つの連続図形とみなして適用しても良いし、ある特定の時間間隔に信号波形を分割して適用しても良い。信号波形を分割してフラクタル次元解析を適用する場合には、分割数に応じた個数のフラクタル次元が得られる。
上記式(A)において、Sは信号波形のデータ列を、δはボックスサイズを、Nδ(S)は信号波形を被覆するために必要なボックス個数を意味する。
フラクタル次元dを算出する際には、以下の第1ステップ~第5ステップを実行する。
(1)第1ステップ:0より大きい適当なボックスサイズδを設定する。
(2)第2ステップ:図2(a)に示すように、信号波形Sを被覆するために必要なボックス個数Nδ(S)をカウントする(図2(a)では、ボックスサイズδ4で信号波形を被覆した例を示している)。
(3)第3ステップ:図2(a)に示すように、ボックスサイズδを変更(図2(a)に示す例では、δ1、δ2、δ3、δ4と段階的に変更)して、第1ステップ及び第2ステップを繰り返すことにより、各ボックスサイズδに対応するボックス個数Nδ(S)を算出する。
(4)第4ステップ:図2(b)に示すように、横軸にlnδ、縦軸にlnNδ(S)をプロットしたグラフを描く。
(5)第5ステップ:図2(b)に示すように、第4ステップで描いたグラフに最小自乗法等により近似直線Lを当てはめ、その傾きをフラクタル次元dとして算出する。
そこで、信号波形の変化のうち、巨視的な(基本的な)変化を近似曲線(高次曲線やスプライン曲線等)の当てはめによって表現し、微小な変化を信号波形に対する上記近似曲線の誤差によって表現すること、すなわち、信号波形の変化を2つの要素に分解して捉えることで、信号波形の情報をより詳細に表現できると考えられる。
(1)図12(a)に示すように、信号波形(図12(a)に示す例では溶接電圧)の実データに近似曲線を当てはめる。図12(a)に示す例では、7次曲線を当てはめている。
(2)図12(b)に示すように、信号波形(信号波形の実データ)に対する近似曲線の誤差を母集団とし、この母集団の累積密度分布に、以下の式(B)で表される自由度νのスチューデントのt分布を当てはめる。この当てはめの際、以下の式(C)で表される密度関数(尤度関数)に最尤推定法を適用して、未知のパラメータである自由度νを決定する。式(C)で表される尤度関数は、自由度νにある値を指定したときに、式(B)で表されるスチューデントのt分布が実データの母集団の累積密度分布にどれだけ当てはまりが良いかを測るための関数である。具体的には、上記の最尤推定法では、式(C)で表される尤度関数の対数をとった対数尤度関数を最大化するように自由度νを決定する。このようにして決定された自由度νは、式(B)で表されるスチューデントのt分布が実データの母集団の累積密度分布に最も当てはまりの良い自由度νの推定値となっていると考えられる。
ここで、上記式(B)、(C)において、zは特徴情報(溶接電流や溶接電圧)の実データを、Γ(・)はガンマ分布を意味する。また、上記式(C)において、μは位置パラメータを、σはスケールパラメータ(σ>0)を意味する。これらのパラメータμ、σは、母集団の累積密度分布から推定可能な値である。
(3)上記(1)で得られた近似曲線の係数パラメータ(例えば、n次の近似曲線P0t+P1t2+・・・Pn-1tnの各係数P0~Pn-1 ここで、tは時間を表す。図12(a)に示す例では、n=7)と、上記(2)で得られたパラメータμ、σ、νとを特徴量とする。
wは、特徴情報を構成する各特徴量の重みを成分とする重み情報(ベクトル)を示す。式(1)に記載のxは、溶接品質A又は溶接品質Bを有する学習用溶接部の特徴情報(ベクトル)を示す。φ(・)は、前記の写像空間に特徴情報を示すデータ点(ベクトルの先端の点)を写像する写像関数であって、正定値性を有するものを示す。正定値性のある写像関数としては、ガウス分布の関数などがある。
αは、特徴情報を構成する各特徴量の重みを示す。k(x,x’)は、要素がk(x,x’)で与えられる行列Kが半正定値であるカーネル関数を示す。式(2)以降に記載のxは、溶接品質Aを有する学習用溶接部の特徴情報(ベクトル)を示す。x’は、溶接品質Bを有する学習用溶接部の特徴情報(ベクトル)を示す。要素がk(x,x’)で与えられる行列Kとは、カーネル関数k(x,x’)に溶接品質Aを有する学習用溶接部の特徴情報xを入力したときに得られるカーネル関数の出力値、及び、カーネル関数k(x,x’)に溶接品質Bを有する学習用溶接部の特徴情報x’を入力したときに得られるカーネル関数の出力値を要素とする行列である。
また、要素がカーネル関数k(x,x’)で与えられる行列Kが半正定値であるカーネル関数k(x,x’)の他の例として、下記式のシグモイド関数とガウス関数とがある。
尚、上記に例示した7つのカーネル関数k(x,x’)において、f(・)は任意の関数を示し、q(・)は非負係数の多項式を示し、k1(・,・)、ka(・,・)及びkb(・,・)は任意のカーネル関数を示し、下付き文字a及びbは、学習用溶接部の識別子を示し、βはシグモイド関数のゲインを示し、σは分散を示す。
と定義すれば、下記式(4)が導出される。
dは、特徴情報を構成する特徴量の数を示す。特徴情報を構成する特徴量の数を十分に多くすれば、上記式(1)から下記式(5)が導出される。
ここで、重み情報wは、下記式(6)で表現される。
カーネル関数k(x,x’)の定義(上記式(3))によれば、上記式(6)を用いて、上記式(1)から上記式(2)が導出される。上記式(2)の判別関数f(x)は、学習用溶接部の特徴情報を構成する特徴量の数に、その次元数が影響される関数となっている。
上付き文字の(i)は、学習用溶接部の識別子を示す。尚、正則化パラメータλは0~1の範囲の値を採る。判別誤差は、溶接品質Aを有する学習用溶接部の特徴情報をカーネル関数k(x,x’)に入力したときの判別関数f(x)の出力値と、溶接品質Aに対応する値との差、及び、溶接品質Bを有する学習用溶接部の特徴情報をカーネル関数k(x,x’)に入力したときの判別関数f(x)の出力値と、溶接品質Bに対応する値との差で規定され、前記2つの差の何れかの絶対値が小さくなれば小さくなり、大きくなれば大きくなるものである。
上記式(8)は下記式(9)を近似した凸関数である。
上記式(9)のyは、各特徴量の重みを成分とするベクトルを示す。判別誤差を求めるために、溶接品質Aを有する学習用溶接部の特徴情報をカーネル関数k(x,x’)に入力する場合は、ベクトルyの各成分を1とし、溶接品質Bを有する学習用溶接部の特徴情報をカーネル関数k(x,x’)に入力する場合は、ベクトルyの各成分を-1とする。上記式(9)のsgn[f(x)]は、下記式(10)で表現される。
下付き文字のiは、溶接品質Aを有する学習用溶接部の特徴情報を構成する特徴量の種類を表わす識別子を示す。下付き文字のjは、溶接品質Bを有する学習用溶接部の特徴情報を構成する特徴量の種類を表わす識別子を示す。
また、重み情報wは、線形和w0に、学習用溶接部の特徴情報を示すデータ点を写像した写像点φ(x(i))に直交するξ成分を加えたものであるので、下記式(13)で表現される。
ここで、重み情報wと写像点φ(x(j))との内積φ(x(j))T・ξは0であるという条件から、前述した式(5)のf(x)は、下記式(14)で表現される。
即ち、式(8)の左辺のγcostは、ξ成分の値に依存しないことが分かる。また、線形和w0とξ成分との直交性から、下記式(15)を導出できる。
式(15)から、λ||w||2はξ=0のときに最小値になることは明らかである。ゆえに、誤差関数が最小となるのは、w=w0のときである。ここで、上記式(12)を利用すると、式(15)から式(11)を導出できる。
最小値となるときの出力ξiをスラック変数と呼び、最小値となるときの出力ξiを上記式(7)に導入することによって、上記式(17)、(18)を制約条件として、上記式(7)は、上記式(16)に変換される。
凸二次計画問題
定義域:Ω⊆Rn
式(19)~式(24)においては、α及びβはラグランジュ乗数を示す。w*は最適化されたときの重み情報を示す。α*及びβ*は、w*が得られたときのラグランジュ乗数α及びβを示す。
上記式(26)から下記式(29)においては、Qはn×n正定値行列を示し、kはn-ベクトルを示し、cはm-ベクトルを示し、wは最適化対象のベクトルを示し、Xはm×n行列を示す。
ここで、Kは対称行列だからKT=Kなので、上記式(30)から下記式(31)を導出できる。
1-βi-γi≠0…(34)
1-βi-γi=0…(35)
このように、ラグランジュ関数が1次式となっている変数はその係数は0なので、双対問題は出力ξiと無関係である。ゆえに、重みαiを上記式(33)で置換し、上記式(35)の制約条件のもとで、下記式(36)のラグランジュ関数を最大化する。
また、βi≧0、及び、γi≧0の条件から、上記式(35)の制約条件は、下記式(37)のようになる。
0≦γi≦1…(37)
最急降下法や内点法といった公知の最適解探索手法を用いて、上記式(36)からγiを算出し(図4のステップS24)、該算出したγiを式(33)に代入すれば、誤差関数を最小とする特徴情報を構成する各特徴量の重みαが特定される(図4のステップS25)。
図1に示すように、引っ張り強さが270MPaで、板厚が0.7mmの鋼板からなる被溶接材M1、M2を重ね合わせて、電極E1、E2で挟み込み、スポット溶接した際の溶接品質を判別する試験を行った。被溶接材M1、M2のスポット溶接にはエア加圧式の定置型スポット溶接機を用い、電極E1、E2による加圧力を150kgf、溶接電流(設定値)を8.5kAに設定して、1打点当たり5周期の通電(1周期が1/60secなので、通電時間にして約80msec)を行った。電極E1、E2としては、先端の曲率半径が40mmで、先端の径が6mmであるドームラジアス型電極を用いた。電流・電圧測定器11としては、ミヤチテクノス社製のウェルドチェッカーを用い、サンプリング速度は0.1msecとした。
以上に説明した破断試験によるナゲット径の測定結果に基づき溶接品質(溶接の良・不良)を判定するという手法は、一般的に利用されており、本判別試験においても、学習用溶接部の溶接品質の判定や、溶接品質判別装置100の判別結果の評価に利用した。
上記式(38)において、下付き文字lは、判別対象溶接部の識別子を示す。また、yは、判別対象溶接部の特徴情報を入力したときの判別関数f(x)≧0のときには各成分が1のベクトルを示し、上記の判別関数f(x)<0のときには各成分が-1のベクトルを示す。
ここで、例えば、判別関数f(x)≧0となる領域が溶接良好であり、判別関数f(x)<0となる領域が溶接不良であるとする。このとき、判別対象溶接部の特徴情報を入力したときの判別関数f(x)≧0で且つその確信度Dl=1であるとすれば、この判別対象溶接部の溶接品質は溶接良好であると判別され、その判別結果(溶接良好)の確らしさは100%(溶接不良である確からしさは0%)であることを意味する。判別結果の確からしさ(%)は、(0.5×Dl+0.5)×100で表される。また、判別対象溶接部の特徴情報を入力したときの判別関数f(x)≧0で且つその確信度Dl=0であるとすれば、この判別対象溶接部の溶接品質は溶接良好であると判別され、その判別結果(溶接良好)の確らしさは50%であることを意味する。同様に、判別対象溶接部の特徴情報を入力したときの判別関数f(x)<0で且つその確信度Dl=1であるとすれば、この判別対象溶接部の溶接品質は溶接不良であると判別され、その判別結果(溶接不良)の確らしさは100%であることを意味する。また、判別対象溶接部の特徴情報を入力したときの判別関数f(x)<0で且つその確信度Dl=0であるとすれば、この判別対象溶接部の溶接品質は溶接不良であると判別され、その判別結果(溶接不良)の確らしさは50%であることを意味する。
図8に示すように、本実施形態に係る溶接品質判別装置100によれば、溶接部ナゲット径による「良・不良」判定のしきい値(2.5mm)付近の被溶接材(No.13~No.16)を除き、ナゲット径による「良・不良」判定と合致する判別結果が得られていることが分かる。そして、溶接部ナゲット径による「良・不良」判定のしきい値付近の上記被溶接材では、判別結果の確からしさが「良・不良」ともに50%程度となり、良・不良の何れの領域に属するか微妙な判別結果となっていることが分かる。
本判別試験でも図1に示す構成を有する溶接品質判別装置100を用いた。ただし、本判別試験では、図13に示すように、引張り強さが590MPaで板厚が2.0mmの鋼板からなる被溶接材M1、M2と、引張り強さが270MPaで板厚が0.7mmの鋼板からなる被溶接材M3とを重ね合わせた、計3枚の被溶接材を電極E1、E2で挟み込み、スポット溶接した際の溶接品質を判別した。被溶接材M1、M2、M3のスポット溶接にはサーボ加圧式で直流電源のスポット溶接機を用い、電極E1、E2による加圧力を3.2kN、溶接電流(設定値)を7.8kAに設定して、1打点当たり通電時間417msecの通電を行った。電極E1、E2としては、先端の曲率半径が40mmで、先端の径が6mmであるドームラジアス型電極を用いた。電流・電圧測定器11としては、ミヤチテクノス社製のウェルドチェッカーを用い、サンプリング速度は約0.38msecとした。
以上に説明した断面観察によるナゲット径の測定結果を、学習溶接部の溶接品質の判定や、溶接品質判別装置100の判別結果の評価に利用した。
図17に示すように、7個の被溶接材(No.53、No.57、No.60、No.67、No.69、No.71、No.81)を除き、実際の「良・不良」判定と合致する判別結果が得られていることが分かる。
図18に示すように、14個の被溶接材(No.53、No.55、No.57、No.60、No.63、No.67、No.70~No.72、No.76、No.78、No.80、No.82、No.83)を除き、実際の「良・不良」判定と合致する判別結果が得られていることが分かる。
本判別試験では図11に示す構成を有する溶接品質判別装置100Aを用いた。ただし、変位計12Aは用いなかった。図11に示すように、引張り強さが590MPaで板厚が2.0mmの鋼板からなる被溶接材M1と、引張り強さが270MPaで板厚が0.7mmの鋼板からなる被溶接材M2とを重ね合わせて、電極E1、E2で挟み込み、スポット溶接した際の溶接品質を判定する試験を行った。被溶接材M1、M2のスポット溶接には、サーボ加圧式で直流電源のスポット溶接機を用い、電極E1、E2による加圧力を2.5kN、溶接電流(設定値)を7.0kAに設定して、1打点当たり通電時間417msecの通電を行った。電極E1、E2としては、先端の曲率半径が40mmで、先端の径が6mmであるドームラジアス型電極を用いた。そしてロードセル11Aから電圧として出力される加圧力の信号波形をデータロガー14で収集し、特徴量抽出手段13に入力した。
以上に説明した断面観察による界面のナゲット径の測定結果を、学習溶接部の溶接品質の判定や、溶接品質判別装置100Aの判別結果の評価に利用した。
図21に示すように、3個の被溶接材(No.101、No.110、No.113)を除き、実際の「良・不良」判定と合致する判別結果が得られていることが分かる。
図22に示すように、4個の被溶接材(No.108、No.111、No.112、No.114)を除き、実際の「良・不良」判定と合致する判別結果が得られていることが分かる。
2・・・決定部
3・・・判別部
11・・・電流・電圧測定器
12・・・コイル
13・・・特徴量抽出手段
100・・・溶接品質判別装置
E1、E2・・・電極
M1、M2・・・被溶接材
S1、S2・・・シャンク
W・・・溶接部
Claims (2)
- 溶接品質が未知の判別対象溶接部を溶接した際の、溶接電流、溶接電圧、溶接用電極の加圧力及び溶接用電極の変位のうちの少なくとも一つの物理量に基づいて得られた複数の特徴量を成分とする特徴情報を示すデータ点を、前記特徴情報を構成する特徴量の数よりも高い次元数の写像空間に写像した写像点が、前記写像空間を二分することによって形成された2つの溶接品質の領域の何れに位置するかを判別し、前記判別対象溶接部の溶接品質を前記写像点が位置すると判別した領域に対応する溶接品質であると判別する溶接品質判別装置において、
前記特徴量を取得するための取得部と、前記写像空間を二分する判別境界を示す判別関数を決定する決定部と、該決定部によって決定された判別関数に前記判別対象溶接部の特徴情報を入力したときの該判別関数の出力値に基づいて、前記判別対象溶接部の溶接品質を判別する判別部とを備え、
前記取得部は、前記判別対象溶接部を溶接する際の、溶接電流、溶接電圧、溶接用電極の加圧力及び溶接用電極の変位のうちの少なくとも一つの物理量を検出する検出手段と、該検出手段によって検出した前記物理量に基づき前記特徴量を抽出する特徴量抽出手段とを具備し、
前記2つの溶接品質のそれぞれは、予め設定された互いに異なる溶接品質であり、
前記決定部は、前記2つの溶接品質の何れかを有することが既知である学習用溶接部の前記特徴情報を用いて前記判別関数を決定し、
前記判別関数は、前記2つの溶接品質のうち一方又は他方の溶接品質を有する前記学習用溶接部の特徴情報が入力されると、前記特徴情報が入力された学習用溶接部の写像点を出力するカーネル関数k(x,x’)と、前記カーネル関数k(x,x’)に付され、前記特徴情報を構成する各特徴量の重みとから構成された関数であり、
前記カーネル関数k(x,x’)は、要素がk(x,x’)で与えられる行列Kが半正定値であるカーネル関数であり、xは、前記一方の溶接品質を有する学習用溶接部の特徴情報であり、x’は、前記他方の溶接品質を有する学習用溶接部の特徴情報であり、
前記決定部は、
前記一方の溶接品質を有する学習用溶接部の特徴情報を前記カーネル関数k(x,x’)に入力したときの前記判別関数の出力値と前記一方の溶接品質に対応する値との差、及び、前記他方の溶接品質を有する学習用溶接部の特徴情報を前記カーネル関数k(x,x’)に入力したときの前記判別関数の出力値と前記他方の溶接品質に対応する値との差で規定され、前記2つの差の何れかの絶対値が小さくなれば小さくなり、大きくなれば大きくなる判別誤差と、前記判別関数の次元数に対し正の相関を有し、前記特徴情報を構成する各特徴量の重みに応じて変動すると共に、正則化パラメータが乗じられた正則化項との和からなる誤差関数の値を最小にするように、所定の正則化パラメータについて、前記特徴情報を構成する各特徴量の重みを特定し、
前記誤差関数の値を最小にするように特定した前記特徴情報を構成する各特徴量の重みを、前記判別関数を構成する各特徴量の重みとして仮に採用したときにおいて、前記一方の溶接品質を有する学習用溶接部の特徴情報を前記カーネル関数k(x,x’)に入力したときの前記判別関数の出力値と前記一方の溶接品質に対応する値との差の絶対値よりも、前記一方の溶接品質を有する学習用溶接部の特徴情報を前記カーネル関数k(x,x’)に入力したときの前記判別関数の出力値と前記他方の溶接品質に対応する値との差の絶対値の方が小さくなる前記一方の溶接品質を有する学習用溶接部の数と、前記他方の溶接品質を有する学習用溶接部の特徴情報を前記カーネル関数k(x,x’)に入力したときの前記判別関数の出力値と前記他方の溶接品質に対応する値との差の絶対値よりも、前記他方の溶接品質を有する学習用溶接部の特徴情報を前記カーネル関数k(x,x’)に入力したときの前記判別関数の出力値と前記一方の溶接品質に対応する値との差の絶対値の方が小さくなる前記他方の溶接品質を有する学習用溶接部の数とを加算した誤判別個数が、所定値以上の場合は、前記正則化パラメータを調整して、再度、前記誤差関数の値を最小にするように、前記特徴情報を構成する各特徴量の重みを特定し、
前記誤判別個数が所定値未満の場合は、前記誤差関数の値を最小にするように特定した前記特徴情報を構成する各特徴量の重みを、前記判別関数を構成する各特徴量の重みとして採用することを確定して、前記判別関数を決定することを特徴とする溶接品質判別装置。 - 前記判別部は、判別した前記判別対象溶接部の溶接品質と共に、当該判別対象溶接部の特徴情報を示すデータ点を前記写像空間に写像した写像点と、前記写像空間を二分する前記判別境界との距離で表される当該判別対象溶接部の判別結果の確信度を算出することを特徴とする請求項1に記載の溶接品質判別装置。
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2012538686A JPWO2012050108A1 (ja) | 2010-10-14 | 2011-10-12 | 溶接品質判別装置 |
US13/878,917 US20130248505A1 (en) | 2010-10-14 | 2011-10-12 | Welding quality classification apparatus |
EP11832543.0A EP2628561A1 (en) | 2010-10-14 | 2011-10-12 | Welding quality determination device |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2010231343 | 2010-10-14 | ||
JP2010-231343 | 2010-10-14 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2012050108A1 true WO2012050108A1 (ja) | 2012-04-19 |
Family
ID=45938337
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2011/073379 WO2012050108A1 (ja) | 2010-10-14 | 2011-10-12 | 溶接品質判別装置 |
Country Status (4)
Country | Link |
---|---|
US (1) | US20130248505A1 (ja) |
EP (1) | EP2628561A1 (ja) |
JP (1) | JPWO2012050108A1 (ja) |
WO (1) | WO2012050108A1 (ja) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111250890A (zh) * | 2020-02-17 | 2020-06-09 | 南京未来网络产业创新有限公司 | 一种对接接头焊缝质量在线监测方法及装置 |
WO2021009963A1 (ja) * | 2019-07-17 | 2021-01-21 | 株式会社日立製作所 | 溶接作業データ蓄積装置、溶接作業支援システムおよび溶接ロボット制御装置 |
CN113505657A (zh) * | 2021-06-18 | 2021-10-15 | 东风汽车集团股份有限公司 | 一种焊点质量检测方法及装置 |
CN115392132A (zh) * | 2021-09-18 | 2022-11-25 | 天津商科数控技术股份有限公司 | 基于深度学习的焊点质量异常检测方法、装置、*** |
Families Citing this family (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE102012025200A1 (de) * | 2012-12-27 | 2014-07-03 | Robert Bosch Gmbh | Schweißverfahren zum Verschweißen von Aluminium |
WO2014136507A1 (ja) * | 2013-03-08 | 2014-09-12 | Jfeスチール株式会社 | 抵抗スポット溶接方法 |
US9314878B2 (en) * | 2013-09-12 | 2016-04-19 | Ford Global Technologies, Llc | Non-destructive aluminum weld quality estimator |
DE102013221273A1 (de) * | 2013-10-21 | 2015-04-23 | Robert Bosch Gmbh | Verfahren zum Überwachen und Regeln einer Qualität von Schweißpunkten |
DE102014008623A1 (de) * | 2014-06-17 | 2015-12-17 | Thyssenkrupp Ag | Verfahren zum Widerstandspunktschweißen eines Sandwichmaterials und Vorrichtung hierfür |
US11137748B2 (en) * | 2014-09-12 | 2021-10-05 | Fronius International Gmbh | Welding control system |
MX2017007020A (es) * | 2014-12-01 | 2017-08-14 | Jfe Steel Corp | Metodo de soldadura por puntos de resistencia. |
GB201509152D0 (en) * | 2015-05-28 | 2015-07-15 | Rolls Royce Plc | Welding method |
DE102015114957A1 (de) * | 2015-09-07 | 2017-03-09 | Harms + Wende Gmbh & Co. Kg | Elektrisches Schweißverfahren |
KR101798110B1 (ko) * | 2016-04-12 | 2017-11-15 | 한전원자력연료 주식회사 | 핵연료봉 저항용접 품질 모니터링 방법 |
KR102215856B1 (ko) * | 2016-06-09 | 2021-02-15 | 제이에프이 스틸 가부시키가이샤 | 저항 스폿 용접 방법 |
US10875125B2 (en) * | 2017-06-20 | 2020-12-29 | Lincoln Global, Inc. | Machine learning for weldment classification and correlation |
WO2019160141A1 (ja) * | 2018-02-19 | 2019-08-22 | Jfeスチール株式会社 | 抵抗スポット溶接方法および溶接部材の製造方法 |
DE102018105893B4 (de) * | 2018-03-14 | 2022-02-10 | GLAMAtronic Schweiß- und Anlagentechnik GmbH | Verfahren und Vorrichtung zum Qualitätsauswerten beim Widerstandsschweißen sowie Computerprogrammprodukt, Steuerungseinrichtung und Verwendung |
FR3095976B1 (fr) * | 2019-05-17 | 2022-01-21 | Psa Automobiles Sa | Dispositif d’analyse automatisée de points de soudure réalisés par un appareil de soudage par résistance en courant alternatif |
US20210341451A1 (en) * | 2020-03-30 | 2021-11-04 | Hitachi Rail S.P.A. | Method and system for monitoring and identifying the weld quality on metallic components |
US11167378B1 (en) * | 2020-05-01 | 2021-11-09 | David W. Steinmeier | Techniques for determining weld quality |
CN112091472B (zh) * | 2020-08-19 | 2021-11-02 | 中车工业研究院有限公司 | 焊接过程质量融合判断方法及装置 |
CN115266141B (zh) * | 2022-07-29 | 2024-05-14 | 广汽本田汽车有限公司 | 基于gru-c网络的点焊质量检测方法、装置及存储介质 |
Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH01271078A (ja) | 1988-04-20 | 1989-10-30 | Yashima Denki Kk | 抵抗溶接機のモニタリング装置 |
JPH06170552A (ja) | 1992-12-01 | 1994-06-21 | Matsushita Electric Ind Co Ltd | 抵抗溶接の溶接品質監視装置 |
JPH07290254A (ja) | 1994-04-22 | 1995-11-07 | Sekisui Chem Co Ltd | 抵抗溶接における溶接品質推定方法 |
JPH10314956A (ja) | 1997-05-14 | 1998-12-02 | Matsushita Electric Ind Co Ltd | 抵抗溶接部の品質評価方法および装置 |
JP2003516863A (ja) | 1999-12-15 | 2003-05-20 | ザ・ユニバーシティ・オブ・シドニー | 溶接評価 |
JP2006110554A (ja) | 2004-10-12 | 2006-04-27 | Dengensha Mfg Co Ltd | 抵抗スポット溶接品質判定方法と監視装置 |
WO2008056638A1 (en) * | 2006-11-06 | 2008-05-15 | Fujifilm Ri Pharma Co., Ltd. | Brain image diagnosis supporting method, program, and recording method |
JP2008287437A (ja) * | 2007-05-16 | 2008-11-27 | Canon Inc | 情報処理方法、情報処理装置 |
JP2009186243A (ja) * | 2008-02-04 | 2009-08-20 | Nippon Steel Corp | 判別装置、判別方法及びプログラム |
JP2010214380A (ja) * | 2009-03-13 | 2010-09-30 | Osaka Univ | リアルタイム溶接品質判定装置及び判定方法 |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7805388B2 (en) * | 1998-05-01 | 2010-09-28 | Health Discovery Corporation | Method for feature selection in a support vector machine using feature ranking |
US7958063B2 (en) * | 2004-11-11 | 2011-06-07 | Trustees Of Columbia University In The City Of New York | Methods and systems for identifying and localizing objects based on features of the objects that are mapped to a vector |
-
2011
- 2011-10-12 WO PCT/JP2011/073379 patent/WO2012050108A1/ja active Application Filing
- 2011-10-12 JP JP2012538686A patent/JPWO2012050108A1/ja active Pending
- 2011-10-12 US US13/878,917 patent/US20130248505A1/en not_active Abandoned
- 2011-10-12 EP EP11832543.0A patent/EP2628561A1/en not_active Withdrawn
Patent Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH01271078A (ja) | 1988-04-20 | 1989-10-30 | Yashima Denki Kk | 抵抗溶接機のモニタリング装置 |
JPH06170552A (ja) | 1992-12-01 | 1994-06-21 | Matsushita Electric Ind Co Ltd | 抵抗溶接の溶接品質監視装置 |
JPH07290254A (ja) | 1994-04-22 | 1995-11-07 | Sekisui Chem Co Ltd | 抵抗溶接における溶接品質推定方法 |
JPH10314956A (ja) | 1997-05-14 | 1998-12-02 | Matsushita Electric Ind Co Ltd | 抵抗溶接部の品質評価方法および装置 |
JP2003516863A (ja) | 1999-12-15 | 2003-05-20 | ザ・ユニバーシティ・オブ・シドニー | 溶接評価 |
JP2006110554A (ja) | 2004-10-12 | 2006-04-27 | Dengensha Mfg Co Ltd | 抵抗スポット溶接品質判定方法と監視装置 |
WO2008056638A1 (en) * | 2006-11-06 | 2008-05-15 | Fujifilm Ri Pharma Co., Ltd. | Brain image diagnosis supporting method, program, and recording method |
JP2008287437A (ja) * | 2007-05-16 | 2008-11-27 | Canon Inc | 情報処理方法、情報処理装置 |
JP2009186243A (ja) * | 2008-02-04 | 2009-08-20 | Nippon Steel Corp | 判別装置、判別方法及びプログラム |
JP2010214380A (ja) * | 2009-03-13 | 2010-09-30 | Osaka Univ | リアルタイム溶接品質判定装置及び判定方法 |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2021009963A1 (ja) * | 2019-07-17 | 2021-01-21 | 株式会社日立製作所 | 溶接作業データ蓄積装置、溶接作業支援システムおよび溶接ロボット制御装置 |
JP2021016870A (ja) * | 2019-07-17 | 2021-02-15 | 株式会社日立製作所 | 溶接作業データ蓄積装置、溶接作業支援システムおよび溶接ロボット制御装置 |
JP7261682B2 (ja) | 2019-07-17 | 2023-04-20 | 株式会社日立製作所 | 溶接作業データ蓄積装置、溶接作業支援システムおよび溶接ロボット制御装置 |
CN111250890A (zh) * | 2020-02-17 | 2020-06-09 | 南京未来网络产业创新有限公司 | 一种对接接头焊缝质量在线监测方法及装置 |
CN111250890B (zh) * | 2020-02-17 | 2021-11-23 | 南京未来网络产业创新有限公司 | 一种对接接头焊缝质量在线监测方法及装置 |
CN113505657A (zh) * | 2021-06-18 | 2021-10-15 | 东风汽车集团股份有限公司 | 一种焊点质量检测方法及装置 |
CN113505657B (zh) * | 2021-06-18 | 2022-05-03 | 东风汽车集团股份有限公司 | 一种焊点质量检测方法及装置 |
CN115392132A (zh) * | 2021-09-18 | 2022-11-25 | 天津商科数控技术股份有限公司 | 基于深度学习的焊点质量异常检测方法、装置、*** |
CN115392132B (zh) * | 2021-09-18 | 2023-07-11 | 天津商科数控技术股份有限公司 | 基于深度学习的焊点质量异常检测方法、装置、*** |
Also Published As
Publication number | Publication date |
---|---|
US20130248505A1 (en) | 2013-09-26 |
JPWO2012050108A1 (ja) | 2014-02-24 |
EP2628561A1 (en) | 2013-08-21 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2012050108A1 (ja) | 溶接品質判別装置 | |
Xing et al. | Quality assessment of resistance spot welding process based on dynamic resistance signal and random forest based | |
Okaro et al. | Automatic fault detection for laser powder-bed fusion using semi-supervised machine learning | |
Karandikar et al. | Tool life prediction using Bayesian updating. Part 1: Milling tool life model using a discrete grid method | |
Thekkuden et al. | Investigation of feed-forward back propagation ANN using voltage signals for the early prediction of the welding defect | |
Boersch et al. | Data mining in resistance spot welding: A non-destructive method to predict the welding spot diameter by monitoring process parameters | |
CN101911272A (zh) | 用于超声波键合的方法和设备 | |
Wan et al. | Quality monitoring based on dynamic resistance and principal component analysis in small scale resistance spot welding process | |
EP2342040B1 (en) | System and process for automatic determination of welding parameters for automated friction stir welding | |
Lee et al. | Development of real-time diagnosis framework for angular misalignment of robot spot-welding system based on machine learning | |
KR102163828B1 (ko) | 용접 파형의 스패터 예측을 위한 머신 러닝 시스템 및 방법 | |
El Ouafi et al. | An on-line ANN-based approach for quality estimation in resistance spot welding | |
Ambrosio et al. | Machine learning tools for flow-related defects detection in friction stir welding | |
US8426771B2 (en) | Method of monitoring machine condition | |
Jin et al. | Quality prediction and control in rolling processes using logistic regression | |
Adams et al. | Correlating variations in the dynamic resistance signature to weld strength in resistance spot welding using principal component analysis | |
CN113487149B (zh) | 基于Catboost K折交叉验证的焊点异常识别***及方法 | |
EP3888839B1 (en) | Method and system for monitoring and identifying the weld quality of a welding performed by a welding machinery on metallic components | |
Dhanraj et al. | A credal decision tree classifier approach for surface condition monitoring of friction stir weldment through vibration patterns | |
Dimla Jr et al. | Automatic tool state identification in a metal turning operation using MLP neural networks and multivariate process parameters | |
Xu et al. | Quality monitoring for resistance spot welding using dynamic signals | |
Sah et al. | An experimental study of contact pressure distribution in panel stamping operations | |
Kershaw et al. | Advanced process characterization and machine learning-based correlations between interdiffusion layer and expulsion in spot welding | |
Wang et al. | INTERPRETABLE DATA-DRIVEN PREDICTION OF RESISTANCE SPOT WELD QUALITY | |
JP5055797B2 (ja) | 溶接品質判定装置および溶接品質判定方法 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 11832543 Country of ref document: EP Kind code of ref document: A1 |
|
DPE1 | Request for preliminary examination filed after expiration of 19th month from priority date (pct application filed from 20040101) | ||
ENP | Entry into the national phase |
Ref document number: 2012538686 Country of ref document: JP Kind code of ref document: A |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2011832543 Country of ref document: EP |
|
WWE | Wipo information: entry into national phase |
Ref document number: 13878917 Country of ref document: US |