JP2007333687A - Quality determination device and method for article using time-series signal - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a quality determination device and a quality determination method capable of reducing an average inspection time required for inspection per article to be inspected. <P>SOLUTION: This quality determination device acquires at least one time-series signal produced from an article in a sensor part and uses it for quality determination in the determination part. The determination part performs a quality determination S104 at a time point S102 when the elapsed time from a predetermined time point timed in a timing part is a first period of time. When the article cannot be determined to be acceptable according to the quality determination S106, the determination part performs the quality determination S104 of the article again at the time point S102 when the elapsed time is a second period of time. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to an article pass / fail judgment device and a pass / fail judgment method using a time-series signal, and more specifically, determines whether or not the assembly is normally performed based on an assembly sound generated when the article is assembled. The present invention relates to a quality determination device and a quality determination method.

  Conventionally, a snap fit mechanism has been widely adopted as a mechanism for assembling a plurality of components. In order to improve productivity when manufacturing an assembly using such a snap-fit mechanism, it is desirable to automatically and quickly determine whether or not the assembly has been performed normally. In view of this, a determination device has been developed that collects fitting sounds generated during assembly and performs automatic determination based on the fitting sounds (see Patent Document 1).

  When the assembly has a plurality of snap-fit mechanisms, the determination device described in Patent Document 1 continues sound collection from the start of assembly until all snap-fit mechanisms are completed, and then performs determination. On the other hand, depending on the configuration of the assembly or the manufacturing apparatus, all the snap-fit mechanisms may not be assembled within a certain time, and may vary from assembly to assembly. Therefore, it is necessary to set the sound collection time sufficiently long in order to prevent an assembly that has been normally assembled from being erroneously determined as a defective product. However, if the sound collection time is set longer, there is a problem that the time required for determination becomes longer.

  Moreover, the determination apparatus described in Patent Document 1 performs pass / fail determination by performing frequency analysis on the collected fitting sound and comparing the sound pressure level of the specific frequency component with a predetermined threshold value. However, when assembling the parts of the assembly, the pressure of pressing the parts may fluctuate, or the characteristics of the fitting sound may change due to variations in the lots of parts constituting the assembly. Similarly, the characteristics may change due to changes in the manufacturing environment (environmental temperature, humidity, etc.). In such a case, the preset threshold value becomes inappropriate, and there is a possibility that the good product and the defective product cannot be accurately distinguished.

Japanese Patent Laid-Open No. 2004-243501

  An object of this invention is to provide the quality determination apparatus and quality determination method which make it possible to eliminate the problem by the prior art mentioned above.

  Another object of the present invention is to provide a quality determination device and a quality determination method capable of reducing an average inspection time required for inspection per article to be inspected.

  It is another object of the present invention to provide a quality determination device and a quality determination method that can accurately determine the quality of assembly of an assembly that is an object to be inspected even when environmental changes or component variations occur.

  According to the first aspect of the present invention, the quality determination device according to the present invention acquires at least one time-series signal generated from the article (2) to be inspected by the sensor unit (10, 11). And it uses for the quality determination in a determination part (19). The determination unit (19) performs pass / fail determination when the elapsed time from the predetermined time measured by the time measuring unit (20) becomes the first period. As a result of the pass / fail determination, if the article cannot be determined to be a non-defective product, the determination unit (19) performs the pass / fail determination again when the elapsed time reaches the second period.

  According to the sixth aspect of the present invention, the quality determination method according to the present invention includes a step (S101) of acquiring a time series signal generated from the article (2) to be inspected, and a time elapsed from a predetermined time point. The step of determining whether or not the elapsed time has reached the first period (S102), and the first pass or fail that determines whether or not the article is good based on the time series signal when the elapsed time has reached the first period In the determination step (S104), the step (S102) for determining whether or not the elapsed time has reached the second period, and the first quality determination step, the article (2) is not determined to be a good product and the elapsed time Has a second pass / fail judgment step (S104) for judging the pass / fail of the article again based on the time series signal when the second period is reached.

  Only when the inspection of the inspection object cannot be completed within the first period set in a relatively short period, the time required for the inspection of each article to be inspected is averaged by performing the pass / fail determination again. It becomes possible to shorten to.

  In addition, as described in claim 2, in the quality determination device according to the present invention, the determination unit (19) measures the first time-series signal measured for a non-defective article and the article for a defective article. It is preferable to have a determination criterion set in advance by a statistical analysis method based on the second time series signal, and determine the quality of the article (2) to be inspected using the determination criterion.

Similarly, according to the seventh aspect of the present invention, in the quality determination method according to the present invention, the first and second determination steps (S104) include the first time-series signal measured for the non-defective article and the first time-series signal. It is preferable to determine the quality of the article (2) based on the second time series signal measured for the non-defective article, using a criterion set in advance by a statistical analysis technique.
By using a statistical analysis method, it is possible to accurately determine the quality of an article even when the discrimination boundary between a non-defective product and a defective product is complicated.

Furthermore, as described in claim 3 or claim 8, the article (2) comprises a plurality of parts (110, 120), and at least one of the time-series signals comprises a plurality of parts (110, 120). It is preferable that the audio signal is generated during assembly.
By using the audio signal at the time of assembling the parts, it can be determined whether or not the assembling has been normally performed, so that it is possible to accurately determine the quality of the article based on whether or not the assembling of the parts is normal.

Furthermore, as described in claim 4, the quality determination device according to the present invention includes a frequency analysis unit (17) that calculates a time-series spectrum for each of a plurality of frequency bands from an audio signal, and the determination unit (19) It is preferable to perform pass / fail judgment based on at least a time-series spectrum for each of a plurality of frequency bands.
Similarly, as described in claim 9, the quality determination method according to the present invention further includes a step (S103) of calculating a time-series spectrum for each of a plurality of frequency bands from the audio signal. In the determination step (S104), it is preferable to perform pass / fail determination based on at least a time-series spectrum for each of a plurality of frequency bands.

  By using the time-series spectra of multiple frequency bands for pass / fail judgment, whether or not assembly was performed correctly even when the time-series spectrum of the mating sound generated during assembly changes due to component variations, etc. It can be determined whether or not.

  Furthermore, as described in claim 5, the sensor unit is a contact pressure detection sensor (11) that measures a contact pressure generated by contact between the components (110, 120) when the components (110, 120) are assembled. The time series signal preferably includes a pressure signal acquired by the contact pressure detection sensor (11), and the determination unit (19) preferably performs pass / fail determination based on the audio signal and the pressure signal. Since the variation factor of the audio signal generated when assembling the parts is also taken into consideration, the quality of the inspection target article can be determined more accurately.

  Similarly, as described in claim 10, the time-series signal includes a pressure signal obtained by measuring a contact pressure caused by contact between the parts (110, 120) when the parts (110, 120) are assembled. In the first and second pass / fail determination steps (S104), it is preferable to perform pass / fail determination based on the audio signal and the pressure signal.

  In addition, the code | symbol in the parenthesis attached | subjected to each said means is an example which shows a corresponding relationship with the specific means as described in embodiment mentioned later.

Hereinafter, an embodiment of a quality determination device according to the present invention will be described in detail with reference to the drawings.
The pass / fail judgment device according to the present invention performs pass / fail judgment at the time when a fitting sound at the time of assembly of an assembly having a plurality of snap-fit mechanisms is collected for a relatively short time, and ends the process when it is judged to be a non-defective product. Only when it is not determined, the fitting sound is continuously collected and the determination is performed again. Therefore, since the determination can be completed in a short time for the assembly that is completed in a short time, the average time required for one inspection can be shortened. In addition, since the pass / fail determination apparatus according to the present invention determines pass / fail statistically in consideration of parameters that cause a variation in the fitting sound, such as contact pressure between components during assembly, the fitting sound varies. Even if it is a case, the quality can be determined accurately.

In FIG. 1, the schematic block diagram of the quality determination apparatus 1 which concerns on this invention is shown.
The quality determination device 1 includes a microphone 10, a contact pressure detection sensor 11, an amplifier 12, an A / D converter 13, a storage unit 14, an operation display unit 15, and a control unit 16. And the sound which generate | occur | produces at the time of the assembly | attachment of the assembly 2 which is a test object is collected with the microphone 10, and it acquires as an electrical signal. The electric signal is amplified by an amplifier 12 electrically connected to the microphone 10. Further, the amplified signal is converted from an analog signal to a digital signal by an A / D converter 13 electrically connected to the amplifier 12. The digital signal is sent to a control unit 16 electrically connected to the A / D converter 13. Similarly, the contact pressure detection sensor 11 detects the contact pressure when the components of the assembly 2 are in contact with each other during assembly, and converts the pressure into a digital signal via the amplifier 12 and the A / D converter 13. The data is sent to the control unit 16. The control unit 16 records these digital signals in the storage unit 14 and determines the quality of the assembly 2 based on the digital signals. The quality determination result is displayed by the operation display unit 15.

In FIG. 2, the schematic of the assembly 2 which is a test object is shown. FIG. 2A is a schematic plan view of the assembly 2. Moreover, FIG.2 (b)-FIG.2 (d) are the schematic side views of the assembly 2 shown along the procedure at the time of an assembly | attachment, respectively.
As shown in each of FIGS. 2A to 2D, the assembly 2 to be inspected includes two parts 110 and 120. A groove 108 is formed at the center of the component 110. The component 110 can be bent along the groove 108. The left part A and the right part B with the groove 108 of the part 110 as a boundary are provided with leg portions extending downward at three places, and each leg portion has a fitting hole 101A. And 101B are formed. On the other hand, the component 120 has protrusions 104A and 104B corresponding to the fitting holes 101A and 101B. Then, a snap-fit mechanism for assembling the component 110 and the component 120 is configured by the fitting holes and the corresponding protrusions.

When assembling the assembly 2, first, as shown in FIG. 2C, the fitting hole 101 </ b> A provided in the part A of the component 110 and the corresponding protrusion 104 </ b> A of the component 120 are fitted. Thereafter, as shown in FIG. 2D, pressure is applied to the B portion of the component 110, and the fitting hole 101B provided in the B portion and the corresponding protrusion 104B of the component 120 are fitted.
At the time of this assembly, a fitting sound is generated when each of the A part and the B part is assembled. However, the length of the interval period from assembling part A to the end of assembling part B varies depending on various conditions during assembly.

Hereinafter, each part will be described in detail.
The microphone 10 collects the fitting sound generated by assembling the assembly 2 and outputs the measurement signal as a time series signal. In this embodiment, a condenser microphone having a frequency band of 20 Hz to 20 kHz is used. However, as the microphone 10, another type of microphone such as a piezoelectric microphone may be used. Further, the frequency band of the sound that can be collected is not limited to the above, and for example, a frequency band of 100 Hz to 18 kHz may be used. As the microphone system and the measurable frequency band, an optimal one can be selected as appropriate depending on the type and size of the inspection object, the structure of the snap-fit mechanism, and the like.

  The contact pressure detection sensor 11 is provided at the tip of a member that presses the component 110 in an assembly apparatus (not shown) for assembling the component 110 and the component 120 of the assembly 2. Then, the contact pressure received from the component 110 when the component 110 is assembled to the component 120 is detected. The contact pressure detection sensor 11 converts the detected contact pressure into a time-series electrical signal and sends it to the amplifier 12.

The amplifier 12 amplifies the time series signal acquired from the microphone 10 and the contact pressure detection sensor 11 and outputs the amplified signal to the A / D converter 13. The A / D converter 13 converts each amplified time series signal from analog to digital and outputs the result. In the present embodiment, the A / D converter 13 samples the input analog signal at a sampling frequency of 44.1 kHz and outputs it as a digital signal with a resolution of 16 bits. Note that the A / D converter is not limited to the above, and ones having different sampling frequencies and resolutions may be used. The sampling frequency that can be used only needs to exceed twice the frequency band for checking the intensity level of the spectrum described later. In addition, the usable resolution is not limited as long as the signal value in each frequency band can be distinguished from a non-defective product and a defective product. For example, the A / D converter 13 may have a sampling frequency of 48 kHz and a resolution of 12 bits. Further, the amplifier 12 may include a plurality of amplifiers, and the time series signal from the microphone 10 and the time series signal from the contact pressure detection sensor 11 may be amplified by separate amplifiers. Similarly, the A / D converter 13 also includes a plurality of A / D converters, and the time series signal from the microphone 10 and the time series signal from the contact pressure detection sensor 11 are separate A / D converters. Analog-digital conversion may be performed.
The digitized measurement signal is sent to the control unit 16.

  The storage unit 14 includes a magnetic recording medium such as a hard disk, a random access memory (RAM), a semiconductor memory such as a flash memory, and a readable / writable optical recording medium such as a CD-RW and a DVD-R / W. The storage unit 14 stores programs used in the control unit 16, various parameters, threshold values, pass / fail judgment results of the assembly 2 to be inspected, time series signals measured by the microphone 10, and the like.

The operation display unit 15 is operated by an operator or displays a pass / fail judgment result, and is configured by a liquid crystal display with a touch panel. Then, according to the operation guidance displayed on the screen of the operation display unit 15, when the measurer touches a predetermined area on the screen, the operation display unit 15 controls a signal associated with the area touched by the measurer. Transmit to unit 16. When the control unit 16 obtains a pass / fail judgment result, the pass / fail judgment result is displayed to notify the operator.
The operation display unit 15 may include a display device other than a liquid crystal display such as a CRT and a pointing device such as a mouse.

The control unit 16 includes a central processing unit (CPU), a read only memory (ROM), a random access memory (RAM), and a program for causing the CPU to perform a predetermined operation.
FIG. 3 shows a functional block diagram of the control unit 16. The control unit 16 includes a frequency analysis unit 17, a signal correction unit 18, a determination unit 19, and a time measurement unit 20, and determines whether the assembly 2 is acceptable based on the digitized time series signal. Each of these units is mounted as a functional module by a program that operates on the CPU. Or you may mount by the arithmetic circuit for exclusive use which implement | achieves the function of each part.

ここでｈ（ｔ）は、測定開始からの経過時間ｔにおける、時系列信号の強度であり、関数Ψ（ｙ）は、マザーウェーブレットである。また、ａは周波数を決定するためのパラメータ、ｂは時間軸上の平行移動量を規定するパラメータである。 The frequency analysis unit 17 obtains a time series spectrum for each of a plurality of frequency bands with respect to the time series signal of the voice acquired from the microphone 10.
Here, the frequency analysis unit 17 calculates the spectral intensity of the time-series signal for each frequency band as a function of the elapsed time from the start of measurement, based on the wavelet transform, for the audio time-series signal. As is well known, the wavelet transform is performed based on the following equation.

Here, h (t) is the intensity of the time series signal at the elapsed time t from the start of measurement, and the function Ψ (y) is a mother wavelet. Further, a is a parameter for determining the frequency, and b is a parameter for defining the parallel movement amount on the time axis.

  In this embodiment, a Gabor function is used as the mother wavelet. The shape of the mother wavelet is defined by a one-dimensional array MW and stored in the storage unit 14 in advance. Then, when performing frequency analysis, the one-dimensional array MW is read from the storage unit 14 and used.

In the present embodiment, the time series spectrum of the measured audio signal is calculated for each frequency band obtained by dividing the frequency from 20 Hz to 20 kHz into 8 logarithms. The frequency band classification is not limited to the above, and the number of frequency bands may be increased or decreased as necessary. Further, the frequency band for calculating the time series spectrum may be narrowed or widened as necessary.
The calculated time-series spectrum for each frequency band is sent to the signal correction unit 18.

  In this embodiment, a Gabor function is used as the mother wavelet. However, instead of the Gabor function, another function used as a mother wavelet in signal analysis such as speech analysis, such as a Mexican hat function, may be used. .

The signal correction unit 18 attenuates a component related to a sound unrelated to the fitting sound generated when the assembly 2 is assembled with respect to the time-series spectrum for each frequency band. For this purpose, the operating sound of an assembling apparatus (not shown) for assembling the assembly 2 and the fitting sound when assembling the assembly 2 are collected in advance and the frequency spectrum is analyzed. For example, it is observed that the frequency spectrum of the operating sound of the assembling device is increased in a frequency band of 2.8 kHz or less, while the frequency spectrum of the fitting sound is observed to be increased in a frequency band of 2.8 kHz or more. In this case, the signal correction unit 18 performs a high-pass filtering process that attenuates the time-series spectrum in the frequency band of 2.8 kHz or less.
Each time-series spectrum subjected to signal correction is sent to the determination unit 19.

The determination unit 19 determines whether or not the assembly 2 has been normally assembled based on each time-series spectrum with the signal corrected and the pressure measurement value acquired from the contact pressure detection sensor 11.
The relationship between the contact pressure between the component 110 and the component 120 and the fitting sound will be described with reference to FIGS. 4 (a) to 4 (e). 4 (a) and 4 (b) are graphs showing the relationship between elapsed time and pressure during assembly. FIG. 4 (a) shows a relatively high pressure, and FIG. 4 (b) shows a relatively high pressure. This corresponds to the case where is small. In FIGS. 4A and 4B, the horizontal axis represents the elapsed time t, and the vertical axis represents the pressure p. In FIG. 4A and FIG. 4B, the snap-fit mechanisms of the parts 3 and 4 are fitted in regions 31 and 32 surrounded by dotted lines, respectively.

  4 (c) and 4 (d) show time-series spectra in a specific frequency band for sounds generated during assembly. FIG. 4C corresponds to FIG. 4A and represents a time-series spectrum when the pressure during assembly is relatively large. FIG. 4 (d) corresponds to FIG. 4 (b) and represents a time-series spectrum when the pressure during assembly is relatively small. 4C and 4D, the horizontal axis represents the elapsed time t, and the vertical axis represents the time series spectrum amplitude s (corresponding to sound pressure). Moreover, the area | region 33 of FIG.4 (c) represents the same timing as the area | region 31 of Fig.4 (a), and the time-series spectrum of this period respond | corresponds to a fitting sound. Similarly, the region 34 in FIG. 4D represents the same timing as the region 32 in FIG. 4B, and the time-series spectrum in this period corresponds to the fitting sound. As is clear from FIGS. 4C and 4D, when the pressure at the time of assembly is large, the amplitude of the time series spectrum of the generated fitting sound also increases.

  FIG. 4E schematically shows the distribution of non-defective products and defective products of the assembling apparatus 2 regarding the amplitude of the time-series spectrum of the fitting sound and the pressure at the time of assembly. The horizontal axis of FIG. 4E represents the value Is obtained by integrating the amplitude of the time series spectrum over the generation period of the fitting sound, and the vertical axis represents the pressure p during assembly. In FIG. 4E, each point 35 indicated by a circle represents a time-series spectrum and pressure for a non-defective product of the assembly 2, and each point 36 indicated by a square represents a time for a defective product of the assembly 2. Represents series spectrum and pressure. As shown in FIG. 4E, when the pressure at the time of assembly fluctuates, if the pass / fail judgment of the assembly 2 is performed based only on the amplitude of the time series spectrum of the fitting sound, the boundary line of the pass / fail judgment criterion is a line. Since it becomes a straight line parallel to the vertical axis of the graph as shown in FIG. 37, it is clear that a good product and a defective product cannot be accurately distinguished. On the other hand, when the quality determination of the assembly 2 is performed in consideration of both the pressure and the time-series spectrum, the boundary line of the quality determination criteria can be defined as shown by the line 38. It becomes possible to distinguish.

  Therefore, the determination unit 19 compares the time series spectrum of the fitting sound at the time of assembly and the measured value of the pressure with a quality determination criterion determined in advance by a statistical analysis method. The assembly 2 is determined as a non-defective product. Otherwise, the assembly 2 is determined as a defective product. Various methods can be used as the statistical analysis method. For example, it is possible to set a linear determination criterion using discriminant analysis, principal component analysis, or the like as a statistical analysis method. In addition, using a neural network or a support vector machine as a statistical analysis method, it is possible to set a nonlinear determination criterion.

  Hereinafter, an example in which the pass / fail criterion is set using these statistical analysis methods will be described. First, the assembly 2 is assembled by varying the pressure during assembly in advance, and the pressure measurement value and fitting sound at that time are collected as data.

Here, with respect to the fitting sound, with respect to a specific frequency band in which the amplitude of the time-series spectrum increases when the fitting sound is generated, for each frequency band from a certain time t0 after the start of assembly to t1 after a certain period of time has elapsed. Integrates the absolute value of the time series spectrum. Then, the calculated integral value of the time-series spectrum of each frequency band (hereinafter referred to as a spectrum area) is used as an input parameter for setting a criterion. As for the pressure at the time of assembly, the average value or the maximum value of the pressure measurement values during the period from t0 to t1 is used as an input parameter for setting a determination criterion.
Note that t0 can be a point in time at which the amplitude of the time-series spectrum exceeds a predetermined value in any of the specific frequency bands. Alternatively, when the control unit 16 is configured to acquire the assembly start control signal from the assembly device, t0 is used as the assembly start control signal to exclude the operation sound of the assembly device unrelated to the fitting sound. It may be determined based on the time of receiving.

  In addition, for each assembly 2 for which data has been collected, the pressure measurement values that are input parameters and the spectrum area of each frequency band are associated with the pass / fail judgment result of the assembly 2 (determination of whether the product is good or defective at this time) Is performed by the operator by visual observation). Then, the average value of each input parameter is calculated for all the non-defective products of the assembly 2 from the collected data. A non-defective product average value vector Gav is expressed as a vector value having the average value of each input parameter as an element. Similarly, the average value of each input parameter is calculated for all defective products of the assembly 2 from the collected data. An average value of each input parameter expressed as a vector value having an element as an element is defined as a defective product average value vector Bav.

  Then, a plane s1 that is orthogonal to the straight line connecting the non-defective product average value vector Gav and the non-defective product average value vector Bav and is equidistant from the Gav and Bav is set as a determination criterion for the non-defective product and the defective product. In this case, when the set of input parameters obtained for the assembly 2 that is the inspection object is a vector V, when the vector V is present on the average value vector Gav side with respect to the surface s1, the assembly 2 is a non-defective product. It is determined.

  Another example of pass / fail judgment criteria will be shown. For example, a normal distribution is assumed for any input parameter. At this time, when the distribution of the normal distribution is different between the non-defective product of the assembled product 2 and the defective product of the assembled product 2, the Mahalanobis distance from the good product average value vector Gav and the defective product average value vector Bav is equal. The surface s2 represented by the discriminant function as follows can be used as a criterion. By setting the determination criterion in this way, the variance of the input parameters is taken into consideration, so that a more accurate pass / fail determination criterion can be set.

  The obtained determination criterion is represented by a discriminant function having each input parameter as a variable. For example, when the output value obtained by inputting each input parameter is larger than a determination value D (for example, D = 0), the assembled product 2 is determined as a non-defective product, and conversely, the determination function D is less than the determination value D. Is set to be judged as defective.

  As still another example of the pass / fail judgment criteria, the pass / fail judgment criteria may be set by using a three-layer perceptron model neural network or nonparametric learning such as a support vector machine. In this case, a set of input parameters relating to the assembly 2 for which data collection has been performed is input, and a pass / fail judgment result for the assembly 2 (for example, a value of “1” for a non-defective product and a value of “−1” for a defective product). A teacher data set is output. Then, the teacher data set is given to a learning system such as a neural network or a support vector to perform learning. Then, the learning is repeated until a correct output is obtained for each set of input parameters. The learning system learned in this way is used as a pass / fail criterion.

  When these non-parametric learning is used, even if it is not possible to linearly separate non-defective products and defective products by the dimension of the number of input parameters, it is possible to set a determination criterion capable of performing good / bad determination with high accuracy.

  Each coefficient and determination value D of the discriminant function that represents the created determination criterion, the coupling load of the neural network, and the like are stored in advance in the storage unit 14. Then, when the pass / fail determination is executed, these values are read out and used by the determination unit 19.

Next, the operation of the determination unit 19 during the pass / fail determination execution will be described.
When the determination unit 19 receives the notification of the determination execution permission signal from the time measuring unit 20, the determination unit 19 reads the pressure measurement value acquired from the contact pressure detection sensor 11 from the time t0 until the notification of the determination permission signal is received. Similarly, the spectrum area is calculated from the time-series spectrum of the specific frequency band obtained based on the digital signal of the fitting sound acquired from the microphone 10 between the times t0 and t1. Note that the period from t0 to notification of the determination execution permission signal is set to Δt1 as in the case of creating the above-described determination reference.

Then, when a determination function corresponding to the surfaces s1 and s2 is given as a determination criterion, the determination unit 19 determines a determination function based on input parameters such as the calculated pressure measurement value and each spectrum area. Calculate the output value of. When the output value is larger than the determination value D, the assembly 2 is determined as a non-defective product. On the other hand, when the output value is equal to or less than the above-described determination value D, it is determined that the assembly 2 may not be a good product.
Alternatively, when a learning system such as a neural network is given as a determination criterion, the determination unit 19 inputs input parameters such as the calculated pressure measurement value and each spectrum area into the learning system, and the obtained output Check the value. When the output value is a value representing a non-defective product or closer to a value representing a non-defective product than a value representing a defective product, the determination unit 19 determines that the assembled product 2 is a non-defective product. Judge that there is sex.
When the determination unit 19 determines that the assembled product 2 is a non-defective product, the determination unit 19 transmits a timing stop signal to the timing unit 20.

  Here, as the period Δt1 from t0 to t1 is set shorter, the time required for pass / fail judgment can be shortened. However, as described above with reference to FIG. 2, in the present embodiment, since the A portion of the component 3 is fitted and then the B portion of the component 3 is fitted, the snap provided in the A portion. There is a slight interval from the fitting of the fitting mechanism to the fitting of the snap-fit mechanism provided in part B. The interval varies depending on the variation of parts 3 or 4, the positional relationship between parts 3 or 4 and the assembly device, etc. There is a case. Therefore, if the period Δt1 from t0 to t1 is too short, the quality determination is performed before the assembly of the assembly 2 is finished, and there is a possibility that the product that should be determined as a non-defective product is determined as a defective product. .

Therefore, in the first pass / fail determination, the determination unit 19 executes the pass / fail determination again when the assembly 2 cannot be determined as a non-defective product. That is, as will be described later, when the determination unit 19 receives the determination execution permission signal again from the time measuring unit 20 at the time t2 after the elapse of the period Δt2 from the time point t1, the determination unit 19 performs the pass / fail determination again.
When the pass / fail determination is re-executed, the determination unit 19 recalculates the pressure measurement value and each spectrum area used for the pass / fail determination by replacing t1 with t2. That is, regarding the pressure, the average value or the maximum value of the pressure measurement values in the period from time t0 to t2 is used as the input parameter. Further, regarding the spectrum area, the integration period is calculated from t0 to t2.

  When the recalculation of the input parameters is completed, the determination unit 19 applies these input parameters to the determination criterion again, and determines whether or not the assembly 2 is a non-defective product. Similarly to the above, the determination unit 19 determines whether the output value of the discriminant function calculated based on the input parameter is larger than the determination value D, or the value indicating the non-defective product or the value of the learning system output value. If it is close to, the assembly 2 is determined as a non-defective product. On the other hand, when the output value is equal to or less than the above-described determination value D, or when the learning system output value is a value representing a defective product or close to the value, the assembly 2 is determined to be a defective product.

  Note that Δt1 and Δt2 can be determined by various criteria. For example, a predetermined number (for example, 100) of assemblies 2 is assembled in advance, and the time (assembly time) required for assembly of each assembly 2 at that time is checked. Then, the longest assembly time among the remaining assembled products 2 is set to Δt1 except for those having an extremely long assembly time or a predetermined ratio (for example, about 10%) of the longer assembly time. . Further, the maximum value of the assembly time is obtained for a predetermined number of the assembled products 2 that have been assembled as described above, and Δt2 is set so that the maximum value is (Δt1 + Δt2). By setting Δt1 and Δt2 in this way, it is possible to make a pass / fail judgment in a short time for most of the assembled products 2, and only one part with a long assembly time is judged again. It is possible to accurately perform the pass / fail determination for all the assemblies 2 while shortening the average value of the time required for the pass / fail determination.

  Furthermore, the determination unit 19 may repeat the pass / fail determination three or more times. In this case, the period Δt1 from t0 to t1 at which the first determination is performed may be set to a shorter value. For example, Δt1 is set to an average value of the time required for one assembling or less than that when a predetermined number of assembling is performed as described above. The average value of the time required for inspecting one assembled product can be further shortened by repeatedly executing the second and subsequent determinations with a short period of about Δt1 / 5.

The timing unit 20 determines the timing for performing the pass / fail determination. Therefore, the time measuring unit 20 measures the elapsed time t from the time t0. Then, when it is time to perform the pass / fail determination, the determination unit 19 is notified of a determination execution permission signal for permitting determination. Here, when the determination unit 19 makes a pass / fail determination at a maximum of two times in one inspection, the time measuring unit 20 notifies the determination unit 19 of a determination execution permission signal when the elapsed time t reaches a predetermined time Δt1. To do.
In addition, when the determination unit 19 determines that the assembled product 2 is a non-defective product when the determination unit 19 receives a time stop signal from the determination unit 19, the time measurement is stopped, and the elapsed time t during time measurement is reset to 0 and initialized. On the other hand, when the time measurement stop signal is not received, the time measurement unit 20 notifies the determination unit 19 of the determination execution permission signal when the elapsed time t has further passed the time Δt2 from Δt1. Then, the timer unit 20 stops timing and initializes the elapsed time t.

  In addition, when the determination unit 19 performs the pass / fail determination three or more times in one inspection, the timing stop signal is received after the second notification or the determination execution permission signal is determined as many times as possible. Until the notification of the determination execution permission signal is repeated at a predetermined interval Δt2.

  Hereinafter, the operation of the quality determination device 1 according to the present invention will be described with reference to the operation flowcharts shown in FIGS. 5 and 6. In the operation described below, the number of pass / fail judgments performed in one inspection is not limited to 2. However, even when the number of pass / fail judgments is 2 at the maximum, the number of possible judgments in step S106 is 2. Then, the same operation flow is obtained. Further, the operation of the pass / fail determination apparatus 1 is controlled by a program read into the control unit 16.

  As shown in FIG. 5, when the inspection is started, the control unit 16 starts acquiring time-series signals from the microphone 10 and the contact pressure detection sensor 11 (step S <b> 101). On the other hand, the timing unit 20 of the control unit 16 starts counting the elapsed time t from a predetermined time t0 after the start of measurement. Then, the timer unit 20 checks whether or not the elapsed time t has passed a predetermined period (step S102). When the elapsed time t is less than the predetermined period, the control unit 16 returns the control to before step S102. On the other hand, when the elapsed time is equal to or longer than the predetermined period, the frequency analysis unit 17 of the control unit 16 performs frequency analysis on the digital audio signal acquired from the microphone 10 and obtains a time-series spectrum for each of a plurality of frequency bands ( Step S103). Further, the signal correction unit 18 corrects each time series spectrum.

  Thereafter, the determination unit 19 of the control unit 16 determines the quality of the assembly 2 based on the spectrum area calculated from the time-series spectrum of a specific band and the pressure measurement value acquired from the contact pressure detection sensor 11 (step S104). ). The pass / fail judgment is performed based on a statistical analysis method such as discriminant analysis as described above. If the determination unit 19 determines that the assembly 2 is a non-defective product as a result of the pass / fail determination, the control unit 16 transmits a signal indicating that the assembly 2 is a non-defective product to the operation display unit 15 (step S105). Then, measurement by the microphone 10 and the contact pressure detection sensor 11 is stopped (step S108), and the inspection is finished.

On the other hand, if it is not determined in step S104 that the assembly 2 is a non-defective product, the determination unit 16 checks whether or not the pass / fail determination count from the start of timing has reached the determination possible count (step S106). When the pass / fail judgment count reaches the countable count, the determination unit 19 determines that the assembly 2 is defective, and the control unit 16 determines that the assembly 2 is defective with respect to the operation display unit 15. A signal indicating this is transmitted (step S107). Then, measurement by the microphone 10 and the contact pressure detection sensor 11 is stopped (step S108), and the inspection is finished.
In step S106, when the pass / fail determination count does not reach the determination possible count, the control unit 16 increments the pass / fail determination count by 1, and returns the control to step S102. And the process after step S102 is repeated again.

  As described above, the quality determination device 1 according to the present invention determines the quality of assembly at a stage where the elapsed time from the start of assembly is small, and performs determination again only for those that could not be determined to be non-defective by the determination. Therefore, the inspection time required when inspecting a large number of assemblies can be shortened. In addition, not only the audio signal but also the contact pressure between the parts of the assembly is used for statistical analysis to determine whether the assembly is good or not. Even when the contact pressure is not constant, Can be determined.

The quality determination device according to the present invention is not limited to the above-described embodiment.
For example, in the above-described embodiment, the quality determination device according to the present invention is described as determining whether the assembly product has a plurality of snap-fit mechanisms, but the quality determination device according to the present invention is Although a predetermined operation is performed over a certain period and a plurality of sounds are generated in the operation, it can be used for quality determination.

  Moreover, in said embodiment, although the time series spectrum of the fitting sound at the time of an assembly | attachment and the pressure which presses the components of an assembly are used as information used for quality determination, the information which can be used is not restricted to these. For example, when the fitting sound changes depending on the environmental temperature or the environmental humidity, a temperature sensor or a humidity sensor is attached to the pass / fail judgment device, and the temperature information or the humidity information acquired from the sensor is combined with the pressure information or It may be used instead of pressure information.

  Further, the elapsed time from the start of measurement or the first pass / fail judgment to the second and subsequent pass / fail judgments may be added to the information used for the pass / fail judgment. In this case, it is estimated that the longer the elapsed time, the higher the possibility that the assembled product is defective.Instead of using this elapsed time as an input parameter for the above statistical analysis, the longer the elapsed time, the better the product. The determination criterion may be corrected so that the rate to be determined decreases.

Furthermore, in the above-described embodiment, the pass / fail determination is performed based on the audio signal and the pressure signal. However, if the pressure signal can be handled as constant, the pass / fail determination may be performed based only on the audio signal. Even in such a case, if the frequency of the fitting sound at the time of assembly fluctuates due to component lot variations, etc., a statistical analysis method using the spectrum signal of each frequency band as an input parameter as described above It is preferable to perform pass / fail determination based on the above. By using a statistical analysis method, it is possible to set a determination criterion in consideration of the frequency variation of the fitting sound, and therefore it is possible to accurately determine whether or not it is good as compared with simple threshold processing.
In addition, the quality determination apparatus may further include an alarm device that notifies the determination result by voice when the assembly to be inspected is determined to be defective.

  As described above, the quality determination device according to the present invention can be appropriately optimized within the scope of the present invention.

It is a schematic block diagram of the quality determination apparatus which concerns on this invention. (A) is a schematic plan view of the assembly to be inspected, and (b) to (d) are schematic side views showing the procedure for assembling the assembly. It is a functional block diagram of a control part. The relationship between the pressure at the time of assembling the assembly and the fitting sound is shown, (a) and (b) are graphs showing the pressure at the time of assembling, and (c) and (d) are time series spectra of the fitting sound. (E) is a graph showing the distribution of the assembly regarding a pressure measurement value and a time series spectrum. It is an operation | movement flowchart of the quality determination apparatus which concerns on this invention.

Explanation of symbols

1 Pass / Fail Judgment Device 2 Assembly 10 Microphone (Sensor)
11 Contact pressure detection sensor (sensor part)
DESCRIPTION OF SYMBOLS 12 Amplifier 13 A / D converter 14 Memory | storage part 15 Operation display part 16 Control part 17 Frequency analysis part 18 Signal correction part 19 Judgment part 20 Timing part 110,120 Parts

Claims (10)

A sensor unit (10, 11) for acquiring at least one time-series signal generated from the article (2);
A timekeeping section (20) for measuring an elapsed time from a predetermined time point and notifying that the elapsed time has become a first period and a second period longer than the first period;
When the time measuring unit (20) notifies that the elapsed time has reached the first period, the quality of the article (2) is determined based on the time series signal, and the article (2) When it is determined that the product is not a non-defective product, when the elapsed time is notified from the time measuring unit (20), the quality of the article (2) is determined again based on the time series signal. A determination unit (19);
A pass / fail judgment device characterized by comprising:   The determination unit (19) is preset by a statistical analysis method based on a first time series signal measured for a non-defective article and a second time series signal measured for a non-defective article. The quality determination apparatus according to claim 1, wherein the quality determination apparatus determines whether the article (2) is good or bad using the determination standard. The article (2) includes a plurality of parts (110, 120), and at least one of the time series signals is an audio signal generated when the plurality of parts (110, 120) are assembled.
The pass / fail determination apparatus according to claim 1 or 2, wherein the sensor unit (10, 11) includes a microphone (10) that acquires the audio signal. A frequency analysis unit (17) for calculating a time-series spectrum for each of a plurality of frequency bands from the audio signal;
The quality determination device according to claim 3, wherein the determination unit (19) performs quality determination based on at least a time-series spectrum for each of the plurality of frequency bands. The sensor unit (10, 11) further includes a contact pressure detection sensor (11) that measures a contact pressure generated by contact between the components (110, 120) when the components (110, 120) are assembled. The time series signal includes a pressure signal acquired by the contact pressure detection sensor (11),
The quality determination apparatus according to claim 3 or 4, wherein the determination unit (19) performs quality determination based on the audio signal and the pressure signal. A method for determining the quality of an article (2),
Obtaining a time-series signal generated from the article (2) (S101);
A step of measuring an elapsed time from a predetermined time point and determining whether or not the elapsed time has reached the first period (S102);
A first pass / fail determination step (S104) for determining pass / fail of the article based on the time-series signal when the elapsed time reaches the first period;
Determining whether the elapsed time has reached a second period (S102);
In the first pass / fail judgment step, if the article (2) is not judged as a good product and the elapsed time has reached the second period, the pass / fail of the article (2) is again determined based on the time-series signal. A second pass / fail judgment step (S104) for judging
A pass / fail judgment method characterized by comprising:   The first and second determination steps (S104) are based on statistics based on a first time series signal measured for a non-defective article and a second time series signal measured for a non-defective article. The quality determination method according to claim 6, wherein the quality of the article (2) is determined using a criterion set in advance by an analytical method.   The article (2) includes a plurality of parts (110, 120), and at least one of the time series signals is an audio signal generated when the plurality of parts (110, 120) are assembled. Or the quality determination method of 7. Calculating a time-series spectrum for each of a plurality of frequency bands from the audio signal (S103),
The quality determination method according to claim 8, wherein the first and second determination steps (S104) determine quality based on at least a time-series spectrum for each of the plurality of frequency bands. The time series signal includes a pressure signal obtained by measuring a contact pressure generated by contact between the parts (110, 120) when the parts (110, 120) are assembled.
The pass / fail determination method according to claim 8 or 9, wherein the first and second determination steps (S104) perform pass / fail determination based on the audio signal and the pressure signal.

Images