JP2004219110A - Oscillatory wave determining device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an oscillatory wave determining device capable of improving determination accuracy by allowing setting of a highly accurate threshold. <P>SOLUTION: The oscillatory wave determining device is provided with an oscillatory wave inputting means 2 for detecting and inputting an oscillatory wave generated during operation of an object, a frequency separation means 6 for forming a first time series signal per each predetermined frequency band by carrying out frequency separation on the inputted oscillatory wave, a signal processing means 7 for extracting a second time series signal per each predetermined frequency band with a target operation noise showing a predetermined operating state of the object from the first time series signal, a determining means 9 for determining a state of the object on the basis of a threshold and sound pressure information of the target operation noise determined from the second time series signal, and a correcting means 20 for correcting the threshold in response to a determination result of the determining means 9. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a vibration wave determination device that determines a state of a target object, for example, whether a fitting state is good or not, based on a vibration wave generated at the time of operation of the target object such as an assembly operation.
[0002]
[Prior art]
Conventionally, for example, in Patent Document 1, an object is lightly hit with a hammer or the like, and the sound or vibration generated at that time is frequency-separated by a wavelet transform operation means to form a time-series signal for each frequency band. A technique is described in which the content of an object is determined and the structure is inspected for cracks based on the determination of the sound pressure level of the series signal. In this specification, sound and vibration are collectively referred to as a vibration wave.
[0003]
[Patent Document 1]
JP-A-10-300730
[0004]
[Problems to be solved by the invention]
In this type of determination device, for example, when it is desired to accurately determine the quality of the state of an object, it is necessary to set a threshold value for the sound pressure level of the time-series signal with high accuracy. However, in the case of an inspection object having a low frequency of occurrence of a defective state, it is difficult to set a threshold value for determination with high accuracy because the sound pressure information itself generated in a defective state is small. Have been.
[0005]
The present invention has been made in view of the above point, and an object of the present invention is to provide a vibration wave determination device that can improve determination accuracy by enabling setting of a highly accurate threshold value.
[0006]
[Means for Solving the Problems]
In order to achieve the above object, the technical means described in claims 1 to 12 is adopted.
[0007]
According to the first aspect of the present invention, the vibration wave input means for detecting and inputting the vibration wave generated when the object is operated, and the first time series for each predetermined frequency band by separating the frequency of the input vibration wave. Frequency separation means for forming a signal, signal processing means for extracting, from the first time-series signal, a second time-series signal for each predetermined frequency band in which a target operation sound indicating a predetermined operation state of the object is generated, A determination unit configured to determine the state of the target object based on the sound pressure information of the target operation sound obtained from the second time-series signal and the threshold value; and a correction unit configured to correct the threshold value according to the determination result of the determination unit. It is characterized by the following.
[0008]
Thereby, it is possible to set a threshold value in accordance with the state of the target object by fixing the threshold value according to the determination result without fixing the threshold value used for the determination unit, thereby improving the determination accuracy. .
[0009]
According to the second aspect of the present invention, the rewritable threshold table storing the threshold used by the determination unit is provided, so that the threshold can be easily corrected in addition to the effect of the first aspect.
[0010]
According to the third aspect of the present invention, there is provided a sound inspection means for obtaining a sound pressure level defect distribution of the object based on the vibration wave obtained by temporarily setting the defect state of the object, and a threshold value used in the judgment means. Is characterized in that a sound pressure level exceeding a range of the defect distribution obtained by the sound inspection means is set as an initial threshold value.
[0011]
This makes it possible to set an initial threshold value in advance by obtaining a virtual failure distribution of the object, and to obtain failure information at an arbitrary frequency even in the case of an inspection object with a low frequency of occurrence of a failure state. Can be set with higher accuracy.
[0012]
According to the fourth aspect of the present invention, the correcting means collects sound pressure information of the target operating sound obtained from the second time-series signal at that time for each of the pass / fail judgment results by the judging means, statistically processes and obtains the sound pressure information. By modifying the threshold value used for the determination means based on the OK distribution information of the good determination result and the NG distribution information of the negative determination result, it is possible to set a threshold value for the sound pressure level according to both distribution states, thereby improving the determination accuracy. It becomes possible.
[0013]
According to the fifth aspect of the present invention, the correction means includes an OK data memory for storing sound pressure information when the object is in a good state, and an OK data memory when the object is in a bad state, according to the quality determination result of the determination means. NG data memory for storing the sound pressure information of the above, first statistical processing means for statistically processing the OK data stored in the OK data memory to obtain an OK distribution, and statistically processing the NG data stored in the NG data memory. Second statistical processing means for obtaining an NG distribution, and threshold calculating means for obtaining a new threshold based on the result of the statistical processing of the first and second statistical processing means and correcting the threshold used in the determining means It is characterized by.
[0014]
Thereby, in addition to the effect of the invention described in claim 5, the sound pressure information when the object is in the good state and the sound pressure information when the object is in the bad state are stored in separate memories. It is possible to set different storage processing methods, and it is easy to deal with defect status information that usually occurs less frequently. By storing both sound pressure information in separate memories, it is easy to obtain an NG distribution and an OK distribution, and it is easy to calculate a threshold.
[0015]
According to the invention described in claim 6, the signal processing means stores the correction means for correcting the level of the first time-series signal to form the second time-series signal, and the correction coefficient used by the correction means. With the correction table means, it is possible to reduce or cut a signal other than the target operation sound indicating the predetermined operation state of the target object, thereby improving the accuracy of the determination processing.
[0016]
According to the seventh aspect of the present invention, an input device that can be selected from the outside is provided, and the threshold value table means selects a threshold value used by the determination means in accordance with a selection command signal instructed by the input device. , The correction coefficient can be continuously selected according to an external selection command, and different objects can be determined continuously.
[0017]
According to the eighth aspect of the present invention, an input device which can be selected and operated from the outside is provided, and the correction table means selects a correction coefficient used by the correction means according to a selection command signal instructed by the input device. This has the same effect as the seventh aspect of the present invention.
[0018]
According to the ninth aspect of the present invention, the signal processing means has a filter processing means for providing the noise-removed time-series signal of the first time-series signal to the correction means, thereby indicating a predetermined operation state of the object. It is possible to reduce or cut a signal other than the operation sound to improve the accuracy of the determination processing.
[0019]
According to the tenth aspect of the present invention, the determining means includes a sound volume calculating means for obtaining sound pressure information by adding an area equivalent value based on the waveform shape of the second time series signal for each predetermined frequency band. The number of target operation sounds generated from the target object is determined based on the sound pressure level of the sound pressure information. As a result, even when the operating environment of the target object is different and the manner in which the target operation sound is generated varies, it is possible to improve the determination accuracy by using the added value, that is, the total volume. In particular, the present invention is effective for a process of determining a target operation sound generated from a plurality of locations on an object.
[0020]
According to the eleventh aspect of the present invention, the volume calculating means obtains an area equivalent value based on the waveform shape of the second time-series signal according to an envelope or a peak value of the second time-series signal, thereby achieving the desired operation. This makes it possible to more accurately detect the total volume of the sound and to improve the determination accuracy.
[0021]
According to the twelfth aspect of the present invention, there is provided a sound inspection means for obtaining a sound pressure level defect distribution of the object based on the vibration wave obtained by temporarily setting the defect state of the object, and the OK data memory and the NG Initially, the data memory initially stores the initial data obtained by the sound inspection means, and thereafter is sequentially rewritten with the latest sound pressure information of the target operation sound obtained from the second time-series signal, and The statistical processing means and the threshold value calculating means initially determine an initial threshold value based on the initial data and give the initial threshold value as the threshold value of the determining means.
[0022]
As a result, it is possible to ensure the accuracy of the determination of the target operation sound by using the relatively accurate initial threshold value from the beginning of the determination, and to rewrite the latest NG data or OK data with time to obtain a threshold value that is more suitable for the current situation. To make it possible to further improve the determination accuracy.
[0023]
BEST MODE FOR CARRYING OUT THE INVENTION
An embodiment of the present invention will be described with reference to the drawings.
[0024]
In the following description, an example will be described in which an assembling sound (fitting sound) generated at the time of assembling a component (or product) as an object is detected to determine a component assembling (fitting) defect. However, the vibration wave determination device 100 of the present invention is not limited to this use, and can be applied to other uses. Further, the present invention can be applied to detection of vibration instead of sound.
[0025]
FIG. 1 shows a system configuration according to an embodiment of the present invention.
[0026]
Here, as an example of the automatic assembly of resin parts as the target object 1 in this example, particularly a case where a loud noise is generated at the time of assembly, a snap-fit mechanism as shown in FIGS. Thus, an example of sound generation from a work process of assembling both parts 110 and 120 is given. FIG. 2 shows each projection 102 provided in each of the plurality of fitting portions 103 of the bottomed cylindrical lower lid component 120 in each hole 102 formed in the plurality of fitting portions 101 of the bottomed cylindrical upper lid component 110. The state of fitting is shown. In this case, the lower lid part 120 is inserted into the upper lid part 110, and the respective projections 104 are fitted into the respective holes 102 substantially simultaneously, and are fixed by the snap-fit mechanism.
[0027]
3A, 3B, and 3C are examples of vibration waves generated when assembling the two parts 110 and 120. FIG. 3A illustrates an assembling operation state, and FIG. (C) shows the vibration wave at the time of good, and at the time of poor assembly. As can be seen from the assembled sound levels in FIGS. 3B and 3C, the sound pressure level of the fitting sound changes according to the quality of the assembly.
[0028]
As a feature of sound generation at the time of assembly, a mechanical operation sound accompanying the movement of the assembly equipment and the like and a fitting sound that conveys the fitting state (that is, a target operating sound to be determined) are generated. In addition to having a sound pressure level corresponding to the magnitude of the assembling speed and the like, the sound pressure levels of both sounds change in correlation with each other.
[0029]
The microphone 2 detects a vibration generated in the object 1 for the vibration wave determination as a sound wave and converts the vibration into an electric signal. The electric signal of the sound pressure input from the microphone 2 is input to the amplifier 3 of the vibration wave determination device 100 and output to the A / D converter 4. The A / D converter 4 converts the sound pressure signal into a digital signal, and outputs the digital signal to the storage device 5 at the subsequent stage for storage processing.
[0030]
The wavelet transform (wavelet transform) calculator 6 constitutes a frequency separating means, takes in the digital sound pressure signal S0 stored in the storage device 5 at a predetermined timing, and converts the digital sound pressure signal S0 into a predetermined value. And separates them into time-series signals S1. In general, the wavelet transform calculator 6 is a calculator that separates the digital sound pressure signal S0 into a time series signal S1 for each frequency band by expanding or reducing a basis function (wavelet function). In this example, a frequency band that matches the fitting sound that is the target operation sound generated by one or more snap-fit mechanisms is set in advance as the assembling sound. This frequency band is composed of one or a plurality of frequency bands according to the characteristics of the target fitting sound.
[0031]
The signal processing means 7 has a filter processor 71 and a corrector 72, and has a function of extracting a time-series signal S2 in a frequency band corresponding to a fitting sound as a target operation sound from the time-series signal S1.
[0032]
The filter processor 71 outputs a time-series signal obtained by reducing or cutting a signal in a frequency band other than the fitting sound from the time-series signal S1 for each frequency band. It is to be noted that the filter processor 71 is omitted, and the time series signal S1 of the wavelet transform calculator 6 is directly supplied to the corrector 72, and a correction coefficient for each predetermined frequency band to be supplied to the corrector 72 is devised. Thus, it is also possible to reduce or cut a signal in a frequency band other than the fitting sound.
[0033]
The corrector 72 is an aggregate of one or more correctors 72a, 72b,... Which normally receives a time-series signal output from the filter processor 71. Each of the correctors 72a and 72b receives a correction coefficient (or a correction amount or a gain) set for each predetermined frequency band from the correction table 8, weights the time-series signal for each frequency band, and performs time-series signal Form S2. This is because the time-series signal S1 in a frequency band other than the fitting sound assumed in advance is regarded as noise and the level is lowered, and on the other hand, the frequency band with less noise in the fitting sound is amplified to improve the S / N ratio.
[0034]
The judging device 9 determines that all the fitting locations are the same even when the fitting sounds (target operating sounds) generated from the plurality of fitting locations in the object 1 overlap or slightly delay on the time axis. It is necessary to determine whether it is in a good state in which it is properly fitted, or in a defective state in which one or more places have a poor fitting. Therefore, the reference value of the sound pressure signal according to the number of overlaps when the fitting points overlap (this is a value obtained by adding an area equivalent value by the product of the sound pressure signal level and the generation period for each predetermined frequency band, A so-called sound volume equivalent value is obtained in advance, and a sound volume equivalent value for distinguishing between a good state and a bad state is set as a threshold value in the threshold table 10. In this example, as shown in FIG. 2, there are three fitting points, and therefore, a threshold value is set which can distinguish between a case where three places are fitted and a case where two places are fitted.
[0035]
In order to determine the area equivalent value based on the waveform shape of the time-series signal S2, the envelope or each peak value of the time-series signal S2 is detected, the sound pressure signal level is grasped, and this sound pressure signal level and time It is converted into an area equivalent value obtained by multiplying the generation period of the series signal S2. This processing is performed for each frequency band in which the fitting sound exists, and a total value is obtained.
[0036]
As an example of the determiner 9, as the area equivalent value of the time series signal S2 portion for each frequency band generated by the corrector 72, or when the fitting sound exists in a plurality of frequency bands, each time series signal S2 portion Fitting sound volume calculating means 91 for obtaining the sum of the area equivalent values, and comparing the value S3 (fitting sound volume value) with a predetermined threshold value set in the threshold value table 10 to determine the quality of the fitting sound. And a judgment processing means 92 for performing the judgment.
[0037]
Here, a description will be given of a procedure for determining the quality of fitting when a plurality of fitting sounds overlap. The vibration wave (sound pressure signal) captured by the microphone 2 includes, in addition to the fitting sound assumed in advance, a mechanical operation sound accompanying the movement of the assembling equipment, an ambient sound of the equipment, and the like. Therefore, if a vibration wave (sound pressure signal) is obtained by cutting a sound other than the fitting sound assumed in advance, the sound pressure level changes according to the number of overlapping fitting sounds. If the position of the microphone 2 and the environment and the like at each fitting location are the same, the sound pressure level increases in proportion to the number of overlapping fitting sounds.
[0038]
The total value of the area equivalent values obtained by multiplying the signal level of each time-series signal S1 and the generation period thereof is correlated with the total volume of fitting sounds or the number of overlapping fitting sounds.
[0039]
Therefore, by comparing the sum of the area equivalent values (fitting sound volume value S3) based on the waveform shape of each time-series signal S2 corrected by the corrector 72 with the threshold value, the fitting sound generated from the target object is compared. The number of overlaps, that is, the quality of fitting can be determined.
[0040]
The determination unit 9 determines that the connection is normal (OK, pass) when all the fittings are properly performed, and determines that the connection is abnormal (NG, rejection) when there is something that is not fitted. To display the judgment result. In the case of an abnormality (NG, reject), the alarm is also output to the alarm device 40 to generate an alarm.
[0041]
Next, the correcting means 20 for correcting the threshold value of the threshold value table 10 will be described.
[0042]
The NG data memory 22 and the OK data memory 24 are rewritable memories, and include NG data (fitting sound volume value) indicating an improper fitting state obtained in advance through experiments or the like, and OK data indicating a good fitting state ( The fitting sound volume value) is stored by a predetermined number N. The NG update range setting means 21 and the OK update range setting means 23 send the NG data memory 22 or the OK data memory 24 to the NG data memory 22 or the OK data memory 24 according to the quality judgment result of the judgment processing means 92 even during the actual judgment processing. After determining the update conditions (that is, whether or not to update and the update speed are set according to the purpose, for example, updating stored data one by one or updating a plurality of data simultaneously), the fitting sound is determined. The volume value S3 is updated and stored.
[0043]
The NG data statistical processing unit 25 performs a statistical process, for example, a normal distribution process on the N NG data stored in the NG data memory 22, and calculates an average value Xngi and a variance σngi. Similarly, the OK data statistical processing means 26 performs a statistical process, for example, a normal distribution process on the N pieces of OK data stored in the OK data memory 24, and calculates an average value Xoki and a variance value σoki.
[0044]
The threshold value calculating means 27 calculates a new threshold value based on the average values Xngi, Xoki and the variance values σngi, σoki obtained by the two statistical processing means 25, 26, and updates and stores the new threshold value in the threshold value table 10. Here, a sound pressure level between the NG distribution and the OK distribution and exceeding the NG distribution is obtained as a new threshold. As an example of calculation, first, the coefficient m of the variance that is farthest from the average value of both distributions is m = (Xoki−Xngi) / (σoki + σngi). The new threshold value L is obtained as L = Xoki−m · σoki or Xngi + m · σngi.
[0045]
In order to increase the accuracy of pass / fail judgment, when taking the safe side of 4σ, if the coefficient m is 4 or more, Xoki−4 · σoki, and if the coefficient m is less than 4, Xngi + 4 · σngi. It is desirable.
[0046]
Here, the procedure for setting the initial value of the threshold value set and stored in the threshold value table 10 used by the determination processing means 92 will be described. Normally, the threshold value is calculated based on distribution information of the sound pressure level by collecting a fitting sound generated at the time of performing an assembling operation of a component as an object in advance. At that time, the distribution information of the sound pressure level in both the good fitting state and the poor fitting state is necessary, but the occurrence frequency of the poor fitting state is lower than that in the good fitting state, so that a sufficient number of It is difficult to obtain information.
[0047]
Therefore, in this example, as shown in FIGS. 7 and 8, N or more OK sound pressure signals indicating a good state in which all three fitting points in the object 1 are properly fitted are collected in the OK database 320. By setting the virtual defective object 1 so that no sound is generated from one of the fitting positions due to the cutting of the protrusion or the overlaying, etc., the virtual NG indicating the poor fitting state is stored. N or more sound pressure signals are collected and stored in the NG database 310.
[0048]
Therefore, in the sound inspection apparatus 100A, the fitting sound is separated from the virtual NG sound pressure signal and the OK sound pressure signal collected in the NG database 310 and the OK database 320 in the same manner as the apparatus 100 shown in FIG. By extracting and statistically processing the sound pressure level, a virtual NG distribution and an OK distribution were obtained, and a threshold L0 was obtained from an average value and a variance value of both. The threshold value L0 is initially set as an initial threshold value in the threshold value table 10, the NG data memory 22 stores virtual NG data (fitting sound data) used for calculating the threshold value L0, and the OK data memory 24 stores OK. Data (fitting sound data) is stored and set for each N pieces.
[0049]
The input device 50A is a device for inputting, for example, the type of the resin molding die of the parts 110 and 120 shown in FIG. 2 as a selection command, and is used to add the type of data stored in both databases 310 and 320. Is done.
[0050]
Next, the determination flow of the vibration wave determination device 100 having the above configuration is summarized as shown in FIG. FIG. 5 is a signal waveform diagram of an oscillating wave, and FIG. 6 is an explanatory diagram showing a state of correction of a threshold.
[0051]
When the determination start is instructed to the apparatus 100, an initial threshold L0, N pieces of OK data and virtual NG data previously obtained by the sound inspection apparatus 100A are read from a storage device (not shown), and the threshold table 10, The storage is set in the OK data memory 24 and the NG data memory 22 (step 201). Needless to say, necessary data may be stored and set in the table 10 and the memories 24 and 22 in advance.
[0052]
When the preparation for determination is completed, first, a vibration wave generated from the object 1 is recorded by the microphone 2 to the storage device 5 as a digital sound pressure signal S0 (FIG. 5A) (Step 202). The wavelet transform calculator 6 separates and extracts the digital sound pressure signal S0 into a time-series signal S1 having a frequency band (FIG. 5B) that matches the fitting sound as the target operation sound (step 203). .
[0053]
In the example of FIG. 5, the sampling frequency of the digital sound pressure signal S0 is 44 KHz, the frequency band 11 is 11 to 22 KHz, the frequency band 10 is 5.5 to 11 KHz, and the frequency band 9 is 2.8 to 5.5 KHz. The sound is noticeable. Also, the machine operation noise appears in the frequency bands 8 to 6 below 2.8 KHz, and it can be seen that the data can be separated by distinguishing the data for each frequency band.
[0054]
Next, the filter processor 71 of the signal processing means 7 reduces or cuts the frequency band other than the fitting sound from these time-series signals S1, and extracts the fitting sound (step 204). Subsequently, the corrector 72 weights the time-series signal using the correction coefficient set for each predetermined frequency band stored in the correction table 8 to form a time-series signal S2, which corresponds to the fitting sound. The S / N ratio of the time series signal is improved (Step 205). FIG. 6C shows a time-series signal S2 obtained by filtering and correcting a signal in a frequency band other than the fitting sound.
[0055]
Next, the determiner 9 calculates an area equivalent value based on the corrected waveform shape of the time-series signal S2, and performs this processing for each frequency band in which the fitting sound exists, and calculates the total value (that is, the volume of the fitting sound). By calculating the total sound volume and the fitting sound volume value S3 which are values (step 206), and comparing the value S3 with a threshold value L (initial threshold value L0 initially) set in the threshold value table 10, the target object 1 is obtained. The number of generated fitting sounds is determined, and the quality of assembly (fitting) of the parts 110 and 120 is determined (step 207).
[0056]
If the fitting sound volume value S3 is equal to or greater than the threshold value L, a pass display (all pieces fit well) is displayed on the display 30 (step 208). Subsequently, the OK update range setting means 23 is set as an update condition for updating the data stored in the OK data memory 24 one by one and collecting the latest N data. Each time a good judgment result of the judgment processing means 92 is generated according to the update condition, the fitting sound volume value S3 at that time is updated and stored in the OK data memory 24 (step 209). Then, statistical processing of the latest N pieces of OK data is performed, and an OK distribution (average value Xoki, variance σoki) is calculated and stored as OK distribution information (steps 210 and 211).
[0057]
On the other hand, if the fitting sound volume value S3 is smaller than the threshold value L, a rejection display (improper fitting) is made on the display 30 and an alarm is given by the alarm 40 (step 212).
[0058]
Subsequently, the NG update range setting means 21 is set as an update condition for sequentially updating the data stored in the NG data memory 22 and collecting the latest N data. In particular, the NG data memory 22 initially stores N pieces of virtual NG data obtained by the sound inspection apparatus 100A, and the initial threshold value L0 is on the safer side (that is, a high sound pressure level that reliably excludes defective products). There is a lot of waste because the threshold value of ()) is set.
[0059]
Therefore, it is necessary to replace the actual NG data as soon as possible to correct the initial threshold value L0 and increase the accuracy of the threshold value L. In this example, until the number K of NG data (fitting sound volume value S3) obtained by the determination processing means 92 reaches N / 2, the number of virtual NG data twice as large as the obtained NG data is replaced. When the number K of NG data becomes N / 2 or more, the same number of data is replaced, and the correction speed of the threshold L is increased (steps 213 to 216). It is also possible to increase the data replacement speed three times or more for the early validation of the threshold value.
[0060]
In accordance with the update condition, the fitting sound volume value S3 at that time is updated and stored in the NG data memory 22 every time a failure determination result of the determination processing means 92 occurs. The NG data statistical processing means 25 performs statistical processing of the latest N pieces of NG data, and calculates an NG distribution (average value Xngi, variance σngi) as NG distribution information (step 217). When the calculated average value Xngi is greater than or equal to the previous value Xngi−1, that is, on the safe side in the direction in which the new threshold value L increases, the calculated NG distribution (average value Xngi, variance σngi) is used to calculate the threshold value. (Steps 218, 219).
[0061]
On the other hand, when the calculated average value Xngi is smaller than the previous value Xngi−1, that is, when the new threshold value L is on the danger side, the obtained NG distribution (average value Xngi, variance σngi) is stored. However, the previous values (average value Xngi-1 and variance value σngi-1) remain (steps 218, 220, 221). However, after the number of NG data K has become sufficiently large (K ≧ K1), it can be determined that the accuracy of the NG data is high, so that a correction to reduce the threshold L is permitted (steps 218, 220, 219).
[0062]
Next, the threshold value calculating means 27 calculates both distributions based on the OK distribution information (average value Xoki, variance value σoki) and the NG distribution information (average value Xngi, variance value σngi) obtained in steps 211, 219, and 221. The variance coefficient m that is the furthest from the average value is m = (Xoki−Xngi) / (σoki + σngi). A new threshold value L is obtained as L = Xoki−m · σoki or Xngi + m · σngi (step 222). The new threshold L is updated and stored in the threshold table 10 (step 223).
[0063]
Therefore, the threshold value L set and stored in the threshold value table 10 is corrected from the initial threshold value L0 each time the determiner 9 detects the fitting sound volume value S3. The upward correction for increasing the threshold value L is immediately performed, but when the NG distribution changes in a lowering direction as shown in FIG. 6, when the NG data number (sample number) K is small, the correction is not performed and the initial threshold value L0 is not changed. When the number K of NG data becomes equal to or larger than the predetermined number K1, the correction of the threshold L is started downward, thereby suppressing the easy correction and ensuring the accuracy of determining the quality of the fitting sound.
[0064]
(Other embodiments)
In the present example, when determining whether or not a plurality of types of objects 1 are assembled or fitted, the correction coefficient or the threshold value is changed according to the types of the objects 1 so that the plurality of types of the objects 1 can be subjected to the determination processing, or a continuous process can be performed. 1 is an example of a determination apparatus 100 that enables determination processing. FIG. 9 shows a system configuration of another embodiment. FIG. 9 differs from FIG. 1 in that an input device 50 for selecting and instructing the type of the object 1 to be assembled and fitted is provided. The difference is that the threshold value is prepared and can be selected by the input device 50, and that a plurality of types of threshold values are prepared as the threshold value table 10 and can be selected by the input device 50.
[0065]
The input device 50 is a device for inputting, for example, the type of the resin molding die of the parts 110 and 120 shown in FIG. 2 as a selection command, and depends on the inner and outer diameters, roundness, dimensions, shape, material, and the like of both parts 110 and 120. It is divided into several types. This is because it has been found that the dispersion of the two components 110 and 120 mainly has a different distribution for each resin mold.
[0066]
In the correction table 8, as shown in FIG. 10A, individual correction tables (Da1, Da2...) 82 are set only for the combination types of the resin molding dies used for the determination of the vibration wave. (For example, Da1) constitutes the individual correction table shown in FIG. In FIG. 10A, A, B, C, and D indicate types of the component 110, and 1, 2, 3, and 4 indicate types of the component 120. Therefore, the table selecting means 81 reads the correction coefficients a11, b11,... Set for each predetermined frequency band from the individual correction table 82 selected according to the selection command signal of the input device 50, and Are provided to the correctors 72a, 72b,...
[0067]
As shown in FIG. 10C, the threshold value L (La1, La2,...) Is set for the combination type of the resin molding dies in the threshold value table 10 as in the correction table 8. In the drawing, A, B, C, and D, 1, 2, 3, and 4 indicate the types of the components 110 and 120 as in FIG.
[0068]
Note that NG data and OK data are stored in the NG data memory 22 and the OK data memory 24 of the correcting means 20 for each type of the object 1 to be determined. This is the same as the embodiment.
[0069]
The operation of the above-described device is the same as the operation of the embodiment shown in FIG.
[Brief description of the drawings]
FIG. 1 is a configuration diagram illustrating a system configuration according to an embodiment of the present invention.
FIG. 2 is an explanatory view showing a part of a process of assembling the parts shown in FIG. 1;
FIG. 3A is an explanatory view showing a main part of an assembling process of the parts shown in FIG. 2; (B), (c) is a figure which shows the sound pressure waveform detected in the assembly | attachment process of components.
FIG. 4 is a flowchart showing a processing flow of the vibration wave determination device 100 shown in FIG.
FIG. 5 is a signal waveform diagram of the vibration wave determination device 100 shown in FIG.
FIG. 6 is an explanatory diagram showing a state of correction of a threshold value L of the vibration wave determination device 100 shown in FIG.
FIG. 7 is an explanatory diagram for obtaining an initial threshold value L0 by the sound inspection device.
8 is an explanatory diagram showing an outline of calculation of an initial threshold value L0 by the sound inspection device shown in FIG. 7;
FIG. 9 is a configuration diagram showing a system configuration of another embodiment of the present invention.
10A and 10B are diagrams showing the contents of an individual correction table 82, and FIG. 10C is a diagram showing the contents of a threshold value table 10.
[Explanation of symbols]
1 Object
2 microphone (vibration wave input means)
6 Wavelet transform calculator (frequency separation means)
7 Signal processing means
71 Filter processor (filter processing means)
72 Corrector (correction means)
8 Correction table (correction table means)
9 Judgment device (judgment means)
91 Fitting sound volume calculating means (volume calculating means)
92 Judgment processing means
10. Threshold table (threshold table means)
20 Correction means
22 NG data memory
24 OK data memory
25 NG data statistical processing means (first statistical processing means)
26 OK data statistical processing means (second statistical processing means)
27 Threshold calculation means
50 input device
100A sound inspection device

Claims (12)

Vibration wave input means for detecting and inputting a vibration wave generated when the object is operated;
Frequency separation means for frequency-separating the input vibration wave to form a first time-series signal for each predetermined frequency band;
Signal processing means for extracting, from the first time-series signal, a second time-series signal for each predetermined frequency band in which a target operation sound indicating a predetermined operation state of the object is generated;
Determining means for determining the state of the target object based on sound pressure information and a threshold value of the target operation sound obtained from the second time-series signal,
And a correcting means for correcting the threshold value according to a result of the determination by the determining means. The apparatus according to claim 1, further comprising a rewritable threshold table storing the threshold used in the determination. Sound inspection means for obtaining a defect distribution of the sound pressure level of the object based on the vibration wave obtained by temporarily setting the defect state of the object,
2. The vibration wave determination device according to claim 1, wherein a sound pressure level exceeding a range of the failure distribution obtained by the sound inspection means is set as the initial threshold as the threshold used by the determination means. 3. The correction means collects sound pressure information of the target operation sound obtained from the second time-series signal at that time for each of the pass / fail judgment results obtained by the judgment means, performs statistical processing, and performs an OK distribution of the obtained good judgment results. The vibration wave determination device according to claim 1, wherein the threshold used in the determination unit is corrected based on information and NG distribution information of a result of the determination. The correction means includes an OK data memory for storing the sound pressure information when the object is in a good state, and the sound pressure information when the object is in a bad state, in accordance with a good or bad judgment result of the judgment means. NG data memory, first statistical processing means for statistically processing the OK data stored in the OK data memory to obtain an OK distribution, and statistical processing of the NG data stored in the NG data memory to obtain an NG distribution And a threshold calculating unit that obtains a new threshold based on the result of the statistical processing of the first and second statistical processing units and corrects the threshold used by the determining unit. The vibration wave determination device according to claim 1, wherein: The signal processing unit includes a correction unit that corrects the level of the first time-series signal to form the second time-series signal, and a correction table that stores a correction coefficient used by the correction unit. The vibration wave determination device according to claim 1, wherein: 3. An input device which is selectable from outside is provided, and said threshold value table means selects said threshold value used by said determination means according to a selection command signal instructed by said input device. The vibration wave determination device as described in the above. 7. An input device which can be selected from outside, wherein the correction table means selects the correction coefficient used by the correction means in accordance with a selection command signal instructed by the input device. 6. The vibration wave determination device according to 1. 9. The vibration wave determination device according to claim 6, wherein the signal processing unit includes a filter processing unit that supplies the time series signal obtained by removing noise of the first time series signal to the correction unit. The determining means includes a sound volume calculating means for obtaining the sound pressure information by adding an area equivalent value based on a waveform shape of the second time series signal for each predetermined frequency band, and a sound of the sound pressure information. The vibration wave determination device according to claim 1, wherein the number of target operation sounds generated from the object is determined based on the pressure level. The said volume calculation means calculates | requires the said area equivalent value based on the waveform or shape of the said 2nd time series signal according to the envelope or peak value of the said 2nd time series signal, The claim of Claim 10 characterized by the above-mentioned. Vibration wave determination device. Sound inspection means for obtaining a defect distribution of the sound pressure level of the object based on the vibration wave obtained by temporarily setting the defect state of the object,
The OK data memory and the NG data memory initially store initial data obtained by the sound inspection means, and thereafter store the latest sound pressure information of the target operation sound obtained from the second time-series signal. 6. The method according to claim 5, wherein the first and second statistical processing means and the threshold value calculating means obtain an initial threshold value based on the initial data and give the initial threshold value as the threshold value of the determining means. 6. The vibration wave determination device according to 4.

Images