JPH03211457A - Crack detecting method - Google Patents

Abstract

PURPOSE:To improve the accuracy of a conforming article decision area and to improve the productivity by updating the conforming article decision area automatically according to the natural vibration frequency and vibration peak value of an inspected article which is decided as a conforming article by inspection. CONSTITUTION:A decision device 8 repeats a process for finding a natural vibration point from the primary - tertiary natural vibration frequencies and output level of a vibration waveform from a frequency analyzer 7 as to a specific number of conforming articles of the same kind in the same shape. Then the point of the center of gravity is found from a distribution on a secondary plane of natural vibration points and the directions of the major and mirror axes of an ellipse are determined from the primary regression expression of a distribution passing the center of gravity to find standard deviations in the direction. An ellipse having a major and a minor axis diameter obtained by multiplying the deviations by a specific coefficient by the directions is set as the conforming article decision area and when an inspected article is in the area, it is decided that the article is normal. Then the natural vibration frequency and vibration peak value of the conforming article are inputted again to update the decision area. Consequently, the accuracy of the conforming article decision area is improved and the productivity is improved.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a method of inspecting a sintered part to be inspected by non-destructive testing, and in particular to a method for inspecting a sintered part to be inspected, and in particular to a method for inspecting a sintered part by non-destructive testing. The present invention relates to a crack detection method using an impact vibration analysis method that determines the quality of the inspected product by analyzing vibration.

[Conventional technology]

Conventionally, one of the nondestructive testing methods for determining the quality of metal members is the so-called impact vibration analysis method, which detects cracks in the inspected member by applying a blow to the metal member and detecting the vibration sound.

As a conventional example of this impact vibration analysis method, there is a method of determining cracks based on the natural frequency of the natural vibration caused by impact, as disclosed in, for example, Japanese Patent Laid-Open No. 48-30983.

In addition, for example, as disclosed in "Evaluation of Mechanical Properties and Cracks of Metal Materials by Impact Vibration Analysis Method; Japan Nondestructive Inspection Association Third Subcommittee Material No. 3844," from the natural frequency and damping coefficient of impact vibration, There are also methods to predict the occurrence of cracks.

This impact vibration analysis method processes the detection signal from the sensor that detects the vibration generated by impact, obtains the auto power spectrum and the envelope of the waveform, calculates the natural frequency and amplitude damping coefficient, and The occurrence of cracks is determined based on the magnitude of the natural frequency and amplitude damping coefficient.

This is the Japan Nondestructive Testing Association Third Subcommittee Material No. 3.
As disclosed in No. 844, the damping coefficient generally increases when a crack occurs, and the natural frequency decreases when a crack occurs.

However, some defective products with cracks have damping coefficients that are smaller than those of good products without cracks, and some have the same natural frequency as good products without cracks. There has been a problem in that the discrimination accuracy is not sufficient based only on the magnitude of the natural frequency or the damping coefficient.

For this reason, the applicant first introduced 50 non-defective products and determined their natural frequency (
f i ; s = 1, 2. shown in FIG. ..., 50)
and the natural vibration peak value, which is the peak value of the vibration level at that natural frequency (Pi shown in Figure 3; i = 1.2.
..., 50) as an ellipse on a two-dimensional plane with the frequency and vibration output level as the balance (see Figure 4)
A method of determining pass/fail based on whether or not the natural vibration peak falls within the ellipse.
and the natural vibration peak value Pi, which is the output level at that natural frequency, is memorized and plotted on a two-dimensional plane with the frequency and vibration output level as parameters. The center of gravity point (point G shown in Figure 4) is determined from the distribution, and the major axis (X direction shown in Figure 4) and short axis direction (the direction shown in Figure 4 Y direction), find the standard deviation (sigma) in each direction, and create an ellipse on a two-dimensional plane with the major axis and minor axis calculated by multiplying the sigma in each direction by 3 for each of the corresponding type of inspected product. The natural frequency ft and natural vibration peak value Pi of each inspected product are set as the non-defective judgment area, and the natural frequency fi and natural vibration peak value Pi are set as the preset frequency and vibration. It is determined whether or not it is within a non-defective judgment area using the output level as a parameter, and if it is determined to be a non-defective product, the natural frequency fi and natural vibration peak value Pi of the non-defective product are stored, and this natural frequency fi and natural vibration are From the natural frequency fi and natural vibration peak value Pi of the latest predetermined number of good products including the peak value Pi, the center of gravity G and major axis of the area are newly calculated.
We have proposed a method in which the short axis is determined, and then a new non-defective determination area for the corresponding type of inspected product is set at the center of gravity G and the long and short axes (Japanese Patent Application No. 1-192389).

[Problem to be solved by the invention]

However, in the above application, when the inspected product is determined to be a non-defective product by discrimination based on the previously set non-defective product discrimination area, the natural vibration point (f
i, Pi), the center point G of the ellipse that newly indicates the outline of the non-defective discrimination area, and the major axis and minor axis of this ellipse were newly calculated each time. The calculation of the regression line and the coordinate transformation using the long axis of the ellipse as the X axis take time, and if the intervals at which the items to be inspected come in are short (for example, in the order of several seconds), the inspection may not be completed in time. However, there was a problem in that the inspection process required a large amount of stock, which became a bottleneck in improving productivity.

Therefore, the applicant focused on the fact that there is a close correlation between the density and natural frequency of the inspected product, and investigated the relationship between the two for each non-defective product and defective product, and as a result, the results shown in Figure 5 are as follows. As shown, it has been found that as the density of the inspected products increases for both non-defective and defective products, the natural frequency fi also increases with an extremely strong correlation. Further, although not shown, it was found that the natural vibration peak value Pi at the natural frequency ft also increases with the density in the same way as the natural frequency fi. From these facts, it can be said that long-term fluctuations in density, shape, etc. have little effect on short-term fluctuations. In other words, it is not necessary to recalculate the center of gravity, major axis direction, and major and minor axes of the initially determined pass/fail determination area every time the product to be inspected is determined to be non-defective. We discovered that it is sufficient to update the data by including the data of the identified inspected products.

Therefore, in the present invention, the center of gravity of the non-defective product determination area is newly determined by adding the data of newly introduced non-defective products to the existing data, and a new non-defective product determination area is set using this center of gravity as the new center of gravity. The purpose is to make it possible to determine whether a good product or a defective product is accurate and in a short calculation time.

[Means to solve the problem]

Therefore, the present invention determines whether the natural frequency of the sound or vibration generated by hitting the product to be inspected and the natural vibration peak value, which is the output level at that natural frequency, are within a preset good product judgment range. In the crack detection method that detects cracks, the natural frequency and natural vibration peak value of a plurality of good products of the same type are memorized, and this frequency and output level are used as parameters. Plot it on a dimensional plane, find the centroid point from the distribution on the two-dimensional plane of the natural vibration points represented by these natural frequencies and natural vibration peak values, and then calculate the ellipse from the linear regression of the distribution passing through the centroid point. Determine the major and minor axis directions of Set a non-defective judgment area for each type of inspected product, measure the natural frequency and natural vibration peak value for each inspected product, and find the natural vibration point represented by these natural frequencies and natural vibration peak values. It is determined whether or not it is within a non-defective product judgment area using preset vibration frequency and output level as parameters, and if the product is determined to be non-defective, the natural frequency and natural vibration peak value of the non-defective product are stored, and the non-defective product is A new center of gravity is determined from the natural frequency and natural vibration peak value of , and the previously stored data of the center of gravity of the non-defective product determination area, the value is updated, and this updated center of gravity is used as the new center of gravity. It is characterized by setting a non-defective product discrimination area.

[Effect]

According to the configuration of the present invention, the non-defective product determination area in the crack detection device that detects cracks by hitting the product to be inspected is
The natural frequencies and natural vibration peak values of multiple non-defective products of the same type and shape are detected, and the detected natural frequencies and natural vibration peak values are displayed on a two-dimensional plane with the frequency and output level as parameters. The centroid point is determined from the distribution of natural vibration points on the two-dimensional plane expressed by the natural frequency and natural vibration peak value, and the long axis of the ellipse and Determine the minor axis direction and find the standard deviation (=sigma) in each direction. Then, the values obtained by multiplying each sigma by a predetermined coefficient for each direction are the major and minor axes of the ellipse, and this ellipse is 2
It is set as a non-defective judgment area of the inspected product on the dimensional plane.

Then, it is determined whether the natural frequency and natural vibration peak value of the product to be inspected are within the above-mentioned non-defective product determination area, and if the product is determined to be non-defective, the natural frequency and natural vibration peak value of this non-defective product are Newly store the natural frequency and natural vibration peak values of a good product of the same type as the non-defective product, and calculate the natural frequency and natural vibration peak value of this non-defective product with the data of a new predetermined number of products first. Then, a new non-defective product determination area is set.

In this way, based on the natural frequency and natural vibration peak value of a predetermined number of non-defective products, a non-defective product judgment area is set on the two-dimensional child surface with the natural frequency and natural vibration peak value as parameters, and Good or bad products are determined based on whether the natural frequency and natural vibration peak value are within the above-mentioned non-defective determination area. A new non-defective judgment area is calculated from the natural frequency and natural vibration peak value of the non-defective product, and this non-defective judgment area is updated sequentially, so it is possible to eliminate variations between manufacturing lofts or seasonal fluctuations in the same type of inspected product. The non-defective product determination area is updated accordingly.

Therefore, during inspection, the quality of the product to be inspected is determined based on the non-defective product determination area that has been updated in accordance with variations between manufacturing lofts or seasonal variations.

〔Example〕

A crack detection method according to an embodiment of the present invention will be explained in detail with reference to FIGS. 1 to 8.

Fig. 1 is a block diagram of an inspection device to which this crack detection method is applied, Fig. 2 is a distribution diagram showing the distribution state of natural vibration peak values, and Fig. 3 is a waveform diagram showing the vibration waveform of impact vibration of the inspected product. , Fig. 4 is a frequency spectrum diagram of the vibration waveform, Fig. 5 is a correlation diagram showing the relationship between the density of the inspected product and the natural frequency, and Fig. 6 is the setting of the non-defective judgment area and the non-defective/defective product in this crack inspection method. FIG. 7 is a plan view of the automatic inspection line viewed from above, and FIG. 8 is a side view of the automatic inspection line of FIG. 7 viewed from the side.

In Fig. 7.8, reference numeral 1 is an item to be inspected, and reference numeral 2 is a cylinder, and the item to be inspected 1 is supported by this cylinder 2 during inspection. Reference numeral 3 denotes a steel ball for striking the product to be inspected 1, which is suspended from a predetermined height by a piano wire of a predetermined length. Reference numeral 4 is an electromagnetic pink that detects vibrations based on changes in the distance to the inspected item 1, and detects impact vibrations when the steel ball 3 hits. Reference numeral 10 denotes an inspection device that processes the detected signal and discriminates between good and defective products, and as shown in FIG. 1, it includes an amplifier 5, a delay device 6, a frequency analyzer 7, and a discrimination device 8.
It consists of Reference numeral 9 denotes an operation panel connected to the inspection device 5, which is used to switch between an inspection mode for performing normal inspections and a registration mode for setting non-defective determination areas, and to set the type of the product 1 to be inspected.

Reference numeral 11 denotes a transport conveyor that transports the inspected item 1 to the automatic inspection line; numeral 12 represents a busher that sends the inspected item 1 transported by the conveyor 11 to the inspection station 13; numeral 14 indicates a conveyor from the inspection station 13; A jig 14 is used to temporarily hold the inspected item 1 that has been carried out.
, reference numeral 15 is a carry-out jig for carrying out the inspected product 1 after inspection;
Reference numeral 16 is an NG chute for discharging defective products, and when opened, the defective products are dropped. Reference numeral 17 is a carry-out conveyor for carrying out non-defective products.

As a result of the above-described configuration, when the electromagnetic pink-up 4 detects the impact vibration during impact with the steel ball 3, the output signal is amplified by the amplifier 5 and input to the inspection device 10. Then, the input signal, that is, the vibration waveform from the point of impact, is taken into the frequency analyzer 7 in accordance with the timing signal output from the delay device 6.

Here, the timing signal will be explained with reference to FIG. 2. This FIG. 2 shows the vibration waveform from the amplifier 5. Waveform A represents the impact noise at the time of impact, and this impact noise A includes a wide range of frequency components similar to white noise.
The amplitude also fluctuates significantly. Waveform B is a vibration waveform of the inspected product 1 after the impact noise A has been attenuated, and shows a resonance phenomenon due to the natural frequency, and has a curve like waveform C shown by a broken line.

Since the impact noise A is a vibration unrelated to the natural vibration of the product to be inspected 1, if this waveform is taken in and subjected to frequency analysis, misjudgment will occur.

Therefore, in order to eliminate the influence of this impact noise A, a timing signal is generated that instructs sampling for a time t1, delayed by t from the moment of impact, and t2 from the moment of impact.
After the elapse of time, a timing signal instructing the second sampling is generated again for the time t1.

In addition, in order to eliminate the influence of resonance, the sampling time t3 is set to be longer than the period of the undulation of the waveform C, and (tt
tl)/2.

By setting the sampling timing 12.1° and the sampling time t in this way, stable and highly accurate discrimination can be achieved.

The frequency analyzer 7 in FIG. 1 analyzes and outputs the frequency component of the vibration waveform detected by the electromagnetic pick 4 and its output level.

The frequency analyzer 7 is capable of detecting the output level of each frequency, as shown in the frequency spectrum diagram of FIG. 3 showing the frequency components of the vibration waveform and their output levels. In this embodiment, the peak value P1 representing the natural frequency of the first to third order natural vibrations shown in FIG. P2. P3 can be detected. High-order natural vibrations of the fourth order or higher have small peak values and lower detection sensitivity, so they are not used in this embodiment.

The discrimination device 8 in FIG. ~Natural frequency of third-order natural vibration and its output level PL P2. From P3, the process of determining the natural vibration point by the peak value detection subroutine (not shown) is performed a preset number of times, that is, a predetermined number of identical types and
Repeated for good products with shape. In the case of first-order natural vibration, this natural vibration point means the natural frequency r.

, when the natural vibration peak value, which is the output level, is Pl, it is expressed as (r+, P,).

The natural vibration point (f,,P

). As shown in Fig. 4, these natural frequencies and natural vibration peak values are plotted on a two-dimensional plane with the frequency and output level as parameters, that is, a two-dimensional plane with the frequency as the x-axis and the output level as the y-axis. Plot the natural vibration peak value on a dimensional plane, find the center of gravity position from its distribution, and calculate the long axis (X axis) of the ellipse from the linear regression equation of the distribution passing through the center of gravity position.
and short axis (y-axis) direction, find the standard deviation (= sigma) in each direction, and multiply the value by the respective predetermined coefficient (
For example, the major axis (α) and minor axis (β
), the closed region surrounded by this ellipse is set as a non-defective determination region for determining non-defective products, and this is set and registered for each of the 1st to 3rd order natural vibrations.

If the above-mentioned procedure is performed individually for a plurality of types of inspected articles l having different types and shapes before the inspection, it is possible to set and register the non-defective product determination areas for the plurality of types of inspected articles l in advance.

Further, when the inspection mode is selected on the operation panel 9, the discrimination device 8 determines the natural frequencies of the first to third natural vibrations of the vibration waveform output from the frequency analyzer 7 and the output level P.
i, P2. It is determined for each 1st to 3rd natural vibration whether or not P3 exists within a non-defective judgment area on a two-dimensional plane with the frequency and output level as parameters, and the 1st to 3rd natural frequency and A product is determined to be non-defective only when all of its natural vibration peak values are within the non-defective product determination area.

The processing of the discrimination device 8 in the inspection device 10 described above will be explained based on the flowchart of FIG. 6.

In step 100, it is determined whether the inspected item 1 is set in a predetermined position. If not, the process does not proceed to the next step, and only if it is set, the process proceeds to step 101 and subsequent steps.

In step 101, the type of the inspected product 1 to be inspected is
In step 102, either the inspection mode or the initial setting mode is input from the operation panel 9.

In step 103, it is determined whether the mode is set to inspection mode or initial setting mode. If the initial setting mode is input in step 102, the process proceeds to step 104 and waits for SRQ (wait for output of frequency analyzer 7). ), and when the waiting state is released, the output from the frequency analyzer 7 is input to the discriminator 8 in step 105.

Then, in step 106, the data (fi, Pi) of the 1st to 3rd order natural vibration points are determined by the peak value detection subroutine.
(i=1 to 3) is obtained.

In step 107, it is determined whether or not the processing from steps 100 to 106 has been performed 50 times. If it is less than 50 times, the process proceeds to step 108, where the counter is incremented by "1", and the process returns to step 100. If it is determined that 50 non-defective products have been measured, step 109 to step 112 is performed.
A non-defective product determination area is set from the distribution of natural vibration points (fi, Pi) determined in step 6.

That is, in step 109, the average value f□(-(Σfi)150) of the natural frequencies fi of the 50 ratio-inspected products is determined.
and similarly the average value psv(-
(ΣPi)150) is the center of gravity of the distribution G (r
ay, Pav) are calculated.

Then, in step 110, the long axis direction and short axis direction of the ellipse serving as the association line of the discriminant region are determined from the linear regression equation of the distribution passing through the centroid point G, and in step Ill, the standard deviation σ8 of the distribution in each direction is determined. ．． Find σ7. In step 112, the standard deviation σ8 in the long axis direction and the short axis direction. Multiply the value from σ7 by a predetermined coefficient to obtain the major axis (H□=nXσX) and the minor axis (H□=nxσ).

).

As described above, a non-defective product determination area whose boundary line is an ellipse defined by the center of gravity G and major axis HIX'HIIY determined in steps 109 to 112 is set.

Next, when the inspection mode is input in step 102, the process proceeds from step 103 to step 120.

Note that steps 120 to 122 are the same as steps 104 to 122.
Since the same processing as in 106 is performed, the explanation thereof will be omitted.

In step 123, it is determined whether or not the first-order natural vibration point (f+, P,) obtained in step 122 is within the non-defective determination area, and if it is within the area, the process proceeds to the next step 124.

Similarly, in steps 124 and 125, it is determined whether the secondary and tertiary natural vibration points (fi, Pf) are within the non-defective determination area, and if both are within the area, proceed to step 126 and OK (pass). After the signal is output, step 12
Proceed to step 8.

In step 128, the natural frequency fi and natural vibration peak value Pi for each of the first to third natural vibrations are newly stored, and the added natural frequency fi and natural vibration peak value Pi are calculated first. The coordinates of the center of gravity G are newly determined from the accumulated value Σfi of the natural frequencies fi of a plurality of non-defective products and the accumulated value ΣPi of the natural vibration peak values Pi.

That is, as shown in step 128 of the drawing, the newly added natural frequency fi and natural vibration peak value Pi
are the integrated value Σfi of the natural frequencies fi and the integrated value ΣPi of the natural vibration peak values Pi of the plurality of non-defective products found previously.
The number N of non-defective data is incremented by 1 (N-N+1), and the division value (Σft
+f, 1. .

)/(N+1) and (ΣPf +PH+1)/
Let (N+1) be the coordinates of the new center of gravity G. Then, a new non-defective product determination area whose boundary line is an ellipse centered on the newly determined center of gravity G is set, and this step is completed.

When this step 12B is completed, the process proceeds to step 129, increments the memory number N of good product data by 1, and returns to step 100.

In addition, if any one of steps 123 to 125 is determined to be outside the range, the result is NG (fail) in step 127.
After the signal is output, the process returns to step 100 and this process is temporarily terminated.

Next, the inspection method of the crack detection device will be explained along the flow of an actual line with reference to FIG. 7.8.

In FIGS. 7 and 8, the product to be inspected 1 is conveyor 1
1, and when it is conveyed to a predetermined position, it is moved from the conveyor container 11 to the inspection station 13 by the busher 12 and set at a predetermined position.

The set item 1 to be inspected is raised to an inspection position by a cylinder 2 and is struck by a steel ball 3. The electromagnetic pick 4 detects the vibration at this time, and the detected detection signal is input to the inspection device 10.

After the inspection of the impact and vibration detection has been completed, the inspected article 1 is carried out to the jig 14, and the receiver waits in the jig 14 until the determination result is obtained by the inspection device FIO.

Then, when the determination result is obtained, the inspected item 1 is transferred to the unloading jig 15.
However, if the inspection result is NG (rejected), the defective product chute 16 is opened and the defective product chute 16 is opened and the product is placed on a defective product pallet (not shown). be done.

However, in the case of the inspected article 1 that is found to be a good product, that is, passed, the defective product shoe 16 is not opened, and the inspected article 1 is moved as it is to the carry-out conveyor 17, and is carried out by the carry-out conveyor F.

In addition, when the setting mode is selected on the operation panel 9, a preset number of good products 1 having the same type and shape are sent to the conveyor 11.
Inspection data of non-defective products is detected at the inspection station 13 and input to the inspection device 10. Then, based on the detected inspection data, a non-defective determination area is set and registered for determining pass/fail during the inspection, and when the next inspection mode is selected, the inspection is performed based on this non-defective determination area. If the inspected item 1 is determined to be a non-defective item through this inspection, the natural frequency and natural vibration peak value of the inspected item 1 are newly calculated from the center of gravity based on the previously stored integrated values. G
Since it is determined and updated, it is possible to correctly set the non-defective judgment area even for variations in the natural frequency between lofts or each season of the inspected product 1. Since the calculations of the long axis and short axis, the standard deviation, and the coordinates of the center of gravity G are not recalculated from the beginning, the process of updating the non-defective discrimination area can be shortened, which greatly contributes to improving productivity.

Note that in this embodiment, the non-defective product determination area is reviewed and updated for each inspection, but
The non-defective product determination area may be reviewed and updated every arbitrary number of times, such as once.

Furthermore, in order to find the coordinates of the new centroid point G based on the data of the natural vibration point of a new non-defective product, we use a method of sequentially adding additional data to the previous integrated value, but this takes some calculation time. However, it is also possible to sequentially erase old data by first-in/first-out data and always find a new center of gravity based on the latest 50 pieces of data.

Although specific embodiments of the present invention have been described above, the present invention is not limited to these embodiments, and includes various embodiments within the scope of the claims, such as , detection means! Instead of a magnetic vibrator, the sound at the time of impact may be detected with a microphone, or the vibration at the time of impact can be directly detected by placing a vibrator directly on the item to be inspected or on a support member that supports the item to be inspected. A similar effect can be obtained by doing so.

〔Effect of the invention〕

As explained above, according to the present invention, when a non-defective product is determined in the inspection, the non-defective product judgment area newly takes in the natural frequency and natural vibration peak value of the non-defective product online, and stores it in advance. Since the center of gravity G is newly determined and updated together with the current data, the good product judgment area is automatically updated even when the natural frequency of the inspected product changes due to changes in each manufacturing loft or seasonal changes. It will be done.

In addition, among the center of gravity, major axis direction, minor axis direction, standard deviation, and major axis and minor axis determined from this standard deviation, which determine the ellipse that indicates the boundary line of the non-defective product discrimination area, the influence of changes in the properties of the inspected product is considered. Since only the coordinates of the centroid point that is most sensitive are updated, the accuracy of the non-defective discrimination area is ensured, and the calculations for updating the non-defective discrimination area can be performed in a short time, greatly improving productivity. It has a great effect.

[Brief explanation of drawings]

1 to 8 are diagrams for explaining a crack detection method according to an embodiment of the present invention, FIG. 1 is a block diagram of an inspection device to which the present crack detection method is applied, and FIG. A distribution diagram showing the distribution state of natural vibration points, Figure 3 is a waveform diagram showing the vibration waveform of impact vibration of the inspected item, Figure 4 is a frequency spectrum diagram of the vibration waveform, and Figure 5 is a diagram showing the density and density of the inspected item. A correlation diagram showing the relationship with the natural frequency. Figure 6 is a flowchart showing the setting of the non-defective judgment area and the process of determining good/defective products in this crack inspection method. Figure 7 is a plane view of the automatic inspection line from above. 8 is a side view of the automatic inspection line of FIG. 7, seen from the side. 1...Product to be inspected

Claims (1)

[Claims] (1) Cracks are detected depending on whether the natural frequency of the sound or vibration generated by hitting the inspected product and the natural vibration peak value, which is the output level at that natural frequency, are within a preset non-defective judgment area. In the crack detection method that performs crack detection, the natural frequency and natural vibration peak value of a plurality of non-defective products to be inspected for each type are memorized, and the frequency and output level are used as parameters on a two-dimensional plane. The centroid point is determined from the distribution of natural vibration points on a two-dimensional plane expressed by these natural frequencies and natural vibration peak values, and the major axis of the ellipse and Determine the minor axis direction, find the standard deviation (sigma) in each direction, and create an ellipse on a two-dimensional plane with major and minor axes equal to the value multiplied by a predetermined coefficient of sigma for each direction. Set a non-defective judgment area for each product, measure the natural frequency and natural vibration peak value for each inspected product, and set the natural vibration point represented by these natural frequencies and natural vibration peak values in advance. If the product is determined to be a good product, the natural frequency and natural vibration peak value of the non-defective product are stored, and the natural frequency of this non-defective product is stored. Then, a new center of gravity is determined from the natural vibration peak value and the previously stored data of the center of gravity of the non-defective product determination area, and its value is updated, and this updated center of gravity is used as the new center of gravity for the non-defective product determination area. A crack detection method characterized by setting.