JPH03191859A - Crack detecting apparatus - Google Patents

Abstract

PURPOSE:To judge the occurring state of cracks in a work at the time of machining by dividing the waveform data which are detected with an acoustic emission(AE) detecting means into specified periods, and detecting the periodicity for every period. CONSTITUTION:The output of an AE sensor 1 is amplified in a pre-amplifier 19 and a main amplifier 21. The noise components at low levels are removed through a filter 23. Then the AE waves are sampled in a sample/hold circuit 25 at a sampling period which is preset at a setting part 29. The analog output which is held with the circuit 25 is converted into a digital value. The value is stored in a memory 31 and accumulated as a discrete waveform data. Then, a periodicity detecting part 15 reads the specified sampling times in the sequence of time series out of the waveform data in the memory 31. The operation of the sum of products is performed with a delay circuit (b) and a sum-of-products circuit (c) in the part 15 for every predetermined number of sampling. In a judging part 37, the autocorrelation function value present in the specified analyzing section which is inputted from the detecting part 15 is compared with a reference value set at a reference setting part 41, and the periodicity is judged.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a rack detection device that is capable of monitoring cracks occurring during processing of a workpiece in real time.

[Prior Art] Magnetic flaw detection is used to detect quenching cracks that occur during induction hardening of steel members. Judgment must be made by immersion in a liquid and visual observation of the fluorescent part, making it difficult to automate the inspection. Therefore, it has been impossible to incorporate the hardening process into an automatic line that requires automatic and real-time determination of defective products as long as the magnetic flaw detection method is employed.

As a technology that enables inspection automation of this induction hardening,
A device using acoustic emission (hereinafter referred to as AEJ) has been proposed (Japanese Patent Laid-Open No. 60-1
35861). This device attempts to automate the determination of the state of cracks by capturing the AE waves generated when cracks occur through filter processing and threshold value determination, and calculating the number of AE waves generated and the average amplitude value.

[Problem to be solved by the invention] However, in general, unless the AE waves are particularly high in a high noise area, only the AE waves caused by crack generation are identified and captured by filter processing and threshold determination. That is difficult. For example, a coolant is used during induction hardening, but when this coolant evaporates on the surface of a heated workpiece, so-called cooling noise is generated. Especially in the initial stage of cooling,
It generates high noise similar to that during heating. If a crack occurs at such a time, the AE wave will be buried in the noise and cannot be detected by filter processing or threshold determination.

Therefore, for workpieces in which no AE waves were detected, crack inspection by magnetic particle testing or the like is still required, and automation of the inspection is still insufficient.

The present invention can reduce A caused by cracks even in the presence of high noise.
The purpose of this invention is to realize a detection device that can identify and detect E waves and determine the state of crack occurrence in a workpiece during processing.

[Means for Solving the Problems] That is, the gist of the present invention is, as illustrated in the basic configuration diagram of FIG. M2, a periodicity detection means M3 that divides the waveform data detected by the acoustic emission detection means M2 into predetermined periods and detects periodicity for each period; and periodicity detected by the periodicity detection means M3. A crack detection apparatus is characterized in that it includes a crack occurrence determination means M4 that determines the crack occurrence state of the inspection object M1 based on the above.

[Effect] The AE wave at the time of crack generation shows high periodicity.

Therefore, based on the height of this periodicity, the state of crack occurrence is determined as follows.

That is, first, the acoustic emission detection means M2
detects the sound waves emitted from the test object M].

Next, in order to determine the periodicity of this waveform data, the periodicity detection means M3 divides the waveform data into predetermined periods and detects the periodicity of the waveform data for each predetermined period. In other words, if you divide the waveform data into predetermined periods and examine periodicity, you can increase the proportion of the AE wave component in the noise during periods when the AE wave level is high, rather than examining the entire waveform data. . This predetermined period is appropriately set depending on the measurement object, measurement conditions, etc. Moreover, the height of periodicity can be obtained, for example, by correlation calculation.

Based on the height of the periodicity thus obtained, the crack occurrence determining means M4 determines the crack occurrence state of the inspection object M1. For example, the periodicity is checked at every predetermined time in a time series, and if it is found that there is a period with periodicity above a certain level, it is determined that a fi crack has occurred.

[Example] Next, the present invention will be explained in detail. The present invention is not limited to these, but includes various embodiments without departing from the gist thereof.

FIG. 2 shows a block diagram of a crack detection device used in the induction hardening device 1 as a first embodiment.

The induction hardening device itself is a device used for boat fishing. This induction hardening apparatus 1 is configured to support a hardening member 3 with chucks 5, 7, etc., and perform induction hardening with a hardening coil 9. An AE sensor 1 is embedded in the upper chuck 7 and detects ultrasonic waves transmitted from the hardened member 3 to the upper chuck 7 during hardening.

The crack detection device includes an AE detection section] 3, a periodicity detection section 1
5, and a crack occurrence determination section 17.

The AE detection unit 13 includes an AE sensor]], a preamplifier]9,
main amplifier 21, filter 23, sample hold circuit 25, A/D converter 27, setting section 29, memory 3],
It consists of a gate circuit 33 and a start signal generator 35.

The output of the AE sensor 11 is amplified by a preamplifier 19 and a main amplifier 21, and a filter 23 removes low-level noise components. If the waveform input to the sample hold circuit 25 at this time is expressed in absolute value, it is shown in Figure 3 (A).
It becomes as follows.

Next, the sample and hold circuit 25 samples the AE wave at a sampling period set in advance by the setting section 29. This sampling period is appropriately set so that the signal waveform passing through the filter 23 can be sufficiently expressed.

The analog output held by the sample hold circuit 25 is converted into a digital value by the A/D converter 27. This digital value is then stored in the memory 31 and accumulated as 2-discretized waveform data. The gate circuit 33 controls the period during which this discretized waveform data is stored in the memory 3]. The period control is performed by the control section of the induction hardening equipment (
This is done by opening and closing the gates using the output signal of a start signal generator 35 provided in a device (not shown) as a trigger signal. For example, if the start signal generator 35 is set to output a high-level signal for a predetermined period of time t immediately after high-frequency heating ends, as shown in FIG. The discretized waveform data of the ultrasonic wave during the cooling process shown in b) will be stored in the memory 3].

Next, the periodicity detection unit]5 reads a predetermined number of samples in chronological order from the discretized waveform data stored chronologically in the memory 3]. Then, for each sampling number, a product-sum operation is performed using an internal delay circuit and a product-sum circuit. This product-sum operation is repeated by increasing the number of delays by one until a predetermined number of delays is reached. Autocorrelation calculation processing is performed in this manner.

The autocorrelation calculation can be expressed numerically as shown in the following equations (1) and (2), and by changing k and performing the calculation, the array of autocorrelation function values R(k) can finally be obtained.

...(1) R(k) = R(k) /R(0)...(2) Here, x O represents the discretized waveform data, and R(0) (Yo formula (
1) represents the value of R(k) when the number of delays = o. Further, according to equation (2), R(k) becomes a numerical array of 1 or less.

FIGS. 4(A) and 4(B) show discretized waveform data with a sampling number N at different sampling positions.

These correspond to one section in Figure 3 (a).

The autocorrelation calculation results are shown in Figures 5 (A) and (B), respectively.
).

Here, in the data in Figure 4 (A), even if the number of delays k increases, the autocorrelation function value R(k) does not easily converge to "0" as shown in Figure 5 (A). I understand. On the other hand, it can be seen that the data in FIG. 4(B) rapidly converges as the number of delays k increases, as shown in FIG. 5(B). It is known that data that does not rapidly converge in autocorrelation calculations, that is, data that has high autocorrelation, has high periodicity. Therefore, it can be seen that the periodicity in FIG. 4(A) is higher than that in FIG. 4(B).

By outputting this array of autocorrelation function values R(k) from the periodicity detection unit 15 to the crack occurrence determination unit 17 in order, it is determined whether the autocorrelation is high or not, that is, whether the periodicity is high or not. is determined.

The crack occurrence determination unit 17 includes a determination unit 37 and a section setting unit 3.
9, and a reference setting section 4].

The determination unit 37 sends autocorrelation function values R(k) existing in a predetermined analysis interval from the array of autocorrelation function values R(k) input from the periodicity detection unit 5 to the reference setting unit 4. It is determined whether the periodicity is high by comparing it with a reference value set by . This reference value is a value that is appropriately set in order to distinguish between highly periodic AE waves caused by cracks and low periodic noise. The predetermined analysis interval is set in advance in the interval setting section 39, and indicates the range of the number of delays.

This process of crack occurrence determination section]7 is carried out by the periodicity detection section]
This is performed every time the array data of autocorrelation function values R(k) from 5 is input.

Comparing FIG. 5(A) and FIG. 5(B), it is clear that as the number of delays k increases, a difference appears in the autocorrelation function value R(k). Therefore, for example, the magnitude of the autocorrelation function value R(k) in a delay number range of 1 or more for a predetermined number of delays is distinguished by an appropriate reference value (L). becomes possible.

The results determined in this way are sequentially output to the outside. For example, it may be output to automatic line control equipment or displayed on a display device such as a CRT.

On the control device side of the automatic line, based on this output, it is possible to determine whether or not cracks have occurred in the quenched member 3 during the cooling process of quenching. Therefore, if even one periodicity is higher than the reference value, it is known that a crack has occurred in the hardened member 3, so that defective products can be immediately removed from the line after the hardening process.

In this way, real-time crack inspection is possible without the need for human intervention, which can contribute to improving the automation rate of the line.

In the above embodiment, there is one periodicity detection unit]5, but
Two or more units may be provided to process the discretized waveform data in the memory 3 and alternately output the data to the crack occurrence determination unit 17.

In this way, multiple autocorrelation calculation processes can be quickly performed on the discretized waveform data. Therefore, the real-time effect can be further improved.

Next, a second embodiment of a crack detection device configured using a computer will be described. The second embodiment differs only partially from the first embodiment.

The different configurations are shown in the block diagram of FIG.

Here, the signal from the filter 23 is input to the A/D board 51. This A/D board 5] corresponds to the sample hold circuit 25, A/D converter 27, and setting section 29 of the first embodiment. The periodicity detection unit 15 detects this A/
Read the data directly from the buffer of D board 5],
Autocorrelation calculation is performed. This calculation result is read into the computer 53, and a process for determining whether or not a crack has occurred is performed. The reading timing of the computer 53 is I1.
Start signal generator 3 input via 0 board 55
This is done based on the start and end signals from 5. Here, the timing of the start and end signals is set to read the discretized waveform data of the cooling process. Similarly, a section setting signal is input from a keyboard (not shown) to designate the range in the gap determination calculation. The result of the autocorrelation calculation and its judgment result are output directly to the CRT 57 and also to the control device of the automatic line as a pass/fail judgment signal.

In addition, the microcomputer 53 includes a CPU, ROM, and RA.
The hardware configuration is generally well known, so a detailed explanation will be omitted.

The control process is shown in the flowchart of FIG. First, input the section setting signal, and perform the judgment calculation (step 1), which will be described later.
60) Set the calculation range (k s, k e) of (
Step 1]0). Next, it waits for a start signal to be output from the start signal generator 35 (step 120).

When the start signal is input, an autocorrelation calculation start command is output to the periodicity detection unit 15 (step 130).
. As a result, the periodicity detection section 5 reads a predetermined number of discretized waveform data from the buffer and performs an autocorrelation calculation while causing the A/D board 51 to perform A/D conversion at a predetermined timing.

The computer 53 side waits for this autocorrelation calculation to be completed (step 140), and when it is completed, reads the calculation result from the periodicity detection unit 15 (step 150).
. Next, perform a judgment calculation on the calculation result (step) 6
0).

Details of this determination calculation are shown in FIG. First, the lower limit value ks read in step 110 is set to delay k, and variable n is further cleared (step 160-1).

Next, the autocorrelation function value R obtained by autocorrelation calculation
We will consider the array (k) one by one. An example of the arrangement is shown in FIG. That is, first, the absolute value of R(k) is 0. 1
It is determined whether or not this is the case (step 160-2). At first, green = kS, so the absolute value of R(kS) is 0. It is determined whether the number is 1 or more.

If the determination is affirmative, the variable n is incremented (step 160-3), and it is determined whether k has exceeded the upper limit value ke (step 160-4). Step 160-
If the determination in step 2 is negative, the process moves to step 160-4 without incrementing the variable n. If a negative determination is made in step 160-4, the delay number k is incremented (step 160-5), and the process starts again from step 160-2.

In the processing of steps 160-2 to 160-5, the absolute value of the array R(k) from ks to ke is 0. The number of items greater than or equal to 1 is counted.

Then, if an affirmative determination is made in step 160-4, the variable n
It is determined whether or not is greater than or equal to the reference value nc (step 160
-6). Note that the process in step 160-6 is the same as the determination of whether or not there is a crack (step 170) in FIG.

Returning to FIG. 7, if an affirmative determination is made in step 170, since n is greater than or equal to nc, it means that the autocorrelation is sufficiently high as shown in FIG. High autocorrelation means high periodicity. Therefore, it can be seen that the ultrasonic waves detected at this time are AE waves generated due to cracks. Therefore, a defective product determination is output (step 190). When the control device on the line side receives this defective product determination output, the hardened member 3 is removed from the line as a defective product.

If a negative determination is made in step 170, it means that the AE wave is not caused by a crack, so the process returns to step 130 and the periodicity detection unit 15 is instructed to calculate the autocorrelation of the next array and send it. Output the command. In this way, similar processing is repeated. If an affirmative determination is made in step 170 even once in this repeated process, it means that the hardened member 3 has cracked and is removed from the line.

In this way, until the end signal is input from the start signal generator 35, an affirmative determination in step 170 must not be made once (
A melon determination output is made (step 200). The control device on the line side receives this non-defective determination output, moves the hardened member to the next process, and sets a new member in the induction hardening apparatus 1.

In this embodiment, part of the processing that was executed by the circuit in the first embodiment is executed by a program. Therefore, it is possible to change the processing content to any desired value by changing the program. For example, the above autocorrelation determination (step 160-
2 to 160-6), the local maximum value and local minimum value of the autocorrelation function value R(k) shown in Fig. 9 with the mark "*" are detected, and if the average value of the absolute values is larger than the predetermined value. It can also be determined that the periodicity is high. It is also possible to calculate the sum of the absolute values of the autocorrelation function values R(k) up to the number of delays ks-ke, and determine that the periodicity is high if the sum is larger than a predetermined value.

Therefore, on the computer 53 side, the autocorrelation function value R(
By configuring the program to select which periodicity determination calculation process to use according to the data situation in k), a more appropriate determination can be obtained.

Further, in this embodiment, there is one periodicity detecting section 15, but in order to speed up the processing, two or more periodicity detecting sections 15 are provided as shown by dotted lines in FIG. The autocorrelation calculation may be executed by switching.

Next, as a third embodiment, a block diagram in which the autocorrelation calculation is also performed on the computer side is shown in FIG. The difference from the second embodiment is that the periodicity detection section 15 is not present and is implemented as software on the computer 300 side.

In terms of processing, steps 130 to 15 shown in FIG.
The point is that the process shown in FIG. 11 is performed instead of the process of 0.

First, a predetermined number N of discretized waveform data is transferred to the A/D board 51.
(step 310).

Next, an autocorrelation function value R(0) is calculated using the following equation (3) (step 320).

...(3) This applies when the number of delays is =o in the above equation (1).

Next, the number of delays k l = k s is set (step 33
0). Next, from k = k s, the value of the autocorrelation function value R(k) is calculated using the above equation (1) (step 340
). Next, an autocorrelation function value R(k) converted to 1 or less is determined using the above equation (2) (step 350). Next, it is determined whether k has exceeded the upper limit value ke (step 360). If a negative determination is made here, k is incremented (step 370), and again at step 340.
- The processing of steps 360 is repeated. Step 36
If an affirmative determination is made in 02, the next determination calculation step 160
After that, processing is performed in the same manner as in the second embodiment.

In addition to the effects of the second embodiment, the present embodiment can calculate the autocorrelation function value R(k) programmatically (this allows the range to be limited to a desired number of delays, thereby eliminating unnecessary autocorrelation function values R(k)). (k
) there is no need to calculate.

In each of the above embodiments, an autocorrelation calculation is used as the correlation calculation, but a cross-correlation calculation, which is another type of correlation calculation, can also be used. When using this cross-correlation calculation, prepare discretized waveform data as a reference in advance, and
With the discretized waveform data detected from [sensor] 1,
The cross-correlation function value Rc (k) may be obtained by executing the calculations shown in the following equations (4) and (5). Then, the periodicity state may be determined by performing the determination processing shown in each embodiment on this array of cross-correlation function values Rc (k).

...(4) Rc (k)=R(k)/R(0)-(5) where xh
<Discretized waveform data detected from the AE sensor 11, y
represents the reference discretized waveform data.

[Effects of the Invention] According to the present invention, the periodicity of the AE wave is checked every predetermined period, and it is detected from the periodicity whether or not a crack has occurred. Noise and AE waves generated due to crack generation are significantly different in their periodicity. Therefore, even in the presence of high noise, the state of crack occurrence can be easily determined by checking the periodicity.

[Brief explanation of drawings]

Fig. 1 is a diagram illustrating the basic configuration of the present invention, Fig. 2 is a system block diagram of the first embodiment, Fig. 3 is an explanatory diagram of the output waveform of the AE sensor and start/end signals, and Fig. 4 ( A)
is the discretized waveform diagram of the AE wave generated due to the crack, the fourth
Figure (B) is a discretized waveform diagram of noise, Figure 5 (A) is an array graph of correlation function values R(k) obtained by autocorrelation calculation of AE waves generated due to cracks, Figure 5 (B ) is an array graph of correlation function values R(k) obtained by calculating the autocorrelation of noise, FIG. 6 is a block diagram of the main part of the second embodiment, and FIG. 7 is a graph of the second embodiment.
FIG. 8 is a flowchart showing the details of the determination calculation in the embodiment; FIG. 9 is an explanatory diagram of the process for determining periodicity from the array of correlation function values R(k); and FIG. The figure is a block diagram of the main parts of the third embodiment,
FIG. 11 is a flowchart of the autocorrelation calculation. 1... Induction hardening device 11... AE sensor 13... AE detection section 1
5... Periodicity detection section 17... Crack occurrence determination section 37... Judgment section 41... Standard setting section 53.
300...computer

Claims (1)

[Claims] Acoustic emission detection means for detecting sound waves emitted from an object to be inspected; and a period for dividing waveform data detected by the acoustic emission detection means into predetermined periods and detecting periodicity for each period. A crack detection device comprising: a periodicity detection means; and a crack occurrence determination means for determining a crack occurrence state of the object to be inspected based on the periodicity detected by the periodicity detection means.