Defect inspection method for steel sheet and defect inspection device for steel sheet

Abstract

To provide a defect inspection method and defect inspection device capable of appropriately detecting defects when inspecting defects in steel sheets by C-direction excitation.SOLUTION: A defect inspection method for steel sheets inspects defects in steel sheets on the basis of a plurality of detection signals detected along a width direction of the steel sheets. The defect inspection method for steel sheets includes: a detection signal acquisition step of acquiring a plurality of data sequences for every time series, with the plurality of detection signals defined as one data sequence; a noise removal step of removing noise included in the plurality of data sequences acquired in the detection signal acquisition step; a detection map generation step of arranging the plurality of data sequences subjected to the noise removal step for every time series to generate a detection map; and a defect data generation step of generating defect data, the data regarding defects in the steel sheets, on the basis of the detection map.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present invention relates to a method and an apparatus for inspecting defects in thin steel sheets, for example, for detecting defects in thin steel sheets. [Background technology]

[0002] Non-destructive detection of defects on the surface or inside of thin steel sheets is carried out on the production line (hereinafter also referred to as online) where soft magnetic thin steel sheets such as tinplate steel sheets for beverage cans and steel sheets for automobiles are manufactured.

[0003] A known example of such an inspection device is a defect inspection device that detects magnetic flux resulting from defects in a thin steel plate that is an object to be inspected. The defect inspection device includes an exciter (magnetizer) that saturates magnetic flux in an inspection area of ​​the thin steel plate, and a detector that detects magnetic flux leaking from the magnetized thin steel plate. That is, the defect inspection device detects defects in the thin steel plate by detecting magnetic flux resulting from defects that have leaked from the thin steel plate.

[0004] In general, defect inspection equipment has magnetic poles that apply a magnetic field to the thin steel plate (the object to be inspected), arranged in the direction of conveyance of the thin steel plate (hereinafter also referred to as L-direction excitation). In L-direction excitation, the thin steel plate is magnetized in its longitudinal direction, and multiple detectors that detect magnetic flux leaking from the thin steel plate are arranged in its width direction.

[0005] However, when the magnetic poles are arranged in the conveyance direction of the thin steel sheet, the ability to detect defects extending in the conveyance direction of the thin steel sheet decreases. That is, the magnetic flux caused by defects extending in the conveyance direction changes little over time and is detected as a low-frequency signal. This makes it difficult to improve the defect detection sensitivity.

[0006] Patent Document 1 discloses a leakage magnetic field inspection method that performs L-direction excitation. In Patent Document 1, information on the size of defects is extracted using a low-frequency side defect detection unit that detects defects in a low-frequency range from the detection signal, and a high-frequency side defect detection unit that detects defects in a high-frequency range.

[0007] One method of magnetic leakage testing is to magnetize a thin steel plate in its width direction and detect magnetic flux leaking from the thin steel plate (hereinafter also referred to as C-direction excitation). For example, Patent Document 2 discloses a configuration of C-direction excitation in which a thin steel plate is magnetized in its width direction to inspect for defects. [Prior art documents] [Patent documents]

[0008] [Patent Document 1] Japanese Patent Application Laid-Open No. 2014-106136 [Patent Document 2] International Publication No. 2021 / 125186 Summary of the Invention [Problem to be solved by the invention]

[0009] However, in Patent Document 1, the thin steel plate is magnetized along the conveyance direction of the thin steel plate, resulting in small leakage magnetic flux and low defect detection sensitivity by the detector, which causes a problem that defects extending along the conveyance direction cannot be detected with high accuracy.

[0010] In Patent Document 2, the signal detected by the detector changes slowly over time, resulting in a low-frequency signal, which means that defects extending along the conveyance direction cannot be detected with high accuracy.

[0011] Furthermore, since the signals detected by the detector contain noise of various frequencies (vibrations, power supply frequencies, drive motors, etc.), it can be difficult to extract signals caused by defects that extend long in the direction of movement of the thin steel plate, which leaves room for improvement.

[0012] The present invention has been made in consideration of the above problems, and aims to provide a defect inspection method and a defect inspection device that can appropriately detect defects when inspecting defects in thin steel plates using C-direction excitation. [Means for solving the problem]

[0013] In order to solve the above problems, the present invention has the following features.

[0014] [1] A method for inspecting defects in a thin steel plate, based on a plurality of detection signals detected along a width direction of the thin steel plate, comprising: a detection signal acquisition step of acquiring the plurality of data sequences for each time series by treating the plurality of detection signals as one data sequence; a noise removal step of removing noise contained in the plurality of data strings acquired in the detection signal acquisition step; a detection map generating step of generating a detection map by arranging the plurality of data strings that have been subjected to the noise removal step in the time series; a defect data generating step of generating defect data, which is data relating to defects in the thin steel plate, based on the detection map. [2] The method for inspecting defects in thin steel sheets according to [1], wherein the noise removal step is performed by at least one of band-pass filtering and wavelet transformation. [3] The method for inspecting defects in thin steel plates according to [1] or [2], wherein in the detection map generation step, the detection map is generated by arranging multiple data sequences corresponding to one thin steel plate in chronological order. [4] A thin steel plate defect inspection device that inspects a thin steel plate for defects based on a plurality of detection signals detected along a width direction of the thin steel plate, a detection signal acquisition unit that treats the plurality of detection signals as one data string and acquires the plurality of data strings for each time series; a noise removal unit that removes noise contained in the plurality of data strings acquired by the detection signal acquisition unit; a detection map generating unit that generates a detection map by arranging the plurality of data strings from which noise has been removed by the noise removing unit in time series; a defect detection unit that detects defects in the thin steel plate based on the detection map. [Effects of the Invention]

[0015] According to the present invention, the method includes a detection map generation step of generating a detection map by arranging multiple data sequences from which noise has been removed in chronological order, and a defect detection step of detecting defects in the thin steel plate based on the detection map. That is, by combining the widthwise data sequences from which defects have been detected in the longitudinal direction, a detection map is generated as a two-dimensional image, which enables accurate defect detection. As a result, for example, defects formed in the longitudinal direction of the thin steel plate can be accurately detected. [Brief explanation of the drawings]

[0016] [Figure 1] FIG. 1 is a configuration diagram showing a configuration of a defect inspection device. [Figure 2] FIG. 2 is an explanatory diagram showing an aspect of defect inspection of an inspection object using a defect inspection device. [Figure 3] FIG. 2 is a block diagram showing the configuration of a defect detection unit in FIG. 1. [Figure 4] 1 shows a process flow of a method for inspecting defects in thin steel plates. [Figure 5] 5 is an explanatory diagram showing how one data string is arranged in the detection signal acquisition step S01 of FIG. 4. FIG. [Figure 6] 1 is a graph illustrating a wavelet function. [Figure 7] 5 is an explanatory diagram showing how one data string is acquired in the detection signal acquisition step S01 of FIG. 4. FIG. [Figure 8]FIG. 5 is an explanatory diagram showing how noise is removed from one data string in the noise removal step S02 of FIG. 4. [Figure 9] 5 is an explanatory diagram showing a manner in which a detection map is generated in the detection map generating step of step S03 in FIG. 4. FIG. [Figure 10] FIG. 4 is an explanatory diagram illustrating an example of a detection map. DETAILED DESCRIPTION OF THE INVENTION

[0017] Fig. 1 shows the configuration of a defect inspection device for thin steel plates according to the present invention. As shown in Fig. 1, the defect inspection device 100 includes a transport unit 10 that transports a thin steel plate (hereinafter also referred to as a strip) SP, which is a plate-shaped object to be inspected, in one direction (a direction perpendicular to the plane of Fig. 1), an exciter 20 that magnetizes the strip SP, a plurality of detectors 31-36 that detect magnetic flux leaking from the strip SP, a defect detection unit 40 that detects defects in the strip SP based on the detection signals of each of the plurality of detectors 31-36, and a display unit 50 that displays the detection results by the defect detection unit 40.

[0018] Examples of the strip SP include soft magnetic materials such as tinplate, TFS (tin-free steel), galvanized steel, and raw sheets of galvanized steel. The thickness of the strip SP is preferably 0.1 to 3.2 mm, and more preferably 0.1 to 2.0 mm. For example, when the strip SP is for beverage cans, it should be about 0.2 mm, and when it is for automotive steel sheets, it should be about 0.8 mm.

[0019] FIG. 2 shows an embodiment of defect inspection of strip SP. As also shown in FIG. 2, conveying section 10 includes multiple non-magnetic rolls 11 made of a non-magnetic material. In this embodiment, two non-magnetic rolls 11 are arranged along the direction of the arrow, which is one direction (flow direction). Therefore, conveying section 10 can convey strip SP in one direction (flow direction). Note that another non-magnetic roll (not shown) may be arranged between the two non-magnetic rolls 11 of conveying section 10 so that strip SP is wound around it, and exciter 20 may be arranged on the opposite side of the contact surface between strip SP and the other non-magnetic roll.

[0020] As shown in FIG. 1, the exciter 20 has an excitation yoke 21 disposed so as to face the strip SP, and an excitation coil 22 as a magnetic field generating means.

[0021] As shown in Fig. 2, the excitation yoke 21 has a pair of columnar legs 21a. The pair of legs 21a have different magnetic poles (N pole, S pole) formed on one end and are arranged along a direction perpendicular to one direction. The excitation yoke 21 has a beam-shaped connecting portion 23 connecting the other ends of the pair of legs 21a. The material of the excitation yoke 21 may be any soft magnetic material. As a specific material for the excitation yoke 21, for example, a steel material such as a general structural rolled steel material (e.g., SS400) can be used in practical terms, taking into consideration economic efficiency.

[0022] The excitation yoke 21 is formed in a gate shape (U-shape) with an open bottom side (one end side) and a closed top side (the other end side) when viewed from the conveying direction (one direction) of the conveying section 10. Therefore, the excitation yoke 21 has an opening OP provided between one end sides of the pair of legs 21a.

[0023] One end of each of the pair of legs 21a is provided close to the surface of the strip SP. That is, the pair of legs 21a are provided on the surface of the strip SP so as to face the surface. In other words, the exciter 20 is provided at a position facing the strip SP.

[0024] Here, facing the exciter 20 and the strip SP means that one end of each leg 21a of the excitation yoke 21 is positioned facing the strip SP, and the angle that each leg 21a makes with respect to the surface of the strip SP is approximately perpendicular.

[0025] In this embodiment, the pair of legs 21a are arranged such that their axial direction is perpendicular to the surface of the strip SP. Each of the pair of legs 21a is provided with a gap between it and the surface of the strip SP.

[0026] The pair of legs 21a are preferably arranged so that the distance between one leg 21a and the surface of the strip SP is substantially the same as the distance between the other leg 21a and the surface of the strip SP. The distance from one end of each leg 21a to the surface of the strip SP is preferably about 0.5 to 7.0 mm.

[0027] The excitation coil 22 is wound around the connection portion 23 of the excitation yoke 21. When a direct current is supplied to the excitation coil 22, the excitation coil 22 generates a direct current magnetic field and magnetizes the excitation yoke 21. In other words, when a direct current is supplied to the excitation coil 22, the excitation coil 22 and the excitation yoke 21 become electromagnets.

[0028] The excitation coil 22 becomes a powerful electromagnet by increasing at least one of the number of turns and the current supplied thereto. The excitation coil 22 supplies magnetic flux to the inspection area of ​​the strip SP via the excitation yoke 21, magnetizing the strip SP.

[0029] The number of turns of the excitation coil 22 for each leg 21a is preferably about 400 to 2,000. The current supplied to the excitation coil 22 is preferably about 1 to 9 A. When the excitation yoke 21 is magnetized, different magnetic poles are formed at one end of the pair of legs 21a. Specifically, the tip of one leg 21a becomes an N pole, and the tip of the other leg 21a becomes an S pole.

[0030] The exciter 20 uses the DC magnetic field to supply a strong magnetic flux to the inspection area of ​​the strip SP, magnetizing the strip SP. Specifically, when the magnetic flux is supplied to the inspection area of ​​the strip SP, the magnetic flux density becomes approximately 1.7 T or more, and the inspection area of ​​the strip SP is in a magnetic flux saturated state or a state close to magnetic flux saturation.

[0031] The length of the exciter 20 in the depth direction (the length perpendicular to the paper surface in FIG. 1, the length along the flow direction in FIG. 2) can be set arbitrarily depending on the size of the strip SP. For example, the length of the exciter 20 in the depth direction is preferably about 40 to 100 mm.

[0032] If the depth of the exciter 20 is less than 40 mm, the magnetic field formed in the opening OP (between the pair of legs 21 a) tends to be non-uniform, which may affect the defect detection accuracy. Also, if the depth of the exciter 20 exceeds 100 mm, more excitation energy than necessary is required to detect defects, which tends to reduce energy efficiency.

[0033] 1, in this embodiment, the strip SP is excited in a cross direction (C direction, width direction of the strip SP) perpendicular to the transport direction (one direction) of the strip SP by the exciter 20 (hereinafter also referred to as C-direction excitation). Therefore, when detecting long defects occurring in the transport direction (one direction) of the strip SP, excitation in the cross direction is effective.

[0034] In the case of C-direction excitation, the number of exciters 20 arranged in the width direction of the strip SP can be reduced by widening the opening OP of the excitation yoke 21 (increasing the distance between the pair of legs 21a), thereby reducing the equipment costs.

[0035] From this viewpoint, it is preferable that the interval L of the openings OP (the distance between the pair of legs 21a) is 40 to 600 mm. If the interval L of the openings OP is less than 40 mm, it is not possible to reduce the number of exciters 20 arranged in the width direction of the strip SP, and there is a risk that the equipment cost cannot be sufficiently reduced.

[0036] Increasing the spacing L between the openings OP is desirable because it widens the widthwise range that can be detected by a single exciter 20. However, the longer the spacing L between the openings OP, the weaker the magnetic flux supplied to the strip SP. Therefore, if the spacing L exceeds 600 mm, there is a risk that the defect detection sensitivity will decrease. Furthermore, if the spacing L between the openings OP exceeds 600 mm, the weight of the exciter 20 will increase, and the accompanying equipment for holding it will also tend to become larger.

[0037] From the viewpoint of ensuring sufficient magnetic flux supplied to the strip SP, the interval L of the openings OP is preferably 300 mm or less, more preferably 200 mm or less, and even more preferably 100 to 150 mm.

[0038] When inspecting a wide strip SP, it is preferable to install a plurality of exciters 20 in the width direction so that the entire width of the strip SP can be inspected.

[0039] The detectors 31 to 36 detect magnetic flux (magnetic lines of force) leaking from the strip SP in a magnetically saturated state. The detectors 31 to 36 include coil elements and Hall elements.

[0040] 1, a plurality of detectors 31 to 36 are arranged in the opening OP in the direction in which the strip SP is magnetized. In other words, the detectors 31 to 36 are arranged on one end side of the pair of legs 21a from one to the other of the pair of legs 21a.

[0041] It is preferable that a plurality of detectors 31 to 36 are provided according to the distance between the magnetic poles, i.e., the interval between the pair of legs 21a. There is no particular limit to the number of detectors 31 to 36 provided, but in this embodiment, six detectors are provided at predetermined intervals from each other.

[0042] When each of the detectors 31 to 36 detects a magnetic flux, it transmits to the defect detection unit 40 an analog voltage signal whose magnitude corresponds to the magnetic flux.

[0043] The defect detection unit 40 detects that the strip SP has a defect based on the detection signals transmitted from the detectors 31 to 36. That is, the defect detection unit 40 has a function of detecting a defect signal corresponding to a defect in the strip SP. When the defect detection unit 40 detects that the strip SP has a defect, it causes the display unit 50 to display the detection result.

[0044] The display unit 50 displays the detection results transmitted from the defect detection unit 40. For example, a display or the like can be used as the display unit 50. The detection results can be displayed, for example, by displaying the words "Defect Found."

[0045] Fig. 3 shows the configuration of the defect detection unit 40. As shown in Fig. 3, the defect detection unit 40 includes a detection signal acquisition section 41 that acquires detection signals transmitted from a plurality of detectors 31 to 36 arranged in the opening OP.

[0046] The defect detection unit 40 includes an amplifier section 42 that performs an amplification process on the signals from each of the detectors 31 to 36 that have passed through the detection signal acquisition section 41, and an A / D conversion section 43 that converts the signals from each of the detectors 31 to 36 that have been amplified by the amplifier section 42 into digital signals. The defect detection unit 40 also includes a noise removal section 44 that removes noise contained in the digital signals converted by the A / D conversion section 43.

[0047] The defect detection unit 40 includes a detection map generation unit 45 that generates a detection map using the data string from which noise has been removed by the noise removal unit 44. The defect detection unit 40 also includes a defect data generation unit 46 that generates defect data, which is data related to defects in the thin steel plate, based on the detection map. The defect detection unit 40 also includes a memory unit 47 that stores various data.

[0048] The detection signal acquisition unit 41 treats a plurality of detection signals as one data string and acquires a plurality of data strings in time series. Specifically, the detection signal acquisition unit 41 acquires each of the detection signals transmitted from the detectors 31 to 36 at a predetermined timing as one data string.

[0049] Note that a bias unit (not shown) that individually applies a bias signal to the detection signals of the detectors 31 to 36 may be provided between the detectors 31 to 36 and the amplifier unit 42. In this case, the bias unit may set a bias signal according to the magnetic environment formed at the opening OP, i.e., the magnetic gradient, and apply the set bias signal. Even if a magnetic gradient is formed at the opening OP, the effect of the magnetic gradient can be canceled from the digital signal converted by the A / D converter 43.

[0050] The amplifier unit 42 amplifies the detection signals transmitted from the detectors 31 to 36. For example, the amplifier unit 42 amplifies each of the detection signals transmitted from the detectors 31 to 36 by the same amplification factor.

[0051] The amplification factor may be set, for example, with the maximum voltage set to 5V, and may be set to a large amplification factor that is within the range of the maximum voltage and that allows defects in the strip SP to be detected with good sensitivity.

[0052] The A / D conversion unit 43 reduces the influence of noise that becomes a disturbance when determining defects detected in the strip SP. This allows the defect data generation unit 46 to accurately determine defects. The signal processing performed by the A / D conversion unit 43 may be set to, for example, about 96 Hz. By setting the A / D conversion unit 43 in this way, defects can be detected effectively even when the transport speed of the strip SP is set to 700 m / min.

[0053] The noise removal unit 44 removes noise contained in the multiple data strings acquired by the detection signal acquisition unit 41. The noise removal unit 44 acquires the digital signals converted by the A / D conversion unit 43 as data strings at predetermined time intervals, and removes noise contained in the acquired data strings.

[0054] The detection map generating unit 45 generates a detection map by arranging a plurality of data strings corresponding to one strip SP in chronological order.

[0055] The defect data generation unit 46 may have the same functions as the defect detection unit used in a conventional defect detection unit. That is, the defect data generation unit 46 has a computer (not shown) including a CPU. For example, the defect data generation unit 46 compares the defect signal with a threshold value stored in the memory unit 47, and generates defect data indicating that the strip SP has a defect if the defect signal exceeds the threshold value. Also, the defect data generation unit 46 generates defect data indicating that the strip SP has no defect if the defect signal does not exceed the threshold value.

[0056] The storage unit 47 is a writable nonvolatile memory such as an EPROM. There are no particular limitations on the storage unit 47, and it may be, for example, a storage device such as an HDD or SSD. The storage unit 47 stores, for example, data sequences of detection signals, threshold values ​​used in defect detection by the defect data generation unit 46, defect detection results, and various programs.

[0057] Fig. 4 shows a process flow of the method for inspecting defects in thin steel sheets. As shown in Fig. 4, the detection signal acquisition unit 41 treats a plurality of detection signals as one data string, acquires a plurality of data strings for each time series, and executes the detection signal acquisition step (step S01).

[0058] The strip SP is excited in the C direction. The detectors 31 to 36 are arranged in the cross direction of the strip SP. Therefore, the detection signal acquisition unit 41 acquires each detection signal corresponding to the position of the strip SP in the cross direction as one data string. The detection signal acquisition unit 41 acquires one data string at every predetermined time. Each data string acquired by the detection signal acquisition unit 41 is stored in the memory unit 47.

[0059] The noise removal unit 44 removes noise contained in the plurality of data strings acquired in the detection signal acquisition step of step S01, thereby performing a noise removal step (step S02).

[0060] The noise removal unit 44 may store data sequences acquired at predetermined intervals (sampling periods) in the memory unit 47, and sequentially remove noise contained in each stored data sequence after a predetermined time (data collection time) has elapsed.

[0061] In this case, the noise removal unit 44 may acquire data strings from the detectors 31 to 36 at a sampling period set to, for example, 10 to 1000 ms (frequency 10 to 100 Hz).

[0062] The sampling period of the data strings acquired from the detectors 31 to 36 may be changed in synchronization with the transport speed of the strip SP. For example, the sampling period of the data strings acquired from the detectors 31 to 36 may be changed so that the pitch in the transport direction of the strip SP becomes constant. Each of the data strings from which noise has been removed by the noise removal unit 44 is stored in the memory unit 47.

[0063] The detection map generating unit 45 generates a detection map by arranging the multiple data strings that have been subjected to the noise removal step of step S02 in time series, and executes the detection map generating step (step S03). The detection map generated by the detection map generating unit 45 is stored in the storage unit 47.

[0064] The defect data generating unit 46 executes a defect data generating step of generating defect data relating to defects in the thin steel plate based on the detection map generated in the detection map generating step of step S03 (step S04). The defect data generated by the defect data generating unit 46 is stored in the storage unit 47.

[0065] Fig. 5 shows how a plurality of data strings are acquired in the detection signal acquisition step S01 in Fig. 4. The detection signal acquisition unit 41 first acquires a data string DC for each sampling period from the start of sampling of the detection signals by the detectors 31 to 36.

[0066] The data string DC includes data detected by the detectors 31 to 36. For example, the data detected by the detector 31 when the time t is "3" is recorded as d(1,3). That is, the data string DC shown in FIG. 5 includes d(1,3), d(2,3), d(3,3), d(4,3), d(5,3), and d(6,3). Note that the time is the elapsed time from the start of sampling.

[0067] The data string DC corresponds to the cross-direction position of the strip SP. The detection signal acquisition unit 41 generates detection data DD as a two-dimensional matrix of the widthwise position x of the strip SP and time t. The detection signal acquisition unit 41 stores the generated detection data DD in the storage unit 47.

[0068] For example, when the detection signal acquisition unit 41 acquires a new data string DC from the detectors 31 to 36, it may delete the oldest data string DC from the detection data DD stored in the storage unit 47 and update the data string as needed.

[0069] As a result, for example, a data string DC acquired during a predetermined data collection time is accumulated in the storage unit 47. It is preferable that the storage unit 47 stores a data string DC corresponding to a distance in the transport direction of the strip SP of up to about 100 m, for example.

[0070] Defects in the transport direction of the strip SP may be up to about 100 m long, so by storing data strings in the memory unit 47 that correspond to such lengths, defects can be detected effectively.

[0071] In step S02, a noise removal step, the noise removal unit 44 performs noise removal on the data string DC acquired as described above. The noise removal by the noise removal unit 44 preferably involves applying at least one of band-pass filtering and wavelet transform processing.

[0072] By removing noise in this manner, it is possible to remove noise mixed into the data string DC along the cross direction of the strip SP from the data string DC, thereby increasing the S / N ratio (signal-to-noise ratio) for defects in the strip SP.

[0073] The noise contained in the data string DC is mainly due to the magnetic gradient formed at the opening OP. When noise removal is performed on the detection data DD, it is preferable to remove noise from each data string DC acquired at the same timing.

[0074] The noise removal step for the data string DC by the noise removal unit 44 is performed as follows: When performing noise removal on the detection data DD shown in Fig. 5, it is advisable to perform noise removal on each data string DC acquired at the same timing.

[0075] In this embodiment, noise removal performed on the data string DC is characterized by being performed in the cross direction of the strip SP. That is, noise removal for L-direction excitation is performed on detection data acquired along the transport direction of the strip SP. The detection data acquired in the transport direction of the strip SP is data acquired over time, in other words, disturbances that occur over time are mixed in as noise. For this reason, when inspecting defects in the strip SP with L-direction excitation, such noise is removed.

[0076] The noise generated over time includes vibration of the strip SP, the power supply frequency caused by the power supply for the excitation coil 22, and the drive motor that drives the conveyor unit 10. Because these noises are disturbances of different frequencies, noise removal for the L-direction excitation requires, for example, complex adjustment of the cutoff frequency of a bandpass filter. Also, depending on the noise frequency, it can be difficult to extract signals caused by defects.

[0077] In contrast, the data string DC acquired when inspecting defects in the strip SP during C-direction excitation is less likely to be mixed with noise that occurs over time because the detectors 31 to 36 detect detection signals at synchronized timing. Therefore, when inspecting defects in the strip SP during C-direction excitation, it is sufficient to remove static noise that is mainly caused by the magnetic gradient of the opening OP.

[0078] Here, noise removal performed when inspecting defects in the strip SP during C-direction excitation, i.e., noise removal performed on the data string DC by the noise removal unit 44, is called C-direction noise removal. Additionally, noise removal performed on the digital signal converted by the A / D conversion unit 43 in the time axis direction or flow direction is called L-direction noise removal.

[0079] The noise removal method by the noise removal unit 44 can be, for example, processing using a filter applied in signal processing. Such noise removal methods can be, for example, low-pass filters, high-pass filters, band-pass filters, Butterworth filters, Bessel filters, etc. In this embodiment, it is preferable to apply at least one of band-pass filter processing and wavelet transform processing among these.

[0080] Bandpass filtering is a process for extracting signals in a predetermined frequency band from a data string. In contrast, the bandpass filtering process for C-direction noise removal in this embodiment does not extract signals in a specific frequency band, but extracts signals in a specific wavelength band.

[0081] Specifically, the bandpass filter process should extract signals in a wavelength band of 0.5 to 2.0λ, where λ is the representative width in the cross direction of the defect to be detected. The wavelength band extracted by the bandpass filter process does not need to be changed according to the transport speed of the strip SP, and should be set according to the size in the cross direction of the defect to be detected (defect width). The bandpass filter process should be set so as to remove signals with wavelengths corresponding to the magnetic gradient formed at the opening OP.

[0082] Wavelet transform processing refers to the process of applying a wavelet transform to an extracted data string. A mother wavelet that has been scaled and translated in the time direction is called a wavelet function. In wavelet transform processing, a Fourier transform is performed on this wavelet function and the signal to be verified. The inner product of these functions is then calculated, and the amount of the wavelet function component contained in the inner product is determined.

[0083] When inspecting defects in the strip SP with C-direction excitation, the detected data sequence is time-independent. In such cases, it is possible to determine the wavelet function, which is a feature of the wavelet transform, and calculate the inner product without performing a Fourier transform.

[0084] In the wavelet transform processing of this embodiment, the well-known Gabor function, Daubechies function, Meyer function, and Mexican hat function can be used as the wavelet function. For example, the wavelet function Ψ(x) shown in equation (1) can be applied. Here, x is the coordinate corresponding to the cross direction of the strip SP, and σ is the representative length of the strip SP in the cross direction. The representative length σ can be set arbitrarily, for example, to the interval L of the openings OP.

[0085] FIG. 6 illustrates the wavelet function shown in equation (1).

number

[0086] In the wavelet transform process, noise is removed by Ψ((xb) / a) using parameter a for scaling the wavelet function Ψ(x) in the C direction of the strip SP and parameter b for moving it in a direction corresponding to the C direction of the strip SP. In other words, using the signal waveform f(x) when the data string DC is continuous data, the wavelet transform W(a, b) can be calculated using equation (2).

[0087]

number

[0088] In the wavelet transform process of the above embodiment, it is preferable to perform a differentiation process on the data sequence DC as a preprocessing before performing the wavelet transform process. The differentiation process is a process for calculating the inclination of the data sequence DC with respect to the C direction of the strip SP. By removing noise contained in the data sequence DC through the preprocessing, it is possible to effectively remove noise caused by the magnetic gradient of the opening OP through the wavelet transform process.

[0089] 7 and 8 show an example in which noise is removed by wavelet transform processing. In the example shown in FIGS. 7 and 8, 80 detectors are provided, and the intensity of the detection signal from each detector is shown. In other words, FIGS. 7 and 8 show 80 pieces of data acquired along the cross direction of the strip SP. The vertical axis in FIGS. 7 and 8 represents the signal intensity after noise removal by wavelet transform processing.

[0090] Figure 7 shows the data string DC before noise removal. As shown in Figure 7, the data string DC contains noise, which increases the difference in signal intensity between adjacent detectors.

[0091] Figure 8 shows the data sequence DF after noise removal by wavelet transform processing has been performed on the data sequence DC of Figure 7. As shown in Figure 8, since noise has been removed from the data sequence DF, the difference in signal intensity between adjacent detectors is smaller than that of the data sequence DC. In the example shown in Figure 8, the signal-to-noise ratio is 3 (S / N=3).

[0092] Fig. 9 shows how a detection map is generated in the detection map generation step S03 in Fig. 4. As described above, data sequence DC is subjected to noise removal to generate data sequence DF. Detection map generation unit 45 generates detection map DM by associating data sequence DF with time t.

[0093] The detection map DM is, for example, numerical data expressed as a two-dimensional matrix of the data string DF after noise removal and the time t. The detection map DM is not limited to this embodiment, and may be, for example, numerical data expressed as a matrix of the data string DF and the transport distance over which the strip SP is transported. In other words, the detection map DM may be generated by converting information on the passage of time into the transport distance of the thin steel plate. In this way, the detection map DM can be generated by arranging multiple data strings corresponding to one strip SP in chronological order. This makes it possible to detect defects throughout the entire strip SP.

[0094] Fig. 10 shows an example of the detection map DM. In Fig. 10, the detection map DM is shown as a two-dimensional matrix of the data string DF and the time t, which is color-coded according to the data value of the data string DF.

[0095] In Figure 10, region 1 of data string DF closest to the threshold is colored orange. Region 2 of data string DF, which is next closest to the threshold after region 1, is colored pink. Region 3 of data string DF, which is next closest to the threshold after region 2, is colored red. Region 4 of data string DF, which is next closest to the threshold after region 3, is colored dark red.

[0096] The detection map DM may also be a map in which a two-dimensional matrix of the data string DF and time t is illustrated as contour lines according to the digital values ​​of the data string DF. By color-coding the detection map DM according to the size of the data string DF as shown in Fig. 10, for example, by displaying the detection map DM on the display unit 50, it becomes easier to identify defects that are long in the flow direction of the strip SP.

[0097] In the defect data generating step S04, if the numerical data of each data string DF exceeds a threshold value, defect data is generated indicating that the strip SP is defective.

[0098] The threshold value may be set based on the signal strength of the defective strip SP. For example, the intensity of a defect signal determined to be a defect based on past operational results may be identified, and the intensity of the defect signal may be used as a reference value to set the threshold value.

[0099] By using the detection map generated by the detection map generation unit 45, defects extending in the longitudinal direction of the strip SP can also be detected. The defect data generation unit 46 may generate defect data to include defect level information such as the length of the defect in the longitudinal direction of the strip SP. The defect level information may be generated in multiple stages, such as "long defect" and "short defect," depending on the level of the defect.

[0100] When generating defect data indicating that the strip SP has a defect, the defect data generating section 46 may cause the display section 50 to display the defect data as "defective."

[0101] The defect data generation unit 46 may generate defect degree information indicating the size of the defect and the degree of importance of the defect based on the difference from a threshold value of the numerical data in the two-dimensional matrix, and display the defect degree information on the display unit 50.

[0102] As described above, in this embodiment, the noise removal unit 44 performs C-direction noise removal on the data sequence DC. However, the noise removal unit 44 may perform L-direction noise removal on the detection data DD in addition to C-direction noise removal on the data sequence DC, which removes noise in the time axis direction or flow direction. When performing L-direction noise removal, at least one of standard bandpass filtering and wavelet transform processing can be applied. Known noise removal methods can be applied for L-direction noise removal, such as the method described in Patent Document 1. L-direction noise removal can remove noise caused by vibrations of the strip SP, power supply frequencies due to the power supply for the excitation coil 22, and the drive motor driving the conveyor unit 10.

[0103] As described above, the defect inspection apparatus 100 of the present invention makes it possible to detect defects with high sensitivity even when the magnetic flux density at the opening OP is non-uniform.

[0104] In this way, by using the data string DF from which noise has been removed in the noise removal step of step S02, it is possible to improve the detection sensitivity for defects present in the cross direction of the strip SP.

[0105] The defect information may include defect size, severity, type, defect degree information, etc. Furthermore, the defect information may be output using a machine learning model. For example, a machine learning model may be generated using data that associates detection maps determined to be defects with defect degree information including defect size, severity, type, etc. based on past operational records. By generating such a machine learning model, it is possible to input the detection map generated by the detection map generation unit 45 and output defect information including defect degree information.

[0106] Specifically, defects in thin steel sheets may be detected using a defect determination model that receives numerical data represented by a two-dimensional matrix as an input, which is a map divided by contour lines or colors, and outputs defect information. The defect information generated in this manner may be displayed on the display unit 50. [Explanation of symbols]

[0107] 100 Defect inspection equipment 10 Conveying section 20 exciter 21a Legs 31~36 Detectors 40 Defect Detection Unit 41 Detection signal acquisition unit 44 Noise reduction section 45 Detection map generation unit 46 Defect data generation unit

Claims

1. A method for inspecting defects in a thin steel plate, based on a plurality of detection signals detected along a width direction of the thin steel plate, comprising: a detection signal acquisition step of acquiring the plurality of data sequences for each time series, the plurality of detection signals being treated as one data sequence; a noise removal step of removing noise contained in the plurality of data strings acquired in the detection signal acquisition step; a detection map generating step of generating a detection map by arranging the plurality of data strings that have been subjected to the noise removal step in the time series; a defect data generating step of generating defect data, which is data relating to defects in the thin steel plate, based on the detection map.

2. 2. The method for inspecting defects in a thin steel plate according to claim 1, wherein the noise removal step is performed by at least one of band-pass filtering and wavelet transformation.

3. 3. The method for inspecting defects in a thin steel plate according to claim 1, wherein in the detection map generating step, the detection map is generated by arranging a plurality of the data strings corresponding to one thin steel plate in chronological order.

4. A thin steel plate defect inspection device that inspects a thin steel plate for defects based on a plurality of detection signals detected along a width direction of the thin steel plate, a detection signal acquisition unit that treats the plurality of detection signals as one data string and acquires the plurality of data strings in time series; a noise removal unit that removes noise contained in the plurality of data strings acquired by the detection signal acquisition unit; a detection map generating unit that generates a detection map by arranging the plurality of data strings from which noise has been removed by the noise removing unit in time series; a defect detection unit that detects defects in the thin steel plate based on the detection map.

Images