JP2020071125A - Method and device for determining defect, method for manufacturing steel plate, method for learning defect determination model, and defect determination model - Google Patents

Abstract

To provide a method and a device for determining a defect which can prevent false detection of detects.SOLUTION: The method for determining a defect according to the present invention generates an eddy current by giving an induction current to an inspection target body made of a conductor and determines the presence or absence of defects in the surface of the inspection target body on the basis of a signal waveform obtained from changes in the eddy current. The method includes the step of determining the presence or absence of defects in the surface of the inspection by inputting a signal waveform from the inspection target body to the defect determination model which has been caused to mechanically learn to determine whether the signal waveform is derived from a defect by using the result of evaluating the state of the inversion in the signal waveform as the feature amount.SELECTED DRAWING: Figure 7

Description

  The present invention relates to a defect determination method, a defect determination device, a steel sheet manufacturing method, a defect determination model learning method, and a defect determination model for determining the presence or absence of a defect on the surface of a subject.

  In general, defects such as balding and ear cracks formed on the surface of a subject such as a steel plate and a steel bar are inspected using an eddy current flaw detector. However, in the inspection using the eddy current flaw detection device, erroneous detection of determining that there is a defect despite the fact that the defect does not actually exist due to noise caused by vibration during transportation of the subject, surface roughness, and the like ( (Also referred to as over-detection) may occur. Therefore, when it is determined that there is a defect in the inspection using the eddy current flaw detector, the presence or absence of the defect is determined by visual inspection. However, the visual inspection requires labor and cost for the inspector.

  From such a background, Patent Document 1 discloses a self-comparison eddy current flaw detection method that includes two coils connected to each other at a certain distance and having different polarities for detecting an eddy current generated by applying an alternating magnetic field to a subject. In the device, as the defect determination circuit, depending on the polarity of the signal from the detector, a + side polarity determination circuit and a − side polarity determination circuit that compare each with a threshold value, and a signal after performing determination of the + side polarity signal And a waiting time measuring circuit that measures the time until the negative-side polarity signal is detected.

JP, 2003-4708, A

  However, in the technique described in Patent Document 1, it is impossible to completely eliminate the influence of noise caused by the vibration during transportation of the subject, the surface roughness, and the like, and therefore, there is still a problem that erroneous detection of defects occurs. is there.

  The present invention has been made in view of the above problems, and an object thereof is to provide a defect determination method and a defect determination device capable of suppressing erroneous detection of defects. Another object of the present invention is to provide a method for manufacturing a steel sheet, which is capable of accurately detecting defects and manufacturing steel sheets with a high yield. Further, another object of the present invention is to provide a defect determination model learning method and a defect determination model capable of accurately determining the presence or absence of a defect.

  The defect determination method according to the present invention is a defect for determining the presence or absence of a defect on the surface of a subject based on a signal waveform obtained by applying an induced current to a subject made of a conductor to generate an eddy current and changing the eddy current. In the determination method, with respect to the defect determination model machine-learned to determine whether or not the signal waveform is a signal waveform derived from the defect using the evaluation result of the state of polarity inversion in the signal waveform as a feature amount The method further comprises the step of determining the presence or absence of a defect on the surface of the subject by inputting a signal waveform from the subject.

  The defect determination method according to the present invention is characterized in that, in the above-mentioned invention, the defect determination model is machine-learned as an evaluation result of an amplitude of the signal waveform.

  In the defect determination method according to the present invention, in the above invention, the machine learning is any one of logistic regression analysis, a decision tree, a neural network, and deep learning.

  The defect determining apparatus according to the present invention is a defect that determines the presence or absence of a defect on the surface of the subject based on a signal waveform obtained by applying an induced current to the subject made of a conductor to generate an eddy current and changing the eddy current. In the determination device, the evaluation result of the state of polarity reversal in the signal waveform is used as a feature value for the defect determination model machine-learned so as to determine whether the signal waveform is a signal waveform derived from the defect. It is characterized by comprising a means for determining the presence or absence of a defect on the surface of the subject by inputting a signal waveform from the subject.

  A method of manufacturing a steel sheet according to the present invention is characterized by including a step of detecting a defect existing on the surface of the steel sheet by using the defect determination method according to the present invention and removing the detected defect.

  The learning method of the defect determination model according to the present invention is an evaluation result of the state of polarity reversal in a signal waveform obtained by applying an induced current to an object made of a conductor to generate an eddy current, and the signal. A determination result of whether or not the waveform is a signal waveform derived from a defect is used as learning data, the signal waveform is an input value, and a signal waveform corresponding to the input value is a signal waveform derived from the defect. It is characterized by including a step of learning a defect judgment model having a judgment value of whether or not as an output value by machine learning.

  The defect determination model according to the present invention is characterized by being learned by the defect determination model learning method according to the present invention.

  According to the defect determination method and the defect determination device of the present invention, it is possible to suppress erroneous detection of defects. Further, according to the method for manufacturing a steel sheet according to the present invention, it is possible to accurately detect defects and manufacture the steel sheet with a high yield. Furthermore, according to the defect determination model learning method and the defect determination model according to the present invention, the presence or absence of a defect can be accurately determined.

FIG. 1 is a diagram showing an example of an eddy current flaw detection signal derived from a defect and an eddy current flaw detection signal at the time of erroneous detection. FIG. 2 is a diagram showing an example of the value of the evaluation function obtained from the eddy current flaw detection signal derived from the defect and the eddy current flaw detection signal at the time of erroneous detection. FIG. 3 is a diagram showing another example of the value of the evaluation function obtained from the eddy current flaw detection signal derived from the defect and the eddy current flaw detection signal at the time of erroneous detection. FIG. 4 is a diagram illustrating an example of the defect determination process based on the model output value. FIG. 5 is a block diagram showing the configuration of the defect detecting apparatus which is an embodiment of the present invention. FIG. 6 is a schematic diagram showing the configuration of the eddy current flaw detector shown in FIG. FIG. 7 is a flowchart showing the flow of the defect determination processing which is an embodiment of the present invention.

  The inventors of the present invention have found that, when there is noise due to vibration or surface roughness during transportation of the subject, the eddy current flaw detection signal output from the eddy current flaw detection device 2 by comparison with a threshold value is a signal derived from a defect. It was found that it is difficult to accurately determine whether or not As a result of repeated intensive research, as shown in FIGS. 1A and 1B, an eddy current flaw detection signal (FIG. 1A) derived from a defect and an eddy current flaw detection signal at the time of erroneous detection (FIG. 1A) are obtained. Since there is a difference in the shape (waveform shape) of the eddy current flaw detection signal from (b)), the difference in the waveform shape of the eddy current flaw detection signal is learned by machine learning, and the eddy current is obtained using the machine learning model obtained by machine learning. It has been conceived to suppress erroneous detection of defects by determining whether the flaw detection signal is a signal derived from a defect.

  Next, the inventors of the present invention have selected a machine learning method. Specifically, as a machine learning method, there are methods such as logistic regression analysis, decision tree, neural network, deep learning, etc., but in the present invention, logistic regression analysis in which the relationship between input data and prediction results is easy to understand is machine learning. It was selected as the method. However, in the present invention, the logistic regression analysis is used as the machine learning method, but other machine learning methods such as decision tree, neural network, and deep learning may be used.

  However, in machine learning of the difference in the waveform shape of the eddy current flaw detection signal, it is difficult to use the waveform data of the eddy current flaw detection signal, which is non-linear data, as it is. Therefore, the inventors of the present invention have conceived to use the digitized waveform shape of the eddy current flaw detection signal as a feature amount (input value) of the machine learning model. Specifically, as shown in the following mathematical expression (1), the value of the evaluation function E (x), which is the numerical value of the amplitude and symmetry of the eddy current flaw detection signal, is set as the feature amount of the machine learning model.

  Here, in the mathematical expression (1), x is the scanning direction (longitudinal length) position of the object by the eddy current flaw detector, and k is the dimension in the scanning direction of the eddy current flaw detector (the distance between the detection coils S1 and S2 shown in FIG. 6). ) And a constant determined according to the scanning speed of the subject (see FIG. 2), f (x) indicates the output level value of the eddy current flaw detection signal at the position x in the scanning direction of the subject. Further, {| f (x) -f (x + k) |} is an index indicating the amplitude of the waveform shape of the eddy current flaw detection signal, and exp {-| f (x) + f (x + k) |} is the waveform of the eddy current flaw detection signal. It is an index showing the symmetry of the shape.

  FIG. 2 is a diagram showing an example of the evaluation function E (x) obtained from the output level value of the eddy current flaw detection signal derived from the defect and the eddy current flaw detection signal at the time of erroneous detection. As shown in FIG. 2, in this example, the eddy current flaw detection signal derived from the defect and the eddy current flaw detection signal at the time of false detection have the same waveform shape amplitude value (both are 2), but the symmetry of the waveform shape is different. It can be seen that there is a difference in the value of the evaluation function E (x). FIG. 3 is a diagram showing another example of the evaluation function E (x) obtained from the output level value of the eddy current flaw detection signal derived from the defect and the eddy current flaw detection signal at the time of erroneous detection. As shown in FIG. 3, in this example, the eddy current flaw detection signal derived from the defect and the eddy current flaw detection signal at the time of erroneous detection have different waveform shape amplitudes and symmetries, and thus the value of the evaluation function E (x) is You can see that there is a difference in. Therefore, it was confirmed that the value of the evaluation function E (x) can be used as the feature amount of the machine learning model in distinguishing the eddy current flaw detection signal derived from the defect and the eddy current flaw detection signal at the time of erroneous detection. Further, the inventors of the present invention used the output level value of the eddy current flaw detection signal at the time of defect detection or erroneous detection as another feature amount of the machine learning model.

Then, the value of the evaluation function E (x) obtained from the plurality of eddy current flaw detection signals, the maximum value of the output level value of the eddy current flaw detection signal at the time of defect detection or erroneous detection, and the determination value of the presence or absence of a defect (for example, defect The value of the evaluation function E (x) and the eddy current are obtained by performing logistic regression analysis using the data of the combination of the judgment value = 1 when the judgment is made and the judgment value = 0 when the judgment is noise. A machine learning model shown in the following mathematical expression (2) in which the maximum value of the output level value of the flaw detection signal is an input value, and the probability that the eddy current flaw detection signal corresponding to the input value is a signal derived from a defect (defect probability) is an output value. Was created as a defect judgment model. In other words, a combination of the value of the evaluation function E (x) obtained from a plurality of eddy current flaw detection signals, the maximum output level value of the eddy current flaw detection signal at the time of defect detection or erroneous detection, and the determination value of the presence or absence of a defect. By performing logistic regression analysis using the data of 1 as learning data, the coefficients a ₁ , a ₂ , and a ₃ of the machine learning model shown in Formula (2) were calculated.

Here, in Expression (2), q _i is the defect probability, x _i1 is the value of the evaluation function E (x), x _i2 is the maximum output level value of the eddy current flaw detection signal in the range of the captured waveform, and a ₁ , A ₂ and a ₃ indicate coefficients.

  Thus, with respect to the eddy current flaw detection signal detected by the eddy current flaw detection device, the value of the evaluation function E (x) and the maximum value of the output level value are input to the defect determination model to calculate the defect probability, and the calculated defect probability. By comparing with the threshold value, it is possible to accurately determine whether or not the eddy current flaw detection signal is a signal derived from a defect. It should be noted that the threshold value may be set to a value of the defect probability that allows the defect and the false detection to be clearly distinguished by increasing the number of samples (the number of data points) at the time of creating the defect determination model. Further, for example, when the test object is a steel sheet having a high Si content, if there is a defect in the steel sheet, fracture is likely to occur during subsequent rolling, so the threshold value is set to be safe in order to reliably detect the defect. Is desirable.

  Even when deep running or a neural network is used as the machine learning method, similarly to the above, the value of the evaluation function E (x) and the maximum value of the output level value of the eddy current flaw detection signal at the time of defect detection or error detection are set. It is also possible to use the learning data as an input and using the determination value of the presence or absence of a defect as an output. However, in the case of deep learning or the like, there is a degree of freedom by selecting the input value to the defect judgment model. Therefore, in the case of deep learning or the like, the symmetry of the waveform shape (the degree of signal polarity reversal) is evaluated in some form without using the value itself of the evaluation function E (x), and the evaluation result shows a defect. The learning data may be any learning data corresponding to the presence / absence determination value. Examples of the evaluation result include a digitized state of signal polarity reversal and a digitized state of signal amplitude.

  FIG. 4 is a diagram showing an example of determination results of defects and erroneous detections using the defect determination model shown in Expression (2). In this example, after a defect determination model is created by logistic regression analysis using 60 samples of defects and 20 samples of erroneous detection, eddy current flaw detection signals of unknown data (defects: 115 samples, erroneous detection: 20 samples) are detected as defects and False detection was determined. Further, in this example, the defect determination model outputs the probability that the input data is a defect (a value of 1 if the data is a defect, and a value that approaches 0 if the data is an erroneous detection). As shown in FIG. 6, in this example, the output value in the case of a defect and the output value in the case of a false detection can be clearly separated without overlapping, and by setting the threshold value to 0.4, the false detection can be performed. Good results were obtained that could suppress the occurrence of.

  Hereinafter, a defect detection apparatus and a defect determination method according to an embodiment of the present invention, which are conceived based on the above concept, will be described with reference to the drawings.

〔Constitution〕
First, with reference to FIG. 5 and FIG. 6, the configuration of a defect detection apparatus according to an embodiment of the present invention will be described.

  FIG. 5 is a block diagram showing the configuration of the defect detecting apparatus which is an embodiment of the present invention. FIG. 6 is a schematic diagram showing the configuration of the eddy current flaw detector 2 shown in FIG. As shown in FIG. 5, the defect detection device 1 according to the embodiment of the present invention includes an eddy current flaw detection device 2, a defect determination device 3, and a database 4 as main components.

  The eddy current flaw detection device 2 is configured by a self-comparison eddy current flaw detection device that applies an induced current to a subject made of a conductor to generate an eddy current and detects a defect existing on the surface of the subject by a change in the eddy current. It is arranged on the inlet side of a rolling line such as a cold rolling line. More specifically, as shown in FIG. 6, the eddy current flaw detector 2 includes an exciting coil (primary coil) P wound around a central leg of an E-shaped ferromagnetic body and both legs outside the central leg. Detection coils (secondary coils) S1 and S2 wound around the section are provided. An AC power supply (not shown) is connected to the terminals of the exciting coil P, and an eddy current is generated in the subject S by the magnetic force generated in the exciting coil P. On the other hand, the detection coils S1 and S2 are differentially connected in series. That is, one terminal of the detection coil S1 and one terminal of the detection coil S2 are connected, and the difference between the induced voltages induced in the detection coils S1 and S2 is detected between the other terminals.

  Therefore, when the surface layer of the subject is normal, the electromotive force induced between the other terminals of the detection coils S1 and S2 is zero. This is because when the surface layer of the subject is normal, the induced voltages generated in the detection coils S1 and S2 are the same and the induced voltages of the two differentially connected detection coils S1 and S2 cancel each other out. On the other hand, when there is a defect near one of the detection coils S1 and S2, the induced voltage generated between the detection coil S1 and the detection coil S2 differs due to the change of the eddy current on the subject. As a result, the difference between the induced voltages of the detection coils S1 and S2 is not zero, and an output is generated at both ends of the differentially connected detection coils S1 and S2. Therefore, by scanning the defective object with the eddy current flaw detection device 2, a sine curve eddy current flaw detection signal centered on the central axis of the E-shaped ferromagnetic material is detected.

  Returning to FIG. The defect determination device 3 includes a processor such as a CPU (Central Processing Unit), a DSP (Digital Signal Processor), and an FPGA (Field Programmable Gate Array), and a storage unit such as a RAM (Random Access Memory) and a ROM (Read Only Memory). , A workstation, or other general-purpose information processing device. The defect determination device 3 functions as the data extraction unit 3a, the prediction model creation unit 3b, and the defect determination unit 3c by the processor executing the computer program stored in the storage unit.

  The data extraction unit 3a extracts the output level value of the eddy current flaw detection signal output from the eddy current flaw detection device 2, and stores the data of the output level value of the extracted eddy current flaw detection signal in the database 4 as detection data.

  The prediction model creation unit 3b learns the above-described defect determination model by machine learning using the detection data stored in the database 4 and the information regarding the determination result of the defect or erroneous detection for each detection data. Although the learning of the defect determination model is performed offline, the defect determination model can be updated by appropriately relearning using newly accumulated data.

  The defect determination unit 3c uses the defect determination model created by the prediction model creation unit 3b to determine whether or not the eddy current flaw detection signal output from the eddy current flaw detection device 2 is a signal derived from a defect. It is determined whether or not there is a defect on the surface of the sample.

  The database 4 stores a plurality of pieces of detection data and the actual value of the defect determination result for each piece of detection data in association with each other.

[Defect determination processing]
Next, with reference to FIG. 7, a flow of the defect determination processing according to the embodiment of the present invention will be described.

  FIG. 7 is a flowchart showing the flow of the defect determination processing which is an embodiment of the present invention. The flowchart shown in FIG. 7 is started at the timing when the scanning of the subject by the eddy current flaw detection apparatus 2 is started, and the defect determination processing proceeds to the processing of step S1. In the present embodiment, it is assumed that the eddy current flaw detection apparatus 2 scans the subject by measuring the eddy current flaw detection signal from the subject conveyed on the manufacturing line with the eddy current flaw detection apparatus 2 fixedly arranged.

  In the process of step S1, the eddy current flaw detection device 2 measures the eddy current flaw detection signal and outputs the data of the output level value of the measured eddy current flaw detection signal to the defect determination device 3 as detection data. As a result, the process of step S1 is completed, and the defect determination process proceeds to step S2.

In the process of step S2, the data extraction unit 3a uses the detection data to detect the value x _{i1 of} the evaluation function E (x) and the output level value of the eddy current flaw detection signal measured by the eddy current flaw detector 2 in the process of step S1. The maximum value x _i2 is calculated. As a result, the process of step S2 is completed, and the defect determination process proceeds to the process of step S3.

In the process of step S3, the defect determination unit 3c _converts the value x _{i1 of} the evaluation function E (x) and the maximum value x _i2 of the output level values calculated in the process of step S2 into the defect determination model shown in Expression (2). The defect probability q _i is calculated by inputting. As a result, the process of step S3 is completed, and the defect determination process proceeds to step S4.

In the process of step S4, the defect determination unit 3c determines whether or not the defect probability q _i calculated in the process of step S3 is equal to or more than a predetermined threshold, and thus the eddy current flaw detection measured in the process of step S1. It is determined whether or not the signal is derived from a defect. Then, as a result of the determination, when the defect probability q _i is equal to or more than the predetermined threshold value (step S4: Yes), the defect determination unit 3c determines that the eddy current flaw detection signal is a signal derived from the defect, and performs the defect determination process. The process proceeds to step S5. On the other hand, when the defect probability q _i is less than the predetermined threshold value (step S4: No), the defect determination unit 3c determines that the eddy current flaw detection signal is a signal derived from noise (erroneous detection), and a series of The defect determination process ends.

  In the process of step S5, the defect determination unit 3c notifies the operator that the surface of the subject is defective. Then, the operator stops the transport line for the subject in response to the notification that the surface of the subject is defective. As a result, the process of step S5 is completed, and the defect determination process proceeds to step S6.

  In the process of step S6, the operator removes the defects formed on the surface of the subject, and restarts the transport of the subject after the removal of the defects is completed. According to the processes of steps S5 and S6, it is possible to prevent the test object from breaking due to rolling of the test object having a defect, so that the test object can be manufactured with high yield. As a result, the process of step S6 is completed, and the series of defect determination processes is completed.

  Although the embodiment to which the invention made by the present inventors has been described has been described above, the present invention is not limited by the description and the drawings forming part of the disclosure of the present invention according to the present embodiment. That is, all other embodiments, examples, operation techniques and the like made by those skilled in the art based on the present embodiment are included in the scope of the present invention.

DESCRIPTION OF SYMBOLS 1 Defect detection device 2 Eddy current flaw detection device 3 Defect determination device 3a Data extraction unit 3b Prediction model creation unit 3c Defect determination unit 4 Database

Claims (7)

In the defect determination method for determining the presence or absence of a defect on the surface of the subject based on the signal waveform obtained by changing the eddy current by giving an induced current to the subject made of a conductor,
From the subject with respect to the defect determination model machine-learned to determine whether the signal waveform is a signal waveform derived from the defect with the evaluation result of the state of polarity reversal in the signal waveform as a feature amount. A defect determination method comprising the step of determining the presence or absence of a defect on the surface of the subject by inputting a signal waveform.   The defect determination method according to claim 1, wherein the defect determination model is obtained by machine learning the evaluation result of the amplitude of the signal waveform as a feature amount.   The defect determination method according to claim 1, wherein the machine learning is any one of a logistic regression analysis, a decision tree, a neural network, and deep learning. In a defect determination device that determines the presence or absence of a defect on the surface of the subject based on the signal waveform obtained by changing the eddy current by giving an induced current to the subject made of a conductor,
From the subject with respect to a defect determination model machine-learned to determine whether or not the signal waveform is a signal waveform derived from the defect using the evaluation result of the state of polarity reversal in the signal waveform as a feature amount. A defect determination apparatus comprising means for determining the presence or absence of a defect on the surface of the subject by inputting a signal waveform.   A method of manufacturing a steel sheet, comprising: detecting a defect existing on the surface of the steel sheet using the defect determination method according to claim 1; and removing the detected defect. Method.   An induced current is applied to an object made of a conductor to generate an eddy current, and the evaluation result of the state of polarity reversal in the signal waveform obtained by the change of the eddy current and whether the signal waveform is a signal waveform derived from a defect or not. Defect as a learning data, using the signal waveform as an input value, and a signal value corresponding to the input value as a defect, the defect value is an output value. A method for learning a defect judgment model, comprising the step of learning the judgment model by machine learning.   A defect determination model, which is learned by the defect determination model learning method according to claim 6.

Images