JP3197193B2 - Two-dimensional signal filtering device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION The present invention relates to a method for filtering a two-dimensional signal if the two-dimensional signal cannot be filtered depending on the shape or size of the signal but can be filtered by utilizing the correlation of a plurality of signals. And a two-dimensional signal filtering device that performs

[0002]

2. Description of the Related Art For example, in a multi-frequency eddy current inspection, which is one method for non-destructively inspecting the state of a tube, when a signal is two-dimensional, a two-dimensional signal filtering device is used.
Conventionally, when filtering is performed on a plurality of two-dimensional signals using the correlation of the signals, the value of the position of the coordinates (x ₀ , y ₀ ) of the two-dimensional result signal is calculated by using each original signal. Are used at the same position on the two-dimensions, that is, only the value of (x ₀ , y ₀ ). If this method is expressed by the function of each device including mathematical formulas, the configuration is as shown in FIG.

First, a description will be given of a method of calculating a two-dimensional result signal when a filtering parameter has already been obtained. The result signal to determine r (x, y), z = z 1, ..., x N, y = y 1, ..., y
_M and the input two-dimensional original signals are represented as s ₁ (x, y),... S _n (x, y). Here, 1,..., N are suffixes for distinguishing a plurality of original signals. The original signal is read by the data reading device 1 as shown in FIG. The original signal read by the data reading device 1 is input to the teacher data creating device 3, the parameter calculating device 4, and the filtering device 5. The teacher signal data created by the teacher data creation device 3 is input to the parameter calculation device 4, and the parameters calculated by the parameter calculation device 4 are sent to the filtering device 5.

In the conventional method, to calculate the value of the position of a certain position (x ₀ , y ₀ ) of the two-dimensional result signal, the same position on the two-dimensional of each original signal, that is, (x ₀ , y ₀ ) , But at this time, the input signal matrix A has the original signal s ₁ (x,
y), ... from s _n (x, y)

[0005]

(Equation 1) It is expressed as Each column vector of the matrix is one of the multiple signals.
Are arranged side by side. Any order may be used as long as they are consistent. The components in each vector of the matrix are obtained by arranging the two-dimensional arrangement of the two-dimensional signals in a line according to some rule. Any order may be used as long as they are consistent. This conversion is necessary to complete the subsequent processing by matrix operation.

The filtering parameter is defined as ＾ W = (W
_1, W _2, ..., expressed as W _n), the filtered signal is calculated by filtering device 5 using the conversion signal A as follows.

＾ r = A · ＝ W ^t The filtering signal ＾ r thus obtained is a vector in which the xy coordinates in the two-dimensional image are arranged in the same order as the column vector of the signal matrix A. Note that ＾ W ^t is the transposed vector of ＾ W.

[0008] Finally, the data inverse conversion device 6 inversely converts the signal into a two-dimensional result signal similar to the original signal. The column direction of the signal matrix A is from (x ₁ , y ₁ ) in the image to (x _N ,
y _M ), the same applies to the column direction of Δr of the filtered signal. This is rearranged two-dimensionally in the opposite manner to the case where the signal matrix A is created. This allows
A final two-dimensional result signal is obtained.

Next, a process for obtaining a filtering parameter will be described. When an input signal with a clear signal generation factor and a teacher signal indicating what the signal should be like as a result of the processing are given, the filtering parameter is calculated by the parameter calculation device 4 as follows.

[0010] The original signal _{s 1 (x, y),} ... s n (x,
y) and the two-dimensional representation of the teacher signal is g (x, y), the filtering parameters are determined so that the error between the result signal and the teacher signal is minimized. That is, the parameter calculation device 4 obtains a parameter that satisfies the following equation.

[0011]

(Equation 2)

ΔW satisfying the above equation (3) can be obtained as follows by a multiple regression analysis which is a known technique. ＾ W ^t = [A ^t · A] ⁻¹ · A · ＾ g (4) where [·] ⁻¹ represents an inverse matrix. ＾ g is a signal obtained by converting g (x, y) of the two-dimensional expression of the teacher signal in the same manner as in the data converter.

[0013]

A conventional two-dimensional signal filtering apparatus uses only the same coordinate value of each of a plurality of input signals when determining a certain coordinate value of a two-dimensional signal. This is based on the following assumptions regarding the effects of the main signal generation factors (for example, target object in the case of an image and damage to a tube in the case of an eddy current inspection) that are to be filtered and retained, on the value of each coordinate of a plurality of input signals. Is equivalent to

(1) The influence range of the main signal generation factor on each of the plurality of original signals is not widened. That is, the signal generating factor in the coordinates (x _0, y _0), for all of the input signals, only affect the values of the coordinates (x _0, y _0), other _{(x 0 + 1, y 0} +1 ) Etc. are not affected.

(2) Therefore, the influence range of the main signal generation factor does not differ among a plurality of original signals. (3) Even when the correlation of the state of the influence range is different for each signal generation factor, the information is not used.

If the filtering parameters obtained based on these assumptions have a wide range of influence in the actual original signal, there is, of course, a risk that sufficient filtering accuracy may not be obtained. The present invention has been made to solve the above-described problem, and has as its object to provide a two-dimensional signal filtering device capable of obtaining high filtering accuracy.

[0017]

SUMMARY OF THE INVENTION The present invention provides a multi-frequency eddy current.
Two-dimensional signal filtering equipment used for flow inspection
And the second order of multi-frequency eddy current inspection for the inspection object
Reads the original signal data, a data reading device for outputting an input signal matrix, and a data converter for converting the format for operation in accordance with input signal matrix data read by the data reading device to the shape of the operation window, want to leave
The signal to be erased that was generated by a factor other than the main signal
A teacher data generating device for generating teacher signal data from a signal including a signal, and using the converted signal data converted by the data converter and the teacher signal data generated by the teacher data generating device, the filtering parameters are minimized. A parameter calculation device that calculates by the square method, and filtering is performed on the converted signal data converted by another or the data conversion device using the filtering parameter obtained by the parameter calculation device, and the main signal to be retained
A filtering device for extracting a signal generated due to an occurrence factor, and a data inverting device for restoring a filtering signal obtained by the filtering device to a two-dimensional result signal identical to the input signal data. And

(Operation) In the data conversion device, window conversion is performed in consideration of the range of influence. The present invention is an extension of the conventional method such that when a resultant signal value at a point at coordinates (x ₀ , y ₀ ) is obtained, the signal values in the vicinity thereof are also referred to by a window operation. become that way. It is assumed that the mathematical symbols used so far hold as they are. In contrast to the input signal matrix A of the conventional method, in the present invention, the following conversion signal matrix A '
Ask for. A ′ [M ₁ ,..., M _n ] (5) Here, M _i , i.

[0019]

(Equation 3)

The columns of the matrix M _i is determined by the shape and size of the window. If the window is a cross, the first column of M _{i is, s i (x 1 -δx,} y 1), s i (x 1, y 1), s i
(X ₁ + δx, y ₁ ), s _i (x ₁ , y ₁ −δy), s
_i (x ₁ , y ₁ + δy).

The method of calculating the composite signal is substantially the same as that of the equation (2), as follows: ＾ r = A ′ · ＾ W ^t (7)

The calculation of the weight is as follows. ＾ W ^t = [A ′ ^t · A ′] ⁻¹ · A ′ ^t · ＾ g (8) In the present invention, the amount of information of the filtering parameter is large by the size of the window, and the surrounding state is used. Therefore, when there is a correlation between the value of a certain place of the original signal and its surroundings, the correlation can be used, and the relationship between a plurality of original signals regarding the influence range of the signal generation factor can also be used.
Filtering accuracy can be improved.

[0023]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a configuration of a two-dimensional signal filtering device according to one embodiment of the present invention. The data reading device 1 reads two-dimensional signal data and outputs an input signal matrix to the data conversion device 2. The data conversion device 2 converts the input signal matrix data read by the data reading device into a calculation format according to the shape of the operation window. The data converted by the data conversion device 2 is transmitted to a teacher data creation device 3, a parameter calculation device 4, and a filtering device 5.
Is input to The teacher signal data created by the teacher data creation device 3 is input to the parameter calculation device 4. The parameter calculation device 4 includes the conversion signal data converted by the data conversion device 2 and the teacher data creation device 3.
The filtering parameters are calculated using the least squares method from the teacher signal data created in step (1), and output to the filtering device 5. The filtering device 5 performs filtering on the converted signal data obtained by another or converted by the data converting device 2 using the filtering parameter obtained by the parameter calculating device 4, calculates a filtered signal, and performs a data inverse conversion device 6. Output to This data inversion device 6 restores the filtered signal calculated by the filtering device 5 to the same two-dimensional result signal as the input signal data.

Next, the operation of the above embodiment will be described. FIG.
Shows the process diagram for obtaining the filtering parameters,
FIG. 3 shows a process diagram for obtaining a two-dimensional result signal.

Hereinafter, a specific problem of extracting a damaged portion of a tube from an image obtained by performing a multi-frequency inspection on the tube will be described as an example. In this example, the plurality of original signals are the real part and the imaginary part of the signal of the eddy current inspection using the plurality of flaw detection frequencies. The teacher signal data is the imaginary part (or the real part) of a signal at a specific flaw detection frequency.
The signal value due to the flaw is stored as it is, and the signal values of the other components (structures and deposits around the pipe) are set to zero. Here, specifically, the flaw detection frequency is set to 100 kHz and 4 kHz.
00 kHz, real part is X signal, imaginary part is Y
As a signal, the original signal is 100 kHz-X signal, 400 kHz
It is referred to as Hz-Y signal. As the teacher signal data, it is assumed that the signal other than the flaw signal is made zero from the 400 kHz-Y signal.

The data reading device 1 takes in the signal obtained by inspecting the tube as two-dimensional signal data for the real part and the imaginary part of each frequency, and creates an input signal matrix in which a plurality of original signals are arranged. FIG. 4 shows the state of the tube and the periphery of the tube represented by the signal, an example of the acquired two-dimensional original signal, and an example of the input signal matrix.

FIG. 4A shows a specific object.
In FIG. 1A, reference numeral 11 denotes a tube sheet. A tube 13 is inserted into a hole 12 provided in the tube sheet 11, and an expanded portion 14 of the tube 13 (tube) 13 is connected at the hole 12 portion. ing. Then, it is assumed that the tube 13 has a flaw 15 at the boundary with the expanded portion 14. FIG.
Ea is a signal of the tube 13, Eb is a signal of the expanding portion 14, and Ec is a signal of the flaw 15 portion. FIG.
(C) shows an input signal matrix, and the original signal 100 kHz-X
The signal, 100 kHz-Y signal, and 400 kHz-X signal are converted into an input signal matrix.

Then, the data data reading device 1
The input signal matrix data read by
To convert the data into a calculation format corresponding to the shape of the operation window. Transform the input signal matrix. At this time, a conversion signal corresponding to the window shape is created. As an example of the window, a case where the window shown in FIG. 5 is used for all original signals will be described. In general, since a window can be defined for each original signal, the size and shape may be different for each signal.

The center of the window shown in FIG. 5 is (x ₀ , y
₀ ), (x ₀ −1, y ₀ ), (x ₀ , y ₀ ), (x ₀ +1, y
₀ ), (x ₀ , y ₀ -1), and (x ₀ , y ₀ +1). The converted signal A 'is as follows: A' = [M _{100 kHz-X} , M _{100 kHz-Y} , M _{400 kHz-X} , M _{400 kHz-Y} ] (9) Here, M _{100 kHz-X} is a matrix relating to the real part of which flaw detection frequency is 100 kHz in the input signal. Specifically, in the window of FIG. 5, the M _{100 kHz-X} signal is as follows.

[0030]

(Equation 4)

Although s (x, y) is a signal which should be described as s _{100 kHz-X} , it is abbreviated here for convenience of description space. Each column vector of the matrix is obtained by arranging each of a plurality of signals. Any order may be used as long as they are consistent.

Next, teacher data is created by the teacher data creating device 3. As a specific example of the teacher signal, of the two-dimensional original signal, a signal corresponding to a flaw from a Y signal having a flaw detection frequency of 400 kHz is left as it is, and a part corresponding to a factor other than the flaw is replaced with a signal value of zero. The signal is obtained by processing. As a process of creating the teacher signal data, first, a target part to be stored and a signal component to be removed are designated, and a signal g to be obtained as a two-dimensional signal is specified.
(X, y) is created, and then the two-dimensional signal g (x, y) is converted in the same manner as the data conversion device 2 performs. In this example, since the teacher signal data is the imaginary part (400 kHz-Y signal) of the signal of the specific flaw detection frequency, the conversion result is a column vector with one column. This is represented by Δg.

[0033]

(Equation 5)

The signal shown in the equation (11) is teacher signal data. The arrangement rule is matched (matched) with the data conversion device 2. Next, the filtering parameter ＾ W is obtained by the parameter calculation device 4. The filtering parameter ＾ W can be obtained by the following equation: ＾ W ^t = [A ′ ^t · A ′] ⁻¹ · A ′ ^t · ＾ g (12)

Further, the filtering device 5 performs filtering on the transformation matrix signal obtained by transforming the original signal. In this case, filtering means that a signal component having a signal value of each flaw detection frequency signal and a signal component having a similar correlation in a window and a flaw signal designated to be stored when generating the teacher signal data are stored, and the filtering is performed when the teacher signal data is generated. That is, a linear operation is performed on the converted signal of the input signal so as to eliminate a signal component similar to a factor signal other than a flaw designated to be erased. Specifically, the following linear operation is performed.

＾ r = A ′ · ＾ W ^t (13) Here, A ′ may be the converted signal data itself used for calculating the filtering parameter, or may be separately generated from the two-dimensional original signal. There is also. The main purpose of the filtering is to save a signal that is close to a signal to be stored (flaw signal) taught by the teacher signal data and delete the signal with the teacher signal data, for a signal including a factor signal that is not known as a flaw. The purpose is to evaluate the soundness of the tube by eliminating signals close to the signals (structures and the like) taught as described above, and extracting only the flaw signal components from the signals of unknown factors. Therefore, in many cases, A 'here is a signal different from the converted signal data used for calculating the filtering parameter.

However, in order to confirm the accuracy of the filtering, the converted signal data itself used for calculating the filtering parameters is used, and the teacher signal data is compared with the result signal to determine how close the filtering is to the teacher signal data. In some cases.

The filtering signal obtained here is 2
Unlike the two-dimensional state of the dimensional original signal, it corresponds to the arrangement of signal values in the converted signal. Therefore, the data inverse conversion device 6 inversely converts the filtered signal into a sequence corresponding to the input signal.

FIG. 6 shows an example of the two-dimensional original signal read by the data reading device 1, and FIG.
This shows a two-dimensional signal of the teacher signal after specifying a location to be saved and a location to be erased in the processing course of the teacher data creation device 3 using the 00 kHz-Y signal.

The present invention relates to an apparatus for filtering a two-dimensional signal when the two-dimensional signal cannot be filtered depending on the signal shape or size but can be filtered by utilizing the correlation of a plurality of signals. Applicable. As a specific example, in a multi-frequency eddy current inspection, which is one method for non-destructively inspecting the state of a tube, the present invention can be applied to a case where a signal is two-dimensional, for example, a case of flaw detection using a rotating probe sensor. Note that a one-dimensional signal is a special case of a two-dimensional signal (having a vertical or horizontal size of 1), and therefore is applicable although the shape of the operation window is restricted.

[0041]

As described above in detail, according to the present invention, since the amount of information of the filtering parameter is large due to the size of the window and the surrounding state is used, the value of the place where the original signal exists and its value When there is a correlation with the surroundings, the correlation can be used, and the relationship between a plurality of original signals related to the influence range of the signal generation factor can also be used, so that the filtering accuracy can be improved.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a two-dimensional signal filtering device according to an embodiment of the present invention.

FIG. 2 is a process diagram for obtaining a filtering parameter in the embodiment.

FIG. 3 is a process diagram for obtaining a two-dimensional result signal in the embodiment.

FIG. 4 is an exemplary view showing a specific object, a two-dimensional original signal, and an input signal matrix in the embodiment.

FIG. 5 is an explanatory diagram of an operation window in the embodiment.

FIG. 6 is an exemplary view showing an example of a two-dimensional original signal according to the embodiment;

FIG. 7 is an exemplary view showing a two-dimensional representation of a teacher signal in the embodiment.

FIG. 8 is a block diagram showing a configuration of a conventional two-dimensional signal filtering device.

[Description of Signs] 1 Data reading device 2 Data conversion device 3 Teacher data creation device 4 Parameter calculation device 5 Filtering device 6 Data inversion device 11 Tube plate 12 Hole provided in tube plate 13 Tube 14 Tube expansion unit 15 Wound

────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP-A-5-240840 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G01N 27/72-27/90

Claims (1)

(57) [Claims] 1. A secondary used for multi-frequency eddy current inspection.
The original signal filtering device reads the two-dimensional signal data of the multi-frequency eddy current test for the inspection object and outputs an input signal matrix. The input signal matrix data read by the data reading device is shaped into an operation window. A data conversion device that converts the data into an arithmetic format according to the data, and the erasure caused by factors other than the main signal
A teacher data generation device that generates teacher signal data from a signal that also includes a signal that the user wants to perform; and using the conversion signal data converted by the data conversion device and the teacher signal data generated by the teacher data generation device, a filtering parameter is set. A parameter calculator for calculating by the least squares method, and performing filtering on the converted signal data converted by another or the data converter using the filtering parameter obtained by the parameter calculator;
A filtering device for extracting a signal generated by a main signal generation factor to be retained; and a data inverting device for restoring the filtered signal obtained by the filtering device to a two-dimensional result signal identical to the input signal data. A two-dimensional signal filtering device, characterized in that: