JPH10115604A - Ultrasonic flaw detection evaluating device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To judge the presence or absence of the flaws in a specimen to be inspected efficiently. SOLUTION: Analysis data 16 obtained in a data analysis part 15 and judged data 24 obtained in a judging part 17 are stored in a object-oriented database 27 as data in which design conditions of a pipe 1 and the set value of a flow detection wave are attributed. Then, the data having the attribute in correspondence with the design conditions of a pipe 1, wherein the presence or absence of the flow is to be judged, and the value of the flaw wave is outputted to a data input part 11 as instructing data 18 among the many data stored in the object-oriented-type database 27. The learning in correspondence with the pipe 1, wherein the presence or absence of the flow is to be judged, is performed in a neural network 17, and the presence or absence of the flaw in the pipe 1 is judged.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an ultrasonic inspection data evaluation apparatus.

[0002]

2. Description of the Related Art In various plants and the like, ultrasonic flaw inspection is widely performed as a means for performing nondestructive inspection of structures and welds.

In this case, it is an inspector having specialized knowledge that evaluates the ultrasonic inspection data obtained by the ultrasonic inspection and determines whether or not there is a defect. The judgment requires skill and there is a possibility that the judgment result by each inspector may vary.

[0004] It is a very difficult task to make a judgment by looking at all the tens or hundreds of ultrasonic inspection data obtained by the inspection.

[0005] Therefore, research and development for automatically judging ultrasonic flaw detection data are being advanced.

FIG. 2 shows an example of an ultrasonic flaw detection data evaluation apparatus which is currently under development. This ultrasonic flaw detection data evaluation apparatus uses a welded joint in a pipe (inspection object) 1 to be subjected to non-destructive inspection. An ultrasonic flaw detector 3 for performing ultrasonic flaw detection of the portion 2 and a data evaluation apparatus main body 7 for determining the presence or absence of a defect based on flaw detection data 6 obtained by the ultrasonic flaw detector 3
And

[0007] The ultrasonic flaw detector 3 includes an ultrasonic probe 5 that can abut on the outer peripheral portion of the tube 1, a flaw-detecting wave that propagates to the ultrasonic probe 5, and a reflected wave from the tube 1. And an ultrasonic transmission / reception unit 4 for outputting flaw detection data 6 in accordance therewith.

An input device 8 such as a keyboard, a pointing device 9 such as a mouse or a tablet, and a display device 10 such as a display are connected to the data evaluation device body 7.

The data evaluation device main body 7 extracts a part of the flaw detection data 6 by a data input unit 11 for receiving the flaw detection data 6 output from the ultrasonic transmission / reception unit 4 and a position instruction input 12 from the pointing device 9. A time gate setting unit 13, a data analysis processing unit 15 for analyzing the extracted data 14 extracted by the time gate setting unit 13 by Fourier transform, and a determination based on the analysis data 16 analyzed by the data analysis processing unit 15. And a neural network (determination unit) 17 using a neural network or the like.

The data input section 11 has teaching data 1 for causing the neural network 17 to perform learning.
8 is transmitted from the input device 8.

The neural network 17 comprises an input layer 20, an intermediate layer 21, and an output layer 22 as shown in FIG. 9 by the neurons 19 shown in FIG. b, c, d, e
The lattice point 23 formed by the vertices corresponds to the neuron 19 of the input layer 20.

When performing a nondestructive inspection of the welded joint 2 of the pipe 1, as shown in FIGS. 2 and 3, an ultrasonic probe 3 A probe 5 is arranged, and the flaw detection wave is propagated from the ultrasonic transmission / reception unit 4 to the welded joint portion 2 of the pipe 1 via the ultrasonic probe 5, and the reflection transmitted from the welded joint portion 2 to the inside of the pipe 1. The waves are received by the ultrasonic transmission / reception unit 4 as flaw detection data 6.

The above operation is performed by moving the ultrasonic probe 5 in a plurality of positions (a, b, c, d, e in FIG. 2) while gradually moving the ultrasonic probe 5 in the axial direction (a) of the tube 1. 5 places).

As a result, a plurality of flaw detection data 6 (see FIG. 4) can be obtained at different positions in one circumferential direction (b) of the welded joint portion 2.

Further, the ultrasonic probe 5 is moved little by little in the circumferential direction (b) of the tube 1 while repeating the above operation.
Flaw detection data 6 is collected over the entire circumferential direction (b) of the welded joint portion 2.

In this manner, when the collection of the flaw detection data 6 by the ultrasonic flaw detection apparatus 3 is completed, the flaw detection data 6
Is input to the data input unit 11 of the data evaluation device main body 7.

The flaw detection data 6 is input to the data evaluation device main body 7 by connecting the ultrasonic transmission / reception unit 4 and the data evaluation device main body 7 via a cable or a network to transmit electric power, or both have infrared ports. In this case, the data may be transferred by infrared communication, or data may be input via a removable medium such as a magnetic tape, a magnetic disk, or an optical disk.

When the flaw detection data 6 is input to the data input unit 11, the time gate setting unit 1 of the data evaluation device main unit 7
3 displays the flaw detection data 6 on a display device 10 such as a display, and extracts a part of the flaw detection data 6 to the operator using a pointing device 9 such as a mouse or a tablet as shown in FIG. Gate T).

A specific manner of setting the time gate T by the operator is performed by designating the vicinity of the peak value of the flaw detection data 6 as the center position T0 and designating a range around the center position T0 with an appropriate width. (Because the range is an object with respect to the center position, you can specify either the front or the rear.)

When the setting of the time gate T is completed, the data analysis processing section 15 of the data evaluation device main body 7 performs the analysis processing of the extracted data 14 of the flaw detection data 6.

A specific method of the analysis processing is an integral conversion method using a trigonometric function called a fast Fourier transform.
The frequency analysis of the extracted data 14 is performed, and two-dimensional data indicating the energy intensity of each frequency of the extracted data 14 is obtained.

At this time, since the frequency analysis is performed on the entire region of the extracted data 14, the concept of time is lost from the two-dimensional data.

Also, the two-dimensional data is a, b,
The positions of c, d, and e are superimposed to form pseudo three-dimensional data as shown in FIGS.

FIG. 5 shows typical analysis data 16 when there is stress corrosion cracking, and FIG. 6 shows typical analysis data 16 when there is no stress corrosion cracking.

The two waveforms shown in FIG. 5 and FIG. 6 are recognized to be clearly different from human eyes, but it is very difficult to make the same recognition by engineering. is there.

Further, the data analysis processing unit 15 normalizes the energy intensity in the analysis data 16 so that the energy intensity is represented by a value between 0 and 1 for the convenience of processing at a later stage.

Based on the analysis data 16 thus obtained, the data evaluation device main body 7 determines the analysis data 16 by the neural network 17.

Here, the neural network 17 has been developed for engineeringly realizing the excellent information processing capability of the human brain.
9 are constructed.

The neuron 19 is, as shown in FIG.
Input Si (0 ≦ Si ≦ 1, i = 1 to n) and weight coefficient W
The internal state u is represented by i (excitation when Wi ≧ 0, suppression when Wi <0, i = 1 to n) and the threshold θ.
Is defined by u = ΣWi · Si-θ, and this internal state u is defined as an output function f (x) = 1 / ｛1 + exp (−x /
t) By substituting into｝, the following output is obtained.

[0030]

F (u) = 1 / {1 + exp (-u / t)}

Note that t is a positive constant (usually 1.0) called the temperature of the neuron 19.

The output function f (x) is called a sigmoid function. As shown in FIG. 8, while the value of x changes from-to +, the value becomes non-linear from 0 to 1. It is a function that changes (0 ≦ f (x) ≦ 1).

The neural network 17 includes various types such as an interconnected type and a hierarchical type, and the hierarchical type is applied to the ultrasonic flaw detection data evaluation device currently under development.

The hierarchical neural network 17 is
As shown in FIG. 9, a neuron 19 is connected so as to have a hierarchy of an input layer 20, an intermediate layer 21, an output layer 22, and the like. Each layer includes one or more neurons 19, The neurons 19 in the same layer are not connected to each other and each neuron 1 in the lower layer
9 is connected to all the neurons 19 in the upper layer.

The input layer 20 stores the analysis data 1
6 and the same number of neurons 19 as grid points 23 set on coordinates formed by the frequencies a and b, c, d, and e.
The intermediate layer 21 is composed of one or more layers of neurons 19 (one layer in FIG. 9), and the output layer 22 is composed of two neurons 19a and 19b.
It is composed of

Further, for each neuron 19,
The learning to the neural network 17 is performed by setting parameters such as the weight coefficient Wi and the threshold value θ.

As a learning method, first, each parameter such as the weight coefficient Wi and the threshold value θ is input to the input device 8 such as a keyboard.
Neural network 1 with random settings using
7, the teaching data 18 is input thereto to generate an output, an error between the output from the neural network 17 and the actual result of the teaching data 18 is calculated, and the error is calculated from the output layer 22 to the intermediate layer. 21, the intermediate layer 21 to the input layer 2
The work is repeated for a large number of teaching data 18 using an error back propagation learning method in which parameters are adjusted and changed for each layer while being propagated back to zero.

At this time, the two neurons 1 of the output layer 22
If the outputs of 9a and 19b are defective, (1.0 to 0.7,
0.0-0.3), if unknown (0.7-0.3, 0.
3 to 0.7), if it is not a defect (0.3 to 0.0, 1..
0 to 0.7).

In the neural network 17 sufficiently learned in this way, the input of the energy intensity at the corresponding grid point 23 to each neuron 19 constituting the input layer 20 is only required, and the output of each neuron 19 is obtained. Are calculated one after another, and a correct judgment result is automatically derived.

As described above, the neural network 17
By using, unlike fuzzy or AI, it is possible to make a correct determination even from a waveform pattern that is difficult to be ruled.

Since the learning contents are distributed and stored in each parameter, the neural network 17
Although a black box itself is formed, by performing more learning, a more accurate determination result can be expected.

The judgment result obtained in this way is also learned by the neural network 17.

The determination result obtained in this way is transmitted to the display device 10 as the determination data 24, and the determination result of the ultrasonic flaw detection by the data evaluation device main body 7 is displayed on the display device 10.

[0044]

On the other hand, the flaw detection data 6 output from the ultrasonic transmission / reception unit 4 of the ultrasonic flaw detection apparatus 3 to the main body of the data evaluation apparatus 7 is based on the design conditions of the tube 1 to be inspected. It varies depending on the set value of the shape, tube diameter, wall thickness, material) and flaw detection wave.

That is, in the ultrasonic flaw detection data evaluation apparatus using the neural network 17, when determining whether or not there is a defect in the welded joint portion of the pipe 1, the flaw detection data 6 for which the flaw is to be determined is determined. In accordance with the obtained design conditions of the pipe 1 and the set values of the flaw detection wave, the evaluated flaw detection data 6 stored in a relational database (not shown) provided separately from the ultrasonic flaw detection data evaluation apparatus. ,
By extracting teaching data 18 having the same design conditions and flaw detection wave settings as the teaching data 18 and inputting the taught data 18 to the data input unit 11, the design conditions and flaw detection wave setting values are met. Learning with neural network 1
7 needs to be done.

However, the relational database applied to the ultrasonic inspection data evaluation device shown in FIG. 2 stores information in a table (rows and columns of data) and stores data in a specified column of a given table. Since the database is used for searching by searching for data in another table, it is possible to quickly select the teaching data 18 from the huge number of evaluated flaw detection data 6 accumulated in the relational database. Therefore, it is not possible to efficiently determine the presence or absence of a defect in the welded joint portions 2 of many pipes 1.

The present invention has been made in view of the above circumstances, and has as its object to provide an ultrasonic inspection data evaluation apparatus capable of efficiently determining the presence or absence of a defect in an inspection object. .

[0048]

In order to achieve the above object, in an ultrasonic flaw detection data evaluation apparatus according to the present invention, a data input section for receiving flaw detection data obtained from an inspection object and a time gate for the flaw detection data are provided. Uses a time gate setting unit that multiplies and extracts extracted data, a data analysis processing unit that analyzes the extracted data by Fourier transform, and a neural network that determines the presence or absence of a defect based on the analysis data obtained by the data analysis processing unit Judgment part
The analysis data obtained by the data analysis processing unit and the judgment data from the judgment unit are stored as data with the design conditions of the inspection object and the set value of the flaw detection wave as attributes, and the presence or absence of a defect is judged from the stored data. An object-oriented database for outputting to the data input unit, as teaching data, attributes corresponding to the design conditions of the inspection object from which the flaw detection data to be obtained and the set value of the flaw detection wave are obtained.

In the ultrasonic flaw detection data evaluation apparatus of the present invention,
The analysis data obtained by the data analysis processing unit and the judgment data from the judgment unit, that is, the presence / absence of a defect in the object to be inspected, as object-oriented data as attributes of the object to be inspected, the design conditions and the set value of the flaw detection wave From the large number of data stored in the database and stored in the object-oriented database, teach what has attributes according to the design conditions of the inspection object to determine the presence or absence of a defect and the set value of the flaw detection wave. The data is output as data to the data input unit, and learning is performed in the determination unit according to the inspection object to determine whether there is a defect.

[0050]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 shows an embodiment of an ultrasonic flaw detection data evaluation apparatus according to the present invention. In FIG. 1, an ultrasonic flaw detection apparatus 3 for flaw detection of a welded joint portion 2 of a pipe 1 and flaw detection data 6 are evaluated. The data evaluation device main body 7, the pointing device 9 and the display device 10 connected to the data evaluation device main body 7 are the same as those shown in FIG. 2, and other portions denoted by the same reference numerals as those in FIG. Represents an object.

Reference numeral 25 denotes an input device such as a keyboard. The input device 25 is used for setting the design conditions (shape, tube diameter, wall thickness, material, and the like) of the tube 1 set by manual operation, and setting of flaw detection waves. The value is configured to be output as the attribute setting command 26.

Reference numeral 27 denotes an object-oriented database. The object-oriented database 27 is an analysis data which has been analyzed by the data analysis processing unit 15 of the data evaluation device body 7 based on an attribute setting command 26 from the input device 25. 16 and the determination data 2 determined by the determination unit 17
Based on the attribute setting command 26 from the input device 25, the defect is selected from the stored data so as to store the design conditions 4 and the set value of the flaw detection wave as attributes. The design condition of the tube 1 from which the flaw detection data 6 for which the presence / absence of the flaw detection data 6 should be determined and the attribute corresponding to the set value of the flaw detection wave are output as teaching data 18 to the data input unit 11 of the data evaluation device main body 7. Is configured.

Hereinafter, the operation of the ultrasonic inspection data evaluation apparatus shown in FIG. 1 will be described.

When performing a nondestructive inspection of the welded joint 2 of the pipe 1, the ultrasonic probe 5 of the ultrasonic flaw detector 3 is arranged near the welded joint 2 of the pipe 1, and an ultrasonic transmitting and receiving unit is provided. 4 from the welded joint portion 2 of the pipe 1 through the ultrasonic probe 5
The reflected wave transmitted from the welded joint portion 2 to the inside of the pipe 1 is received by the ultrasonic transmission / reception section 4 as flaw detection data 6.

When the collection of the flaw detection data 6 by the ultrasonic flaw detection apparatus 3 is completed, the flaw detection data 6 is input to the data input unit 11 of the data evaluation apparatus main body 7.

The design conditions (shape, tube diameter, wall thickness, material) of the tube 1 to be inspected and the set values of the flaw detection wave are input to the input device 2.
5 and input conditions from the input device 25 to the design conditions of the pipe 1;
The set value of the flaw detection wave is output to the object-oriented database 27 as an attribute setting command 26.

When the flaw detection data 6 is input to the data input unit 11, the time gate setting unit 1 of the data evaluation device main body 7
3 displays the flaw detection data 6 on a display device 10 such as a display, and extracts a part of the flaw detection data 6 to the operator using a pointing device 9 such as a mouse or a tablet as shown in FIG. Gate T).

When the setting of the time gate T is completed, the data analysis processing section 15 of the data evaluation device main body 7 performs the analysis processing of the extracted data 14 of the flaw detection data 6.

The analysis data 16 obtained by the above-described analysis processing in the data analysis processing unit 15 is transmitted to the neural network 17 and the object-oriented database 27.
Is output to both.

In the neural network 17,
The determination of the analysis data 16 is performed, and the obtained determination result is transmitted as the determination data 24 to the display device 10 and the object-oriented database 27, and the display device 1
The result of the ultrasonic flaw detection by the data evaluation device main body 7 is displayed at 0.

On the other hand, the object-oriented database 2
In step 7, data is added to the analysis data 16 and the determination data 24 (presence / absence of a defect) to which attributes such as the design condition of the tube 1 as the inspection object and the set value of the flaw detection wave are added.

This object-oriented database 27
Are used as teaching data 18 for the data evaluation device body 7.

That is, as described above, the input device 2
5, when the design conditions of the pipe 1 for which the presence or absence of a defect is to be determined and the set value of the flaw detection wave are output to the object-oriented database 27 as the attribute setting command 26, a large number of data stored in the object-oriented database 27 are obtained. Among them, those having attributes corresponding to the design conditions of the tube 1 to be inspected and the set values of the flaw detection wave are output to the data input unit 11 of the data evaluation device main body 7 as teaching data 18, and the above-described design is performed. Learning according to the conditions and the set value of the flaw detection wave is performed in the neural network 17.

As described above, in the ultrasonic inspection data evaluation apparatus shown in FIG. 1, the analysis data 16 obtained by the analysis processing in the data analysis processing unit 15 and the determination data 24 obtained by the neural network 17 The design conditions are stored in the object-oriented database 27 as data having the set value of the flaw detection wave as an attribute, and the design conditions of the tube 1 to be inspected are selected from among a large number of data stored in the object-oriented database 27. Since the data having the attribute corresponding to the set value of the flaw detection wave is output to the data input unit 11 of the data evaluation device main body 7 as the teaching data 18, the learning to the neural network 17 can be performed appropriately and promptly. It is possible to efficiently determine the presence or absence of a defect in the welded joint portions 2 of many pipes 1.

The ultrasonic flaw detection data evaluation apparatus of the present invention is not limited to the above-described embodiment, but may be variously modified without departing from the gist of the present invention. .

[0067]

As described above, in the ultrasonic inspection data evaluation apparatus of the present invention, the analysis data obtained by the data analysis processing unit and the presence / absence of a defect in the inspection object determined by the determination unit are determined. Inspections that should be stored in the object-oriented database as data with the design conditions of the inspection object and the set value of the flaw detection wave as attributes, and for which the presence or absence of a defect should be determined from a large number of data stored in the object-oriented database The object having the attribute corresponding to the design condition of the object and the set value of the flaw detection wave is output to the flaw detection data evaluation device data input unit as teaching data, so that learning according to the inspected object to determine whether or not there is a defect is determined. In such a case, it is possible to achieve an excellent effect that the inspection can be performed quickly and the presence or absence of a defect in the inspection object can be efficiently determined.

[Brief description of the drawings]

FIG. 1 is a schematic system diagram showing an example of an embodiment of an ultrasonic inspection data evaluation device of the present invention.

FIG. 2 is a schematic system diagram showing an ultrasonic flaw detection data evaluation device currently being developed.

FIG. 3 is a partial cross-sectional view showing a relationship between a tube and an ultrasonic probe.

FIG. 4 is a graph showing an example of ultrasonic flaw detection data to determine the presence or absence of a defect.

FIG. 5 is a diagram showing typical analysis data when there is stress corrosion cracking.

FIG. 6 is a diagram showing typical analysis data when there is no stress corrosion cracking.

FIG. 7 is a diagram showing a model of a neuron.

FIG. 8 is a graph showing a sigmoid function.

FIG. 9 is a system diagram showing a neural network part.

[Explanation of symbols]

 1 tube (inspection object) 6 flaw detection data 11 data input unit 13 time gate setting unit 14 extracted data 15 data analysis processing unit 16 analysis data 17 neural network (judgment unit) 18 teaching data 24 judgment data 27 object-oriented database

Claims (1)

[Claims] 1. A data input unit for receiving flaw detection data obtained from an inspection object, a time gate setting unit for multiplying flaw detection data by a time gate to extract extracted data, and data for analyzing and processing the extracted data by Fourier transform An analysis processing unit, a determination unit using a neural network that determines the presence or absence of a defect based on the analysis data obtained by the data analysis processing unit, and analysis data obtained by the data analysis processing unit and determination data from the determination unit. The design conditions of the inspection object and the setting of the inspection wave, which are stored as data having the set value of the inspection wave as an attribute and from which the inspection data to determine the presence or absence of a defect are obtained. An object-oriented database for outputting, to the data input unit, data having an attribute corresponding to a value as teaching data. Wave inspection data evaluation device.