JPH05256632A - Damage inspecting method for metal member and damage inspecting device - Google Patents

Abstract

PURPOSE:To provide a damage inspecting device of a metal member capable of precisely inspecting the damage of the metal member in a short time, easily controlling the obtained damage data, and automatically and precisely judging the residual life of the metal member based on the degree of the damage. CONSTITUTION:A metal member (shaft) 1 is set on an inspecting device, an optical microscope 11 is moved to an inspection portion, and the image of the optical microscope 11 is picked up by a video camera 12 and processed by an image processing device 16. A computer 20 judges whether cracks are generated or not on the shaft 1 as damage based on the image-processed data. When the computer 20 recognizes the cracks, it confirms the shape of the cracks via the connection processing or integration processing with other cracks. The computer 20 analyzes the distribution of the confirmed cracks, determines the maximum length of the cracks, and judges the residual life of the shaft 1 based on a master curve of crack progress.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a damage inspection method and damage inspection apparatus for metal members, and more particularly, it processes a microscope image of a rotating body such as a boiler feed pump shaft of a thermal power plant,
A device that determines corrosion fatigue cracks and corrosion pits on the surface of a rotating body from the obtained luminance distribution, finds the maximum length of the distribution of corrosion fatigue cracks, and automatically diagnoses the remaining life of metal members such as rotating bodies. Regarding

[0002]

2. Description of the Related Art There are high-speed pumps for power plants and the like as mechanical structures that are subjected to repeated loads under a high temperature atmosphere.
In such a high-speed pump, a mechanical structure that constitutes the pump is subjected to thermal fatigue and simple fatigue damage at the same time due to repeated start and stop or load fluctuation. As a result, fatigue and damage due to aging of materials accumulate,
It causes cracks in parts and reduces structural strength.

[0003] In such a case, if the remaining mechanical strength as a mechanical structure is not evaluated and properly managed, damage to the equipment or a large accident in the plant may result. In particular, the shaft of a high-speed pump is in an important position in the equipment configuration, so the remaining life in terms of strength must be evaluated correctly. Therefore, it is required to inspect whether the shaft has a crack as a damage.

Various techniques for inspecting mechanical structures have been proposed and put to practical use. As a general nondestructive inspection method, an inspector visually confirms a crack on the surface of a metal material, a dye flaw detection method (PT), a magnetic particle flaw detection method (M
The inspection methods such as T) are mainly used to determine the presence or absence of damage such as cracks. The dye flaw detection method (PT) is a method in which a dyeing solution is sprayed onto the surface of the shaft to be inspected, the surface of the shaft is wiped off, and the dyeing solution soaked in the cracked part of the shaft is checked to check for damage such as cracks. It is a method of recognition.
The magnetic particle flaw detection (MT) is a magnetic flux disturbance when a magnetized shaft has a damaged portion such as a crack. Therefore, a small magnetic pole is attracted to the damaged portion and fluorescent light is applied to the damaged portion for easy viewing. Is a method of recognizing by shining. As these nondestructive inspection methods, a method in which an ultrasonic wave or X-ray photograph and an image processing device are combined is also known.

However, in the dye flaw detection and the magnetic particle flaw detection, which are general nondestructive inspection methods, the inspector visually confirms the presence or absence of damage and the size of the flaw. It was difficult to accurately determine the presence or absence of significant damage and the extent of damage.

[0006]

DISCLOSURE OF THE INVENTION Problems to be Solved by the Invention
Japanese Patent No. 149970 proposes a damage detection method in which a cross section or a polished surface of a sample taken from a high temperature pressure resistant member is observed with a microscope. Also in this conventional example, since the inspector visually confirms the presence or absence of damage and the size thereof, it is difficult for even a skilled inspector to accurately determine the presence or absence of minute damage and the degree of damage. It was

Japanese Unexamined Patent Publication (Kokai) No. 1-187454 briefly describes a method of capturing a microscopic image of a metal sample with a television camera and analyzing the image with an image analyzer. But,
No substantial analysis method is described.

"Material Testing Technology" Vol.1, No1 (January 1989)
"Image processing technology for surface microcrack detection in material strength" discusses detection of crack propagation and fatigue behavior by an optical microscope, a video camera, and an image processing device.

Japanese Unexamined Patent Publication No. 61-151411 proposes a method for evaluating defects based on the dot density in a binary image obtained as a result of imaging. In this conventional example, it can be used even at a position where the dot density is low. The fact that defects such as cracks exist continuously is not considered. Therefore, the length and size of the crack may be underestimated.

Japanese Unexamined Patent Publication (Kokai) No. 2-208563 proposes a steel material evaluation apparatus for calculating and determining the precipitation position and variation of point inclusions from a virtual grain boundary. When the point sequence extraction mechanism of point inclusions in this determination method is applied to the inspection of crack damage, the characteristics of the crack are not taken into consideration, so it is not possible to determine how far the crack extends continuously. It was difficult.

Japanese Patent Application No. 59-260759 discloses a technique for evaluating damage in a boiler or turbine of a power generation facility.
Using a video camera and image processing means, microscopic cracks in the creep fatigue region are sampled over a wide area at once, and the maximum crack length is estimated by the extreme value statistical method.

[0012] In this conventional example, the remaining crack life can be estimated by using the extreme value statistical method or the like, so that the reliability is low in practical use. Moreover, not only does it take time to collect data on shaft damage, process data, and analyze data,
It was difficult to grasp and accumulate quantitative data.

Japanese Unexamined Patent Publication No. 61-201105 proposes a method for evaluating the surface condition of a coating layer or the like based on the area of a shadow in a binary image.

Generally, when the damage is a crack, the crack may further grow as fatigue accumulates. Therefore, replacement of the metal member should be considered.
If the damage is a corrosion pit, the metal member can be continuously used by scraping off the corrosion pit with a grinder or the like. When this conventional example is applied to a damage inspection of a metal member, it is impossible to distinguish between a crack defect and a corrosion pit, and it is necessary to use other evaluation methods together.

Japanese Patent Laid-Open Nos. 62-53791 and 2-269967 propose a method of detecting the number of pixels from a digital image to detect microorganisms and cells.

When these methods for detecting living microorganisms and cells are applied to the damage inspection of metal members, the difference between the distribution pattern of microorganisms and cells and the generation mode of cracks is not taken into consideration. There was a risk of misjudging the shape and length.

It is an object of the present invention to inspect a metal member for damage in a short time with high accuracy, to easily manage the obtained damage data, and to automatically determine the remaining life of the metal member based on the degree of damage. It is to provide a damage inspection method and a damage inspection device for a metal member that can be determined in a very simple manner.

[0018]

In order to achieve the above object, the present invention enlarges an image of a portion to be inspected of a metal member with an optical microscope and converts the luminance distribution of the imaged enlarged image into a digital signal. In a damage inspection method for a metal member for determining the presence or absence of a defect as a damage based on a digital signal, a cluster which is a set of pixels left after image processing of an image luminance distribution includes a predetermined number or more of pixels. If so, it is determined that the cluster is a defect, the approximate inclination of the defect is obtained from the coordinates of the pixels at both ends of the cluster, and a pixel in a predetermined range is vector-searched in relation to the approximate inclination direction from the end of the cluster. If so, it is determined that the pixel is included in the same defect, vector search is again performed from the tip of the pixel determined to be included, and a new pixel is determined. Lost the point of the cluster determines that the entire defects, proposes a damage inspection method for a metal member for calculating the length and angle of the defect.

In the vector search, each time the vector search from both ends of the cluster determined to be defective is completed, within a circle with a predetermined diameter or within a predetermined rectangle from the ends of the pixels recognized as within the cluster. A vector search can be performed again.

In order to achieve the above object, the present invention also enlarges and images an inspection target portion of a metal member with an optical microscope, converts the luminance distribution of the captured enlarged image into a digital signal, and based on the digital signal. In the damage inspection method for a metal member for determining the presence or absence of a defect as a damage, if the cluster, which is a set of pixels left after image processing the brightness distribution of the image, includes a predetermined number or more of pixels,
The cluster is determined to be a defect, the approximate inclination of the defect is obtained each time the vector search is completed from the coordinates of the pixels at the both ends of the cluster, and the vector is calculated from the tip of the cluster within an angle range of ± 90 degrees with respect to the longitudinal direction of the defect. If there is a pixel, it is determined that the pixel is included in the same defect, and a vector search is performed again from the tip of the pixel determined to be included, and the cluster at the time when there is no new pixel is determined as the entire defect. However, the present invention proposes a method for inspecting a metal member for damage, which calculates the length and angle of a defect.

When the metal member has a rotation axis, if the inclination of the defect is within 45 degrees with respect to the rotation axis, vector search is performed within an angle range of ± 90 degrees with respect to the axial direction from the tip of the cluster. , The inclination of the defect is 4 with respect to the rotation axis.
If it exceeds 5 degrees, it will be ±
A vector search is performed within an angle range of 90 degrees.

In any case, a plurality of inspection target portions of the metal member are continuously imaged, and when the cluster recognized as a defect straddles an adjacent image, the adjacent image data are connected and connected. A vector search is performed from the tip of the defective pixel.

Further, when it is determined that there is no new pixel by repeating the vector search from the tip of the pixel recognized as within the cluster, the processing level of the luminance distribution is lowered by a predetermined value and the vector search is performed again from the tip of the defective pixel. By repeating the above, the cluster at the time when the predetermined processing level is reached is determined as the entire defect, and the length and angle of the defect can be calculated.

Further, the tips of adjacent defects have a distance of 0.
If they are close to each other within 2 mm, they may be treated as the same defect.

The length of the cluster determined to be a defect and the width in the direction orthogonal to this length direction are obtained, and if the ratio of the width to the length is 0.5 or more, the defect is determined to be a corrosion pit, If the width to length ratio is less than 0.5, the defect is determined to be a crack.

For a defect determined to be a crack, the frequency distribution of crack lengths is calculated to determine the maximum crack length, and the maximum crack length is calculated from the maximum crack length based on a master curve of crack growth which is determined in advance. Estimate the remaining life of metal members.

In any of the metal member damage inspection methods, as a pretreatment, the surface of the metal member is polished with emery paper, chemical polishing or electrolytic polishing is performed, and then observation with an optical microscope is performed. Alternatively, the surface of the metal member is polished with emery paper, and further chemical polishing or electrolytic polishing is performed to create a replica of the metal structure of the metal member,
This replica is observed with an optical microscope.

In order to achieve the above object, the present invention further provides an optical microscope for magnifying and observing a metal member to be inspected, and an optical microscope for moving a desired member to be inspected by moving the metal member and the optical microscope relatively. A camera for capturing a magnified image of the optical microscope, an image processing unit for converting the luminance distribution of the captured magnified image into a digital signal, and a predetermined luminance level of the image output from the image processing unit. If the cluster, which is a set of pixels, contains a certain number of pixels or more, a means for determining the cluster as a defect, a means for obtaining the approximate inclination of the defect from the coordinates of the pixels at both ends of the cluster, and A vector search is performed for a predetermined range of pixels in relation to the general tilt direction, and if there is a pixel, it is determined that the pixel is included in the same defect, and A vector search again from the tip of the pixels that have been determined to be a means for determining the overall defect clusters at the time a new pixel is exhausted,
An object of the present invention is to propose a damage inspection device for a metal member, which comprises means for calculating the length and angle of a defect.

The metal member damage inspection apparatus includes means for determining a maximum crack length by obtaining a frequency distribution of crack lengths,
Means for estimating the remaining life of the metal member from the maximum crack length based on a master curve for crack growth which has been obtained in advance can be provided.

[0030]

In the present invention, a metal member such as a shaft is set in the inspection device, the optical microscope is moved to the inspection site such as the positioned shaft, and the surface of the shaft or the like is enlarged and observed by the optical microscope. The image of the optical microscope is picked up by a video camera or the like and processed. The computer determines, based on the image-processed data, whether the shaft has a crack as a damage. When a crack is recognized, the shape of the crack is determined by a connection process or a coalescing process of the crack and another crack. The determined crack distribution is analyzed, the maximum crack length is determined, and the remaining life is judged based on the master curve of crack growth.

Therefore, a series of operations from the detection of minute cracks to the judgment of the remaining life of the shaft can be automatically executed. At that time, only by inputting the inspection conditions into the computer, the presence / absence of cracks on the shaft surface and the diagnosis of the remaining life are automatically executed, so that the inspection time can be greatly shortened. It goes without saying that data acquisition, processing, and analysis can be executed in a short time, but data management is facilitated and data can be quantitatively grasped and managed.

Further, since all the parts such as shafts where cracks may occur are inspected, inaccuracies due to estimation and interpolation occur in the part where data is not taken in as in the conventional extreme value statistical method. There is nothing to do. Further, even an operator who is not an expert inspector can accurately detect a crack, and therefore the remaining life of the shaft can be quantitatively and highly accurately diagnosed.

[0033]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A crack detection and life evaluation device for a rotating body, which is an embodiment of the damage inspection device for metal members according to the present invention, will now be described with reference to the drawings.

FIG. 1 is a block diagram showing the overall construction of a crack detection and life evaluation device for a rotating body, that is, a metal member damage inspection device according to the present invention for evaluating the life of a boiler feed water pump shaft, and FIG. FIG. 3 is a flowchart showing a crack detection procedure by image processing and a residual life evaluation procedure based on the distribution of cracks, and FIG. 3 is a flowchart showing the crack image processing procedure in more detail.

The damage inspection apparatus of this embodiment is roughly classified as follows.
The inspection apparatus main body 2, the chuck 3 and the centering mechanism 4 for holding the shaft 1 to be inspected, the rotating mechanisms 5 and 6 for rotating the shaft 1, the optical microscope 11 for observing the surface of the shaft 1, and the optical microscope 11 for the shaft. 1, a moving mechanism 7, 8, 9, 10 for moving in the central axis direction, an image processing device 16, and a computer 20.

One of the shafts 1 is fixed to the chuck 3, and the other is supported by the centering mechanism 4. The chuck 3 is connected to the stepping motor 5 via a speed reducer 6, and is rotationally driven under the control of the stepping motor driving device 13.

In order to observe the surface of the shaft 1, the field of view of the optical microscope 11 is imaged by a video camera 12 such as a CCD camera. The observation magnification of the optical microscope 11 can be adjusted according to the observation target by exchanging the objective lens and the eyepiece lens. The observation magnification is set to several tens to 200 times for a crack and several times to 20 times for a corrosion pit. Optical microscope 11 and camera 12
And are attached to the measurement stand 10. Measuring stand 10
When the driving ball screw 8 is rotated by the stepping motor 7, the guide shaft 9 parallel to the central axis of the shaft 1
Move along. The stepping motor 7 is controlled by the stepping motor driving device 14.

The stepping motor driving device 13 and the stepping motor driving device 14 are the interface 1
5 and the interface 19 to the computer 20. Stepping motor drive device 1
3 and 14 send a voltage pulse signal to the stepping motors 5 and 7 to rotate the shaft 1 and drive the measurement stand 10 in the axial direction according to a control procedure described later to observe the surface of a desired portion of the shaft 1. To do so.

The image signal of the video camera 12, that is, the microscope image of the shaft surface is taken in by the image processing device 16 and subjected to lateral differentiation processing or binarization processing of the brightness distribution.
Converted to digital signal. The processed data is sent to the computer 20 via the interface 19 and undergoes various analyses. The image signal of the video camera 12 is also displayed on the video monitor 17 as necessary,
It is recorded in the video recorder 18.

The computer 20 sends a command to the stepping motor drive devices 13 and 14. The stepping motor driving devices 13 and 14 rotate the shaft 1 to determine the position in the circumferential direction, and move the measurement stand 10 in the axial direction to determine the position in the axial direction of the optical microscope 11. The computer 20 images the surface of a desired inspection portion of the shaft 1, executes image processing, measures cracks and corrosion pits, obtains their distribution, and diagnoses the remaining life.

Next, the crack detection procedure by image processing and the residual life evaluation procedure based on the distribution of cracks will be described with reference to the flowchart of FIG. In step 1, the shaft 1 is attached to the chuck 3 and the centering mechanism 4 of the inspection device body 2.
Fixed by. In step 2, using the axial coordinate Z and the circumferential coordinate θ as parameters, the measurement start position (Z
₀ , θ ₀ ) and the measurement end position (Z ₁ , θ ₁ ) are set, and the measurement range of the shaft 1 is designated. In step 3, the stepping motor drive device 14 sends a predetermined voltage pulse signal to the stepping motor 7 in accordance with a command from the computer 20, and moves the measurement stand 10 to the measurement start position Z ₀ . Further, the stepping motor drive device 13 sends a predetermined voltage pulse signal to the stepping motor 5 in response to a command from the computer 20, rotates the shaft 1, and moves it to the measurement start position θ ₀ .

In step 4, the shaft surface is observed by the optical microscope 11. In step 5, the image processing device 16 image-processes the video signal from the video camera 12 and measures the luminance distribution. In step 6, the computer 20 detects a crack, and in step 7, confirms the presence or absence of a crack. If there is a crack, the image data is recorded in a storage means such as a floppy disk in step 8. If there is no crack, image data is not recorded.

In step 9, the circumferential position is confirmed. If the final angular position θ ₁ is not reached, in step 10, the step number n is incremented by 1, and step 11
At, the stepping motor 5 is rotated so as to advance the circumferential angle to θJ = θJ + θa. At this time, the computer 20 sends to the stepping motor driving device 13 a number of pulses corresponding to the diameter of the shaft 1 corresponding to the increment θa of the rotation angle and the observation magnification of the optical microscope 11.
The stepping motor 5 is rotated.

The above steps 4 to 9 are repeated,
In step 9, the circumferential position is the final angular position θ ₁
If so, in step 12, the axial position is confirmed. If the final position Z ₁ is not reached, in step 13, the number of steps m is advanced by 1, and step 14
At, the stepping motor 7 is rotated so as to advance the axial position to ZJ = ZJ + Za. At this time, the computer 20 sends a pulse corresponding to the increment Za of the axial coordinate according to the observation magnification of the optical microscope 11 to the stepping motor driving device 14 to rotate the stepping motor 7.

Steps 4 to 11 are repeated, and in step 12, the axial position is the final position Z.
If it has reached _1, then go to step 15. Here, strictly speaking, in step 9, if the circumferential position has reached the final angular position θ _1, it is necessary to rotate the stepping motor 5 and return the shaft 1 to the initial angle θ ₀ .

In step 15, the crack is connected based on the image data recorded in step 8. In step 16, the crack distribution is analyzed, and in step 17
In step 18, the maximum crack length is determined, and in step 18, the remaining life is evaluated based on the master curve 19 for crack growth. Details of the image processing, crack connection processing, residual life evaluation, and the like will be sequentially described.

FIG. 3 is a flowchart showing the crack image processing procedure in more detail, and corresponds to steps 5 and 6 in FIG. In step 21, the maximum brightness Imax and the minimum brightness Imin are measured from the brightness distribution of the microscope image. In step 22, the level of binarization processing is initialized. For example, the initial level is 50% of the maximum brightness Imax, that is, Im = 0.5Imax. In step 23,
Binarization processing is executed, and in step 24, a cluster that becomes a crack bud is searched for. When the binarization processing is performed, the image data becomes digital data obtained by dividing the monitor screen into 400 pixels in the vertical direction and 600 pixels in the horizontal direction. 40
In the case of 0 × 600, one screen is composed of 24,000 pixels. The defect damages on the pump shaft surface of interest here are fatigue cracks and corrosion pits. A method for discriminating between a fatigue crack and a corrosion pit will be described later, but here, assuming a case of a fatigue crack, a treatment method thereof will be described.

FIG. 4 is a schematic diagram of a microscope image of a fatigue crack defect. The defective portion becomes black in the microscope visual field. When the microscopic image is binarized, the pixels corresponding to the defective portion are displayed in black.

FIG. 5 is a schematic diagram of digital data obtained by the binarization process. The defective pixels are displayed with fine shading. Since defects have continuity, search for clusters in which adjacent pixels are connected in black from all pixels, and for example, if 20 or more pixels are connected in black, it is temporarily determined to be a crack. To do. In FIG. 5, there are 12 portions corresponding to the defects displayed by fine shading.

Next, at step 25, the tip of the cluster is detected. The procedure is to simply look for pixels on both ends that are identified as defective. In the case of the example in FIG. 5, the pixel with a dot in the shade is determined to be the top of the cluster. Although the defect itself has continuity, when the binarization process is performed, the pixels determined to be defective are not always continuous. Therefore, various processing is required for the data after the binarization processing.

In step 26, the slope of the cluster is calculated. If the coordinates of the pixel determined to be the tip of the cluster are (Zi, θi) and (Zj, θj), the angle Θ with respect to the axial direction of the shaft is Θ = arc tan {r (θi−θj) / (Zi−Zj )}. Where r is the radius of the shaft.

In step 27, a vector search for defective pixels is performed from the tip of the cluster.
In the vector search, as shown in FIG. 6, the tip coordinates (Zi, θi) and (Zi) of the pixel determined to be the tip of the cluster are used.
j, θj), for example, in the range of R = 0.1 mm, the defect is executed within a range of ± 90 degrees in the longitudinal direction. Expressed in terms of angles, at the tip of each cluster, Θ ₁ = Θ + 90 Θ ₂ = Θ + 180 ± 90. That is, search for black pixels in a semicircular range from the tip of the cluster. In FIG. 6, the pixels displayed with diagonal shading are included in the semicircular range.

In step 28, it is assumed that the clusters are merged and the pixels displayed by diagonally shaded areas are also included in the clusters.

Next, as shown in FIG. 7, a new tip of the cluster is detected, and a semicircular vector search is executed again from the tip. These steps are repeated to determine the total defect coverage.

In the binarization processing of step 22, the recognition of cracks may be insufficient with only one processing level. Therefore, the binarization processing level is gradually lowered and a vector search is performed from the crack tip.

In step 29, it is determined whether the binarization processing level Im has reached the final level Ic. If the binarization processing level Im has not reached the final level Ic, the step number n is advanced by 1 in step 30, and the binarization processing level is reset in step 31. Here, it is assumed that the maximum brightness Imax is decreased from 50% by 10% by 10% as in the case of Im = Im-0.1 · Imax, and the final level Ic is set to 20%.

FIG. 8 is a schematic diagram of the result of vector search from the crack tip while gradually lowering the binarization level. Here, the pixels discriminated at the binarization processing level 50%, the pixels discriminated at 40%, the pixels discriminated at 30%, and the pixels discriminated at 20% are distributed in this order from the center to the outside. The state is shown.

Next, steps 23 to 31 are repeated, and if the binarization processing level Im reaches the final level Ic in step 29, the inclination Θ and the tip coordinate (of the cluster) finally determined to be defective in step 32 ( Zti, θti),
(Ztj, θtj) and the defect length 2c are calculated.

In step 33, the cluster width b
To calculate. As shown in FIGS. 9 and 10, the width b in the direction orthogonal to the length 2c in the longitudinal direction of the cluster is obtained.
In step 34, the aspect ratio b / c of the cluster
Is larger or smaller than 0.5. If the cluster is elongated and the aspect ratio b / c is small as shown in FIG. 9, the defect is determined to be a crack in step 35, and the cluster is almost round and the aspect ratio b / c is shown in FIG.
Is larger, the defect is determined to be a corrosion pit in step 36.

11 to 14 show an example of another defect damage determination method. There are various methods for the vector search from the cluster tip in step 27 of FIG. Figure 1
1 indicates a square vector search. At the tip coordinates (Zi, θi) of the pixel determined to be the tip of the cluster, (Zi + A, θi + α), (Zi + A, θi−α),
Within the square range of (Zi-A, θi + α), (Zi-A, θi-α), at the other tip coordinate (Zj, θj), (Zj +
A, θj + α), (Zj + A, θj−α), (Zj−A, θj +
Pixels are searched for in the square range of (α), (Zj-A, θj-α).

FIG. 12 shows a circular vector search. In the tip coordinates (Zi, θi) of the pixel determined to be the tip of the cluster, the origin is (Zi, θi) (A, α)
Pixels are searched for in the circular range of, and the tip coordinates (Zj,
In θj), pixels are searched for in the circular range of (A, α) with (Zj, θj) as the origin.

FIG. 13 shows the shaft 45 with respect to the axial direction.
This is a vector search at the tip of the cluster tilted more than one degree, and at the tip coordinates (Zi, θi) of the pixel determined to be the tip of the cluster, (Zi, θi) is set as the origin (A,
In the semicircular range of α = 0 to 180 degrees, at the other tip coordinate (Zj, θj), (Zj, θj) is the origin (A,
Search for pixels in the semi-circular range of α = 0 to −180 degrees.

FIG. 14 shows 45 in the axial direction of the shaft.
This is a vector search at the tip of the cluster tilted by less than or equal to a degree, and at the tip coordinates (Zi, θi) of the pixel determined to be the tip of the cluster, (Zi, θi) is the origin (A, α = 90 to 270). In the semi-circular range of (degrees), at the other tip coordinate (Zj, θj), the pixel is defined in the semi-circular range of (A, α = 90 to −90 degrees) with (Zj, θj) as the origin. look for.

FIG. 15 is a diagram for explaining the cluster connection processing method of step 15 in FIG. Observe the surface of the shaft with a microscope while shifting it in the circumferential and axial directions by the size of the screen. At that time, since the error becomes large in the area around the screen where the aberration is large, it is removed from the image data with some overlap. Therefore, the amount of movement in the circumferential direction and the amount of movement in the axial direction is the size of the screen that is actually used except for the overlapping portion. In the screen data collected in this way, the defects are not always contained in one screen data. When the defect is large or when the magnification of the optical microscope 11 is increased, the result may spread over several screens. In such a case, as shown in FIG. 15, when it is determined that the defect is in the periphery of the screen, the processing from step 23 to step 31 in FIG. 3 must be executed together with the adjacent screen data. I have to.

FIG. 16 is a schematic diagram showing a method of a coalescing process. The crack distribution analysis in step 16 of FIG. 2 includes a crack coalescence process. Fracture mechanics is required for the residual life evaluation based on crack growth described later. This is because when adjacent cracks are very close to each other, they often coalesce with a small crack growth. A circular vector search is performed to find adjacent cracks from the final ends of the cluster determined to be a crack, and if not only the crack tip but a part of the crack is within the search range, it is regarded as one crack. To coalesce. Then, the crack length is obtained from the combined crack tip. If some other cracks are not in the search range after searching, the crack is treated as existing.

When all the crack detection and coalescence processing are completed as described above, the crack distribution analysis is completed. At this time, the inclination Θ of all cracks and the coordinates of the crack tip (Z
ti, θti), (Ztj, θtj), and the defect length 2c are stored as data in a storage device such as a floppy disk.

Using the data of the crack distribution, the maximum crack length in step 17 in FIG. 2 is determined. A frequency distribution graph such as that shown in FIG. 17 is displayed on the CRT screen of the computer 20, and the maximum crack length 2c is displayed.
Calculate and display max.

FIG. 18 is a flow chart showing the procedure of the remaining dead life evaluation in step 18 in FIG. When the operation of the boiler feed pump is started, in step 41,
Record the number of revolutions Ni. In step 42, a crack is detected, and in step 43, the presence or absence of a crack is determined. If there are no cracks, operation continues at step 44. Steps 41 to 42 are repeated, and if there is a crack in step 43, the maximum crack length 2c is substituted into the master curve for crack growth as shown in FIG. 18 in step 45 to calculate the life consumption rate N / Nf. calculate. In step 46, the remaining life Nr is calculated based on the life consumption rate N / Nf by Nr = Ni (1-N / Nf) / (N / Nf). If in step 47, the remaining life N
If r is greater than the number of revolutions Nc until the next periodic inspection,
In step 48, the operation is continued and steps 41 to 46 are repeated. On the other hand, in step 47, if the remaining life Nr is smaller than the rotation speed Nc until the next regular inspection,
Proceeding to step 49, the pump operation is stopped. In step 50, if the crack is short, the crack is deleted with a grinder or the like, and if the crack is long, the shaft is replaced with a new one.

With the fatigue characteristics of load control shown in FIG. 19, a crack with a length of about 0.1 mm that can be observed by an optical microscope and can be detected by image processing has a life consumption rate N / Nf = 65.
% Until a macroscopic crack of about 3 mm is reached.
It is considered that the logarithm log2c of the crack length increases almost in proportion to the life cost ratio N / Nf. Therefore, the life consumption rate N / Nf is given by N / Nf = (log2c + 4.20) /4.924.

FIG. 20 is a flow chart showing a processing procedure including pretreatment of the shaft surface in the image processing of defects. In the defect inspection of the boiler feed pump, first, in step 51, the surface of the pump shaft is directly observed without pretreatment. In step 52, the oxide film on the surface of the pump shaft is removed by polishing with emery paper. In step 53, observation with an optical microscope is performed, and in step 54, an inspection mainly including detection of corrosion pits is executed. Since detection of fatigue cracks is difficult only by polishing with emery paper, the pump shaft surface is mirror-finished by chemical polishing or electrolytic polishing in step 55, and the fatigue cracks are detected in step 56.

Instead of directly observing the surface of the pump shaft, a replica of the pump shaft surface may be made using a molten acetyl cellulose film or the like after the mirror finishing, and this replica may be observed with an optical microscope.

Further, in the above description of the embodiments, the crack detection and life evaluation device for the pump shaft, that is, the rotating body has been described, but the present invention is not limited to the rotating body, and for detecting cracks in other metal members. Obviously, it can be applied to a life evaluation device.

[0073]

According to the present invention, the inspection time and the like can be automatically detected by simply inputting the inspection conditions and the like into the computer and automatically detecting the presence or absence of a crack on the surface of the shaft or the like and diagnosing the remaining life. It can be greatly shortened. It goes without saying that data acquisition, processing, and analysis can be executed in a short time, but data management is facilitated and data can be quantitatively grasped and managed.

Further, since all the parts such as shafts where cracks may occur are inspected, it is inaccurate by estimating or interpolating the part where data is not taken in as in the conventional extreme value statistical method. Does not occur.

Furthermore, even an operator who is not a skilled inspector can accurately detect cracks, so that the remaining life of the shaft can be quantitatively and highly accurately diagnosed.

Therefore, it is possible to improve the reliability of the boiler water supply pump as a whole.

[Brief description of drawings]

FIG. 1 is a block diagram showing an overall configuration of a crack detection and life evaluation device for a rotating body, that is, a metal member damage inspection device according to the present invention for evaluating the life of a boiler feed pump shaft.

FIG. 2 is a flowchart showing a crack detection procedure by image processing and a residual life evaluation procedure based on a crack distribution.

FIG. 3 is a flowchart showing in more detail a crack image processing procedure.

FIG. 4 is a schematic diagram of a microscope image of a fatigue crack defect.

FIG. 5 is a schematic diagram of digital data obtained by binarization processing.

FIG. 6 is a schematic diagram of a semicircular vector search from the cluster tip.

FIG. 7 is a schematic diagram of a semicircular vector search from a new tip of a cluster.

FIG. 8 is a schematic diagram of a crack search in which the binarization level is sequentially lowered.

FIG. 9 is a schematic diagram showing a method for discriminating a fatigue crack from a corrosion pit when the target is a fatigue crack.

FIG. 10 is a schematic diagram showing a method of discriminating a fatigue crack from a corrosion pit when the target is a corrosion pit.

FIG. 11 is a schematic diagram of a square vector search from the cluster tip.

FIG. 12 is a schematic diagram of circular vector search from the cluster tip.

FIG. 13 is a schematic diagram of a semicircular vector search from the cluster tip according to the cluster inclination.

FIG. 14 is a schematic diagram of a semicircular vector search from the cluster tip according to the cluster inclination.

FIG. 15 is a schematic diagram showing a process of connecting clusters across different screens.

FIG. 16 is a schematic diagram showing a coalescence process of cracks.

FIG. 17 is a schematic diagram of frequency distribution of crack lengths.

FIG. 18 is a flowchart showing an example of a remaining life evaluation procedure.

FIG. 19 is a diagram showing an example of a master curve for fatigue crack growth.

FIG. 20 is a flowchart illustrating an example of a preprocessing method of image processing.

[Explanation of symbols]

 1 Pump Shaft 2 Inspection Device Main Body 3 Chuck 4 Centering Mechanism 5 Shaft Rotating Stepping Motor 6 Reduction Gear 7 Measuring Platform Moving Stepping Motor 8 Measuring Platform Guide 9 Ball Screw 10 Measuring Platform 11 Optical Microscope 12 Video Camera (CCD Camera) 13 Shaft Rotating stepping motor driving device 14 Measuring platform moving stepping motor driving device 15 Interface 16 Image processing device 17 Video monitor 18 Video recorder 19 Interface 20 Computer

Continuation of front page (51) Int.Cl. ⁵ Identification code Office reference number FI technical display location G06F 15/62 400 9287-5L

Claims (14)

[Claims] 1. A region to be inspected of a metal member is magnified and imaged with an optical microscope, the luminance distribution of the magnified image is converted into a digital signal, and the presence or absence of a defect as damage is determined based on the digital signal. In the damage inspection method for a metal member, wherein a cluster, which is a set of pixels left after image processing the brightness distribution of the image, includes a predetermined number or more of pixels, the cluster is determined to be a defect, and the cluster The approximate inclination of the defect is obtained from the coordinates of the pixels at both tips, and a vector search is performed for a predetermined range of pixels from the tip of the cluster in relation to the approximate inclination direction. If there is a pixel, that pixel is included in the same defect. Then, the vector search is performed again from the tip of the pixel determined to be included, and the cluster at the time when there is no new pixel is determined. A method for inspecting damage to a metal member, which comprises determining the entire defect and calculating the length and angle of the defect. 2. The metal member damage inspection method according to claim 1, wherein each time a vector search from both tips of a cluster determined to be defective is completed, a pixel having a predetermined diameter from the tip of a pixel recognized as inside the cluster is detected. A method of inspecting a metal member for damage, which comprises performing a vector search again within a circle. 3. The damage inspection method for a metal member according to claim 1, wherein each time a vector search from both ends of a cluster determined to be defective is completed, a predetermined rectangle is formed from the ends of the pixels recognized as in the cluster. A method for inspecting a metal member for damage, which comprises performing a vector search again in the inside. 4. An inspection target portion of a metal member is magnified and imaged by an optical microscope, the luminance distribution of the magnified image is converted into a digital signal, and the presence or absence of a defect as a damage is determined based on the digital signal. In the damage inspection method for a metal member, wherein a cluster, which is a set of pixels left after image processing the brightness distribution of the image, includes a predetermined number or more of pixels, the cluster is determined to be a defect, and the cluster Each time the vector search is completed from the coordinates of the pixels at both tips, the inclination of the defect is determined, and ± from the tip of the cluster with respect to the longitudinal direction of the defect.
When vector search is performed in the angle range of 90 degrees, if there is a pixel, it is determined that the pixel is included in the same defect, vector search is again performed from the tip of the pixel determined to be included, and when there is no new pixel Is determined as the entire defect, and the length and angle of the defect are calculated. 5. An inspection target portion of a metal member having a rotation axis is magnified and imaged with an optical microscope, the luminance distribution of the magnified image thus taken is converted into a digital signal, and a defect as a damage is generated based on the digital signal. In a method for inspecting a metal member for the presence or absence of a defect, if a cluster, which is a set of pixels left after image processing the brightness distribution of the image, includes a predetermined number or more of pixels, the cluster is determined to be a defect. Then, each time the vector search is completed from the coordinates of the pixels at both ends of the cluster, the approximate inclination of the defect is obtained. If the inclination of the defect is within 45 degrees with respect to the rotation axis, the axis from the tip of the cluster is calculated. Vector search in an angle range of ± 90 ° with respect to the direction, and if the inclination of the defect exceeds 45 ° with respect to the rotation axis, ± 90 with respect to the circumferential direction from the tip of the cluster. If there is a pixel, it is determined that the pixel is included in the same defect, the vector search is performed again from the tip of the pixel determined to be included, and the cluster at the time when there is no new pixel Is determined as the entire defect, and the length and angle of the defect are calculated. 6. The damage inspection method for a metal member according to claim 1, wherein a plurality of inspection target portions of the metal member are continuously imaged, and clusters recognized as defects are adjacent to each other. In the case of straddling an image, the adjacent image data are connected, and a vector search is performed from the tip of the connected defective pixel. 7. The metal member damage inspection method according to claim 1, wherein a vector search is repeated from a tip of a pixel recognized as within the cluster, and it is determined that there is no new pixel. Then, the processing level of the brightness distribution is lowered by a predetermined value, the vector search is repeated from the tip of the pixel of the defect, and the cluster at the time when the predetermined processing level is reached is determined as the entire defect, and the length and angle of the defect are determined. A method for inspecting damage to a metal member, the method including: 8. The metal member damage inspection method according to claim 1, wherein if the tips of adjacent defects are close to each other within 0.2 mm, they are regarded as the same defect. A method for inspecting damage to a metal member, the method including: 9. The damage inspection method for a metal member according to claim 1, wherein a length of the cluster determined to be the defect and a width in a direction orthogonal to the length direction are obtained. If the ratio of the width to the length is 0.5 or more, the defect is determined to be a corrosion pit, and if the ratio of the width to the length is less than 0.5, the defect is determined to be a crack. A method for inspecting a metal member for damage. 10. The method for inspecting damage to a metal member according to claim 9, wherein the maximum crack length is determined by obtaining a frequency distribution of the lengths of the cracks, and based on a master curve of the crack growth obtained in advance. And a method for inspecting a metal member for damage, wherein the remaining life of the metal member is estimated from the maximum crack length. 11. The method for inspecting a metal member for damage according to claim 1, wherein the surface of the metal member is polished with emery paper, and chemical polishing or electrolytic polishing is further performed. A method for inspecting damage to a metal member, which comprises observing with a light microscope. 12. The metal member damage inspection method according to claim 1, wherein the surface of the metal member is polished with emery paper, and chemical polishing or electrolytic polishing is further performed. A method for inspecting damage to a metal member, which comprises making a replica of a metal structure of a member and observing the replica with the optical microscope. 13. An optical microscope for magnifying and observing a metal member to be inspected, means for relatively moving the metal member and the optical microscope so that a desired inspection object portion faces the optical microscope, and the optical microscope. A camera that captures an enlarged image of the image, an image processing unit that converts the luminance distribution of the captured enlarged image into a digital signal, and a cluster that is a set of pixels of a predetermined luminance level in the image output from the image processing unit. , If it includes a predetermined number of pixels or more, means for determining the cluster as a defect, means for obtaining the approximate inclination of the defect from the coordinates of pixels at both tips of the cluster, and the approximate inclination from the tip of the cluster A vector search is performed for a range of pixels in relation to the direction, and if there is a pixel, it is determined that the pixel is included in the same defect, and A metal member including means for performing vector search again from the tip of the pixel determined to be covered and determining a cluster at the time when no new pixel is present as an entire defect, and means for calculating the length and angle of the defect. Damage inspection device. 14. The damage inspection device for a metal member according to claim 13, wherein a means for determining a maximum crack length by obtaining a frequency distribution of the lengths of the cracks, and a master curve for crack growth obtained in advance. And a means for estimating the remaining life of the metal member from the maximum crack length based on the above.