RU2323492C2 - Method for detecting surface flaws on cylindrical pieces of equipment - Google Patents

Abstract

FIELD: nuclear fuel production.

SUBSTANCE: proposed method for detecting surface flaws on cylindrical pieces of equipment includes sequential delivery of piece of equipment under inspection to surface inspection position. Butt-end surfaces of piece of equipment delivered to inspection position are illuminated by radiating flux. Radiation detectors receive radiation reflected from butt-end surfaces. Images received from detectors are treated in analyzing device. Side surface of piece of equipment under inspection is illuminated by radiation flux passed at angle φ to normal to its surface. Image reflected at angle to normal equal to incident angle of radiating flux is received. Butt-end surfaces of piece of equipment under inspection are illuminated by radiating flux passed at angle α to normal to butt-end surface. Radiation reflected from butt-end surfaces at angle to normal equal to incident angle of radiating flux is received. Image boundaries of piece-of-equipment surfaces are determined in image frames by means of analyzing device using boundary tracing method. Same method is used to find surface flaw sections on surface images. Surface flaws are described by geometric figures. Surface areas of these figures are calculated. Type of flaws is determined, and decision is taken on fitness of piece of equipment under inspection basing on logic decision rules.

EFFECT: enhanced reliability of on-line inspection of cylindrical pieces of equipment for surface flaws and their type.

1 cl 4 dwg

Description

The invention relates to a technology for the production of nuclear fuel, in particular to methods for controlling the appearance of fuel pellets.

A known method for detecting defects on the end surfaces of fuel pellets, implemented in a device for monitoring cylindrical objects (US patent No. 4226539, IPC G01N 21/48).

The method is based on moving the controlled object to the first control position, rotating the object around the longitudinal axis, illuminating the side surface from the radiation source and receiving the radiation scattered by the side surface using a one-dimensional multi-element photodetector, moving the object at a constant speed without rotating it through the second control position in the direction perpendicular to the axis of the object, illuminating the end surfaces of the object with inclined parallel light fluxes from light sources with collies by sensors, reception of radiation scattered by end surfaces with multielement linear photodetectors, threshold processing of video signals from photodetectors.

The disadvantages of the above control method are its low sensitivity to minor surface defects, since the threshold signal levels used in the threshold processing of the video signal are determined taking into account the integrated assessment of the intensity of the optical signal from the tablet in a separate optical channel. It is obvious that small rare defects will have a weak effect on the average signal level in this channel. In addition, in this method, scattered radiation from the object is recorded in the main optical channels, giving an image with low contrast, which also reduces the likelihood of detecting surface defects. Also, this method uses mechanical scanning of the end surfaces of the tablet in one coordinate, which requires high stability of the linear velocity of the object, which in production conditions is almost impossible due to the presence of vibrations that lead to a “blur” of the video signal and, as a result, distort the true geometry defects. Moreover, this control method has limited functionality, since it does not provide a determination of the type of detected defect (crack, chip, pore).

There is also known a method of complete visual control and remote sorting of nuclear fuel tablets (patent WO No. 00/28549, IPC G21C 17/06, G01N 21/88).

The method includes feeding the tablets to the control position, rotating the tablet to the control position around its axis, while illuminating the side and end surfaces of the tablet with illuminators emitting light flux in a direction at an angle to the normal to the end surfaces of the tablet, registering images of the end surfaces of the tablet with two progressive cameras scan, oriented at an angle to the normal to the end surfaces of the tablet, registration of the image of the side surface of the tablet with a third camera, at the same time full display of arranged images of tablet surfaces on one control monitor, visual or automatic (using a computer) image analysis and remote sorting of tablets by an operator (or using a computer).

The disadvantages of the above method for monitoring fuel pellets are the low reliability of detection of defects on the end surfaces, due to the low contrast of the image due to the registration, mainly, of the radiation scattered by the end surfaces, which is formed when the illuminators and cameras are placed on one side with respect to the normals to the end surfaces of the tablet.

Closest to the proposed method is a method for automatically controlling the surface of nuclear fuel pellets (Japanese Patent No. 06-051091, IPC G21C 17/06, G01N 21/88).

The method consists in the fact that the tablet with the help of a transporting device is fed sequentially to the control position of its surfaces. One of them controls the lateral (cylindrical) surface of the tablet, on which it undergoes rotation around its longitudinal axis. During rotation, a strip of the side surface of the tablet is illuminated from a one-dimensional irradiating device, and the rays from it are directed normal to the surface of the tablet. The rays reflected and scattered by the surface of the tablet are received by a CCD camera, the optical axis of which is located at an angle of 25-60 ° to the direction of the illuminating beams. An electrical signal from the camera is transmitted via a communication channel to the analytical device. In another control position, the end surfaces of the tablet are controlled. The end surfaces are alternately illuminated from an annular illuminator arranged so as not to impede the passage of reflected rays into the radiation receiving device. The rays reflected and scattered by the surface are received by a CCD camera, the optical axis of which is perpendicular to the end surface of the tablet. An electrical signal from the CCD camera enters the analytical device. In the analytical device, the value of such an electrical signal, which occurs during the control of tablets of the highest quality, is taken as zero (norm value). This value is obtained during the training (calibration) of the system and stored in memory. The area with a surface defect will correspond to a signal with a lower value than the norm (it will be negative). The detection of negative signals is a processing method that identifies defects. Information about defects is recorded in the memory of the analytical device and then a program is connected that detects low-quality tablets, which are then automatically separated from high-quality ones.

Signs of the prototype that coincide with the essential features of the claimed invention are the sequential supply of the controlled object to the position of control of its surfaces, irradiation of the strip of the side surface of the object at the position of control of the side surface of the radiation stream, reception of the CCD reflected from the side surface of the radiation by the camera, transmission of the electrical signal from the camera to the analytical device, irradiating the first end surface of an object with a radiation flux, receiving reflected from the first end surface radiation from the CCD camera, transmitting an electric signal corresponding to the first end surface from the CCD camera to the analytical device, irradiating the second end surface of the object being monitored with a radiation stream, receiving the radiation from the second end surface of the radiation from the CCD camera, transmitting an electric signal corresponding to the second end surface from the CCD cameras into the analytical device, processing of electrical signals from all surfaces of the object in the analytical device, sorting the objects of control I'm on high-quality and low-quality.

The reasons that impede the achievement of the following technical result when using the known method adopted for the prototype include:

- low probability of detection of objects of control with surface defects due to the low contrast of the obtained images of its surfaces due to their illumination by the radiation flux directed perpendicular to the irradiated surface, as a result of which the camera perceives not only the rays reflected from the surface, but also a significant part of the scattered radiation;

- low information capabilities of the method, as the method of signal processing in the analytical device adopted in the prototype does not allow the analysis of defects by type (crack, chip, pore ...) and, therefore, does not provide important technological information.

The invention proposed by the authors sets the solution to the following problems:

- increasing the likelihood of detecting objects with surface defects;

- improving the information capabilities of the control method.

When implementing the invention, the following technical result can be obtained: operational, highly reliable control of cylindrical objects for the presence and nature of surface defects, high information content of the control.

The specified technical result in the implementation of the invention is achieved by registering the radiation reflected from the controlled surfaces, parallelizing the processes of image registration and processing them in an analytical device, an original method of processing information in an analytical device.

Distinctive features from the prototype are: illumination of the side surface of the controlled object with a radiation flux directed at an angle φ to the normal to its surface, receiving reflected radiation at an angle to the normal equal to the angle of incidence of the irradiating flux; illumination of the end surfaces of the controlled object with a radiation flux directed at an angle α to the normal to the end surface, receiving reflected from the end surfaces of the radiation at an angle to the normal equal to the angle of incidence of the irradiating stream; processing images obtained from radiation detectors in an analytical device by determining the boundaries of images of the surfaces of an object in image frames by the method of bypassing boundaries, by finding on the images of surfaces by the method of bypassing the boundaries of defective surface sections, describing defective surfaces by geometric figures, calculating the area of these figures and determining the type of defects. The decision on the suitability of the controlled object is made on the basis of logical decision rules.

The achievement of the technical result is possible due to illumination of the end surfaces of the test object by the radiation flux directed at an angle to the normal to the end surface, reception of radiation reflected from the end surfaces at an angle equal to the angle of incidence of the irradiating flux, illumination of a strip of the lateral cylindrical surface of the object located along its longitudinal axis, radiation flux directed at an angle to the normal to the surface of the object, receiving reflected radiation at an angle equal to the angle of incidence of the irradiating beam processing of information in the analytical device, which consists in determining the boundaries of images of the surfaces of an object in image frames by the method of bypassing borders, searching on images of surfaces by the method of bypassing the boundaries of defective surface sections, describing defective surfaces by geometric figures, calculating the areas of these figures, determining the type of defects and deciding on the suitability of the tablet based on logical decision rules.

The invention is illustrated by drawings, where figure 1 presents a diagram of a device that implements the inventive method; figure 2 - image of the end surface without large surface defects; figure 3 - image of the end surface with characteristic defects; figure 4 is a scan image of the side surface with characteristic defects.

The implementation of the method is proposed, for example, in the control device for tablets of nuclear fuel, shown in figure 1, which operates in accordance with the inventive method as follows.

Fuel tablets 1 are fed through a vibrating tray 2 in a continuous flow parallel to each other to a conveying device 3. The conveying device 3 consists of a movable rotor 4 with grooves 5 for gripping the tablets. In the housing 3 there is an annular groove 6, the cylindrical surfaces of which 7, 8 limit the movement of the tablets 1 in the radial direction of the cylindrical surfaces 7, 8. In the axial direction, the movement of the tablets is limited by the walls 9, 10. From the vibrating tray 2, the tablets arrive at position 11 of the conveying device 3 and then, using a step drive, they are fed to position 12 of device 3. At position 12 of the conveying device 3, a hole is made in the side wall 9 (section AA), through which the tablet is illuminated with light, directed at an angle α to the normal to the end surface 13 of the tablet, from the radiation source 14. The light reflected from the flat surface of the tablet at an angle α to the normal is collected by the lens 15 onto the photosensitive area of the multi-element two-dimensional photodetector 16 of the television camera 17. The image of the tablet surface 13 recorded by the camera 17 is transmitted to the analytical device 18, where it is digitized and processed. The image of the end surface of the suitable tablet during registration of the rays reflected from the surface 13 will have the form shown in Fig.2, where dark dots 19 against a bright background highlight small areas with defects that are not a sign of rejection of the tablet. Figure 3 shows the image of the end surface 13 of the tablet, on which there are large defects in the form of chips 20 at the edge of the surface, chips 21 at the interface between the surface and the hole 22, pores 23. Defects on a bright background of the surface in the image will be high-contrast, since light scattered on defects, practically does not get into the camera lens.

The image processing of the end surface in the analytical device 18 is performed in the following sequence.

The defect search algorithm on end surfaces processes black-and-white images of size n × m pixels (the number of matrix elements). The end image in this image is located in a square plot A of size k × k pixels. The coordinates of section A and the expected coordinates of the center of the butt are known to the program. If the end face has no defects, then its image contains a flat part in the form of a light ring and a central hole in the form of a dark circle. The presence of a chamfer and small defects at the boundaries of the ring affects the fact that the shape of the outer and inner borders of the ring has local deviations from the circle.

The end image processing algorithm is divided into the following steps.

1. Search for external contours of white areas.

In the beginning, the outer borders of all white areas in the square A are searched. For this, an algorithm for bypassing the borders is used (see Methods of computer image processing. Edited by V. A. Soifer, Fizmatlit, 2001). A horizontal horizontal scan of this square is made. If a white dot a occurs after black dots, the scanning process is interrupted and a check is made to see if this white dot is the border of the white area or an occasional minor image defect. To do this, 8 neighboring points surrounding the point a are checked. If all of them turned out to be black, then point a is declared random. It is ignored, and the sweep process continues. If bypassing the neighbors of point a a white point b is detected, then points a and b are declared boundary, and the check of neighboring points of point b begins. If a white dot c is found among them, then it is also declared boundary. This process continues until the next white point is point a, from which the border bypass process began. All selected boundary points describe the external contours of the first white area, inside which, possibly, there are also black areas. It may turn out that the white ring of the flat end is divided by dark sections into several disconnected parts. The search for other white areas begins at point a, at which the scan was interrupted. If during the passage of the next line there is a white point that has already been declared a boundary point, then the scan continues further from the first black point in this line. The process of outlining white areas ends after passing all the points of square A.

2. Building the border of the white area

The two most distant boundary points of the white section are selected. They determine the "diameter" of the white area. If there is more than one white area, then the maximum diameter is selected. A circle O ₁ is constructed with this diameter and center at the midpoint of the selected diameter. If the white section was one, then this circle will be the approximate boundary between the flat part of the end and the chamfer. If the diameter does not fit the specified restrictions or if the center of the circle deviates from the proposed center of the end by a distance greater than the permissible, then the tablet is considered defective.

3. Correction of the boundary of the flat part of the end face obtained in stage 2.

For images that have passed the limitations of claim 2, the outer boundary of the light ring is refined. It is assumed that the real boundary is not a circle, but deviates from it by no more than some given value V ₁ . A ring of width V ₁ and an inner boundary coinciding with the circle O ₁ is constructed. The ring is divided into angular sectors of the same angular size. Each sector has a white dot farthest from the center of the ring. These points of neighboring sectors are connected by straight line segments. The result is a piecewise linear boundary G ₁ that describes the external contour of the strip of the end part.

4. Search for the boundary of the inner hole.

Information on the coordinates of the center of the end and the diameter of the central hole allows you to build a circle O ₂ , which is the first approximation of its boundary. A ring of width V ₂ and an external boundary coinciding with the circle O ₂ is constructed. The ring is divided into angular sectors of the same angular size. In each sector there is a white dot closest to the center of the ring. These points of neighboring sectors are connected by straight line segments. The result is a piecewise linear boundary G ₂ that describes the outer contour of the hole or the inner contour of the strip of the end part.

5. Search for defects.

Inside between the boundaries of G ₁ and G ₂ , a search for black areas is done. To do this, use the same method as described above. Detected single black pixels surrounded by white pixels are ignored. For each black section, its dimensions are determined by the area of the describing rectangle. For this purpose, in each black spot there are two points farthest from each other. The distance between them is taken as the side d of the describing rectangle. Then there is the point farthest from the line connecting these two points. The distance from this point to the line is equal to half the length of the second side e of the rectangle. The area of such a defect is taken to be s = 0.5ed.

It is believed that defects whose boundaries partially coincide with the boundary G ₁ , have their continuation on the chamfer, not visible in the picture. For such defects, there is an additional area equal to the product of the length of the coincident part of the boundaries and the width of the chamfer. Defects smaller than certain sizes are not taken into account in further processing.

6. Rejection.

The decision to reject the product is made using the methods of pattern recognition (see Zagoruyko N.G. Applied methods of data and knowledge analysis. Publishing House of the IM SB RAS, Novosibirsk, 1999, 270 pp.), Based on logical decision rules (see. Lbov G.S., Startseva N.G. Logical decision rules and questions of statistical stability of decisions.Id. IM SB RAS, Novosibirsk, 1999, 211 pp.). Upon detection of each successive defect, its area s is compared with the allowable area of an individual defect s ₁ . If this threshold is exceeded, the product is rejected. If not, then the area s is added to the defect area counter. If the sum of the areas of defects S exceeds the threshold S *, then the product is rejected. If none of these conditions is met, then the product is considered suitable.

The logical decision rule looks like this:

Next, the tablet, the image of which is processed in the analytical device 18, moves from position 12 to position 25. At this position, the tablet engages with a drive roller 26, which rotates at a constant angular velocity. Through a rectangular hole 27 in the upper part of the conveying device 3, the lateral cylindrical surface of the rotating tablet is illuminated by light from the light source 28 at an angle φ to the normal to the generatrix of the side surface. The image of the illuminated strip of the side surface using the lens 29 is projected onto the photosensitive area of the linear multi-element photodetector 30 of the one-dimensional receiving camera 31. The signal from the camera 31 is fed to an analytical device 18, where a digital scan image of the entire cylindrical surface of the tablet is formed from a sequence of lines. An example of a scan image of a cylindrical surface 32 with defects 19, 20, 23 and 24 (crack) is shown in Fig.4. The scan image of the cylindrical surface is processed in the analytical device 18 in the process of moving the tablet from position 25 to position 33.

The defect search algorithm on the side surfaces analyzes black-and-white images with size h × g pixels with 256 shades of gray. The image of the tablet in these frames is a bright rectangle formed by scan lines along the cylinder generatrix of the side surface of the tablet. The area on the frame, not occupied by the analyzed image, is a dark area.

The processing algorithm is divided into the following steps.

1. The search for the borders of the side surface inside the frame is required to clarify the position of the vertical borders of the scan image in the frame. These boundaries correspond to the corner points separating the side surfaces from the ends of the tablet. To search for boundaries, a one-dimensional function of average pixel brightness of each vertical line of the frame E (x) is determined. The average brightness value over the whole frame E ₀ and the brightness variance D _{0 are} calculated.

Initially, approximate positions of the image boundaries are found. To determine the left border, the values of the function E (x) are analyzed when moving along the x axis from the left border of the frame to its center. An approximate left border is the first point x ₁₀ at which the brightness exceeded the threshold P ₀ = E ₀ + D ₀ × k ₀ . Similarly, the approximate right boundary is found - the point x _r0 . The value of the coefficient k ₀ is selected when setting up the system.

Knowing the approximate positions of the boundaries, we can proceed to refine them. To do this, the value of the average image brightness E ₁ and the image brightness variance D _{1 are} determined from the x-axis from the x ₁₀ to the x _r0 point. If you move along the x axis from the center of the image to its edges, then you will find points x ₁₁ and x _r1 , in which the values of the function E (x) for the first time become less than the threshold value P ₁ = E ₁ -D ₁ × k ₁ . The value of the coefficient k ₁ is selected when setting up the system. The found points x ₁₁ and x _r1 are taken as the specified position of the left and right border of the image.

2. Screening by image width. If the distance between the left and right borders does not fit into the given restrictions, then the tablet is considered non-standard or defective and rejected.

3. Rejection by slipping tablets. It is believed that when shooting the image there was slippage if there is a scan section consisting of more than 10 horizontal lines of equal brightness. Two adjacent lines are considered the same if the sum of the squares of the differences in brightness of their vertically adjacent pixels is less than a certain threshold. When slipping, not the entire side surface is represented in the image. There may be a defect on the invisible part of the surface. For this reason, the tablet cannot be considered guaranteed to be defect-free and it is rejected.

4. Isolation of dark defects. At this stage, dark defects are distinguished, which usually correspond to chips. To isolate them, the threshold filtering method is used. An entire surface with a brightness below a certain threshold is considered defective. The threshold is calculated based on the average value and variance of the sweep brightness: P ₂ = E ₁ -D ₁ × k ₂ , where k ₂ is the algorithm parameter.

5. Isolation of light defects. At this stage, light defects are highlighted. Isolation occurs by threshold filtration. An entire surface with a brightness above a certain threshold is considered defective. The threshold is calculated based on the average value and the variance of brightness: P ₃ = E ₁ + D ₁ × k ₃ , where k ₃ is a parameter of the algorithm.

6. Determining the size of the defect. The boundaries and nature of the defect are determined by the same method that was used when analyzing the image of the ends.

7. Exclusion of minor and false defects. At this stage, too small defects and defects whose contrast is less than a certain threshold are excluded from further consideration. The defect contrast in this case is defined as the difference between the average brightness of the defect and the remaining part of the describing rectangle.

8. Exclusion of nested defects. If the describing rectangle of one defect is embedded in the describing rectangle of another defect, then the smaller defect is considered to be embedded, and is excluded from further consideration.

9. Search for cracks. Dark defects whose contours do not have common points with the image boundaries and for which the length to width ratio exceeds k ₄ (algorithm parameter) are considered cracks.

10. Rejection. Let the area of the ith defect be equal to s _i , the allowable area of an individual defect is equal to s _о , the total area of all p defects is S _p , the allowable total area of all defects is S _p0 , the length of the jth crack is l _j , the allowable length of an individual crack is l _o , the total length of all q cracks is L _q , the allowable total length of all cracks is L _q0 . The following logical decision rule is used to reject for cracks and other defects.

At position 33, the image of the second end surface 34 of the tablet is recorded using the illuminator 35, lens 36, camera 37 with a two-dimensional array photodetector 38. Next, when the tablet moves to position 39 of the transporting device 3, information from the camera 37 is processed in the analytical device 18 as described earlier for the end surface 13 of the algorithm. By the time the tablet arrives at position 39, a decision is made on the tablet's suitability in the analytical device and a control signal is issued to the rejector 40, which sends the controlled tablet to the appropriate receiving container (41 for rejects, 42 for suitable tablets). Thus, the described device implements the principle of conveyor processing with parallelizing the processes of moving the tablet, receiving and processing video information.

Hitachi CCD cameras (model KR-M1AP) with a resolution of 795 × 596 pixels with a Leutron Vision Pie Port-Mono-H4 controller (WWW.leutron.com) can be used as photodetector cameras 17, 37. As the linear camera 31, the VS-Ld-751 camera of NPK Videoscan CJSC (WWW.VIDEOSCAN.RU) with the VS-2001 controller can be used. As the lenses 15, 36, you can use the lens "World 38B", as the lens 29 lens "Jupiter 3 (15/50)." The analytical device can be implemented on a Pentium-4 computer. All three camera controllers are inserted into the computer. In illuminators 14, 28, 35 it is possible to use AL 336K LEDs (and AO. 336. 364 TU). The angles α and φ can be approximately 15 and 35 degrees, respectively. All drives of the moving parts of the device can be implemented on DSHI-200 stepper motors (Ya2M3. 595.057 TU) controlled by a controller based on the ATMEL AT 8958252 microprocessor. Computer and controller are connected via RS 232 channel.

Claims (1)

A method for detecting surface defects of cylindrical objects, namely, that the controlled object is sequentially fed to the surface control position, the lateral surface control position is illuminated by the radiation flux, the side surface of the controlled object is rotated during its rotation, the radiation reflected from the side surface is received by the radiation receiver, at the control position end surfaces illuminate with a stream of radiation end surfaces of the controlled object, receive reflected radiation from end surfaces with radiation detectors, they process the images received from the receivers in an analytical device, characterized in that they illuminate the side surface of the object being monitored with a radiation stream directed at an angle φ to the normal to its surface, receive reflected radiation at an angle to the normal equal to the angle of incidence of the irradiating stream ; illuminate the end surfaces of the controlled object with a radiation flux directed at an angle α to the normal to the end surface, receive radiation reflected from the end surfaces at an angle to the normal equal to the angle of incidence of the irradiating flux; in the analytical device, determine the boundaries of the images of the surfaces of the object in the image frames by the method of bypassing the boundaries, find the defective sections of the surfaces on the images of the surfaces by the method of bypassing the boundaries, describe the defective surfaces by geometric figures, calculate the areas of these figures, determine the type of defects and make a decision based on logical decision rules about the suitability of the controlled facility.

Images