Inspection system and method for automatically detecting, locating and characterizing defects on the surface of nuclear fuel rods

Abstract

System and procedure for the inspection of the surface (7) of a nuclear fuel rod (2) to automatically detect, locate and characterize defects comprising a plurality of acquisition systems (1) of point clouds and a translation device for the nuclear fuel rod (2); wherein the system further comprises a processing device capable of transforming the point clouds into height maps; segmenting the height maps and obtaining discontinuities in the height maps; classifying the discontinuities of different geometries to obtain area, length, width, position and orientation; projecting the defects on the Y axis to obtain the depth of the defects; and calculating the equivalent function and its second derivative to obtain the point where the depth of the defects is maximum.

Description

Technical Field

[0001] This application relates to the field of manufacturing nuclear fission fuel for nuclear reactors, and more specifically, to the field of testing equipment for fresh fuel rods (non-irradiated) or tubes made of zirconium alloy. Background Technology

[0002] In a nuclear reactor, fuel elements consisting of multiple nuclear fuel rods are used, arranged in rows and columns in a matrix. The fuel rods are made of a metallic alloy containing more than 95% zirconium and are characterized by a highly mirror-like appearance. The fuel rods comprise fuel pellets, in which case uranium (U) (in the form of uranium oxide (UO2)) is typically... 235 U is enriched at up to 4.9%.

[0003] To avoid potential reactor malfunctions, leaks of radioactive materials (uranium and other fission products) in the nuclear power plant loop, and the resulting loss of productivity and increased radioactive risks, fuel rod manufacturing imposes quality requirements on various aspects of the fuel rod surface, including notches, cracks, and pores. The most restrictive conditions occur in certain fuel rods used in PWR reactors, where the maximum permissible depth of defects is 25 μm and the surface area is 2 mm². 2 .

[0004] During manufacturing, the traditional method for verifying whether these quality requirements are met is a visual inspection of the entire bar surface by a qualified inspector. While this inspector uses a visual model, the inspection is still subjective, and defects are judged based on their appearance. Of course, manual inspection lacks the precision required to characterize deep and area defects.

[0005] Another approach is to use eddy current devices to identify whether the rod has defects and their location on the rod (without characterizing the defects). Currently, using these systems, the defect detection accuracy is about 50 µm, but it is not possible to characterize the defects in terms of depth, area, and orientation.

[0006] The nuclear fuel manufacturing industry has shown extensive interest in technologies for characterizing defects on the surface of fuel rods. However, attempts to date have not yielded the expected results, and there is no evidence that any system can provide information and in-depth characterization of discovered defects with sufficient accuracy, beyond reliably and automatically inspecting the entire rod surface. Summary of the Invention

[0007] The purpose of this application is to provide a system and program that, in addition to characterizing surface defects in nuclear fuel rods, can automatically detect and locate surface defects in nuclear fuel rods. To this end, the system proposes a method for translating fuel rods and multiple systems for acquiring point clouds of the fuel rod surface, wherein each acquisition system consists of a laser, a camera, and a point acquisition card. This application also includes a processing device capable of converting the point cloud into a height map, performing local thresholding on segments of the height map to obtain discontinuities in the height map; classifying discontinuities of different geometries to obtain area, length, width, location, and orientation; projecting the defects onto the Y-axis to obtain the depth of the defects; and calculating an equivalent function and its second derivative to obtain the point where the depth of the defects is maximum. The processor can also selectively eliminate voids, noise, and / or edges, discontinuities caused by vibration or reflection. Attached Figure Description

[0008] To aid in a better understanding of the features of this application and to supplement this specification, the following figures are attached as an integral part thereof, and their features are illustrative rather than limiting: Figure 1A , Figure 1B and Figure 1C A view of the point cloud acquisition system according to this application is shown.

[0009] Figure 2A and Figure 2B The data acquisition system associated with the rod and its coordinate system is shown.

[0010] Figure 3A and Figure 3B Perspective and front views of an embodiment of a point cloud acquisition system with four points are shown.

[0011] Figure 4 A perspective view of the rod translation system is shown. In the illustrated embodiment, the system includes a motorized wheel, although other translation devices may also be used.

[0012] Figure 5 This is a flowchart of the method described in this application.

[0013] Figure 6A and Figure 6B The X and Y axes of the defect are shown, as well as the projection onto the Y axis of the longitudinal defect, and the minimum value is obtained by the second derivative.

[0014] Figure 7 It shows in Figure 6B The cut formed at the minimum value, and the depth obtained as the height of the defective rectangle increases. Detailed Implementation

[0015] The previously exposed problems are satisfactorily resolved through the process and system proposed in this application.

[0016] This application relates to a method and system for detecting, locating, and characterizing the width, length, and depth of defects on the surface of unirradiated fission fuel rods. The inspection utilizes 3D vision technology based on laser triangulation. A number of point cloud acquisition systems (1) are used, pre-calibrated, and adjusted; each system consists of a linear laser beam (3), an optical sensor (4), and an acquisition card (5).

[0017] The acquisition system operates by having a linear laser beam (3) fall on an object (2) previously placed in the field of view at a certain time point, causing the object (2) to diffusely reflect the laser light. At the same time, the light is received by an optical sensor (4) at a given distance and angle. The sensor takes a photograph, obtaining the intensity distribution of the diffuse light reflected by the laser from the photograph and extracting the contour (6). It also extracts a 2D image from the photograph and finally extracts the time (called a timestamp). This information is stored in a three-channel RGB image (red: contour; green: 2D image; blue: timestamp). As the fuel rod moves, the photographing process is repeated over time, thereby generating contours along the fuel rod axis (Y-axis). These contours form a point cloud (7). Finally, in order to process the point cloud into a two-dimensional image, it is converted into a height map so that each point of the cloud corresponds to a pixel, so that the X and Y coordinates of the cloud correspond to the height and width of the image, and the Z coordinate of the cloud corresponds to a gray level from 0 to 255.

[0018] Preferably, the laser type is blue light, which has a shorter wavelength (approximately 470 nm), thus having less penetrating power and a larger focal point compared to other types of lasers (e.g., red light (approximately 700 nm)). This minimizes the light reflected back to the detector, resulting in three times less noise than a red detector and significantly improved accuracy.

[0019] According to this application, the point cloud (7) of the surface is processed by a process (PC, industrial computer, tablet computer, or any device with a processor) and by a series of specially developed artificial vision operations. According to the method of this application, defects on the surface of the rod (2) are automatically detected, measured, and classified. For this purpose, a reference system is used such that the axis of the cylinder constituting the rod is the Y-axis, and the axes transverse to it at each moment are the X-axis and Z-axis. Figure 2B According to this application, the fuel rod moves linearly along the Y-axis ( Figure 4The acquisition system is positioned laterally, with the linear laser beam corresponding to the XZ plane. Therefore, the profile (6) obtained by each optical sensor (4) of the acquisition system (1) at each moment will be a circumferential arc within the sensor's field of view, corresponding to the cross-section of a rod defined by the XZ plane coinciding with the laser beam at that particular moment. In the XZ plane containing the laser beam, the camera's field of view is a trapezoid, which corresponds to the triangle of the field of view of the optical sensor (rectangle).

[0020] During the linear motion of the rod, each acquisition system acquires a profile with a specific frequency, and the realization of these profiles along the Y-axis forms a point cloud corresponding to the surface of the rod. Figure 1C According to this application, a limited number of systems are used to cover the entire surface of a rod. Since the rod itself does not rotate, a number of point clouds corresponding to the number of systems is obtained, and the combination of all point clouds covers the entire surface (360°) of the rod.

[0021] Furthermore, according to this application, a data acquisition system with a defined frequency obtains profiles that accumulate one after another as the rod moves, and the distance between the profiles is a function of the rod's velocity.

[0022] The depth of the defects to be detected and measured will be greater than or equal to 25 µm; therefore, the ideal Z-accuracy of the measuring equipment will be on the order of 0.5 µm (25 / 50). Depending on the nuclear model to be inspected (PWR, VVER, or BWR), the diameter of the rod (Ørod) varies from 8.9 to 10.7 mm, and the length from 1500 to 4510 mm. For equivalent inspection and according to this application, the number of acquisition systems (minimum number 3) for 360° inspection of cylindrical rods will be calculated as follows: Adq.Syst. = f (Ørod) = [ 3 + 0.1 Ørod ] According to this application, considering that the acquisition systems are identical or have similar characteristics, this formula is valid for a given value of Z accuracy (0.5 μm). In determining this formula, in addition to the diameter, the sensor's field of view and the rod's velocity are also considered.

[0023] Each acquisition system has a finite field of view, which varies depending on the Z-axis accuracy. Higher accuracy results in a smaller field of view and a smaller trapezoidal size constituting the field of view. This occurs in a similar manner to speed; the ends of the curves forming the contour are not perpendicular to the acquisition system, requiring longer exposure times than for flat and perpendicular surfaces. Therefore, longer exposure times result in shorter sampling rates. This creates a maximum speed that can be checked. Below this speed, at lower speeds, more contours will be obtained, leading to higher accuracy of the point cloud on the Y-axis.

[0024] Thus, according to the previous formula, the number of fuel rod (PWR, VVER, or BWR) collection systems with a diameter between 9.5 and 10.7 mm will be at least 4. Figure 3A and Figure 3B ): Adq.Syst. = [ 3 + 0.1 × 10.7 ] = [ 4.07 ] = 4 This application may also include a support structure for the acquisition system through which the fuel rod passes. At the inlet and outlet of the structure, there may be separate conical nylon centering devices to consistently control the distance between the acquisition system and the rod during inspection and to limit unwanted vibrations. The structure is robust enough that the position of the acquisition system will not change; the acquisition system will be placed and tightened (or secured by a similar method) to a cross-sectional structure transverse to the axis of the rod, in the same plane, but spaced 90° apart from each other (in the case of four). Proper positioning of the acquisition system relative to the rod will allow for measurements with acceptable quality and reliability. In a preferred example, the structure is designed to dissipate as much heat as possible generated by the operating acquisition system through convection; the structure is made of a conductive material such as aluminum and has as few air chambers as possible.

[0025] Furthermore, the structure may include a system for adjustment and installation (using gauges, micrometers, etc.) to significantly alter the vertical distance between the acquisition system and the rod. It may also have a similar system for adjusting the entire support structure and acquisition system, and can be rotated to adjust the alignment of the entire assembly relative to the rod.

[0026] For linear movement of the bar, the device may include a motorized roller or belt system on which the bar rests. Figure 4 According to this application, the longitudinal velocity component of the velocity is obtained via an idler-wheel probe located closest to the acquisition system, which always replicates the motion of the rod and is positioned at the location of the encoder used to measure the velocity of the rod. This is necessary to establish a sampling frequency so that the sampling frequency is constant across the entire rod and constructs a point cloud in a uniform manner.

[0027] This application also includes a specific hardware system for coordinating, synchronizing, and communicating between the acquisition systems. This specific hardware system is connected to the acquisition systems. The hardware provides single-point power supply, security, triggering, and digital signals, and is essential for the proper operation of the system. Finally, the encoder and processing device (PC, industrial computer, or tablet computer) are connected to the synchronization hardware to perform point cloud processing. This synchronization hardware outputs N point clouds (one point cloud for each acquisition system, or a joint point cloud connecting some clouds with other clouds in overlapping areas).

[0028] The acquired information is a 16-bit root image with 3 channels (red, green, and blue), which stores all 2D and 3D information.

[0029] According to the embodiments of this application, in Figure 5 The diagram shows the operations performed in the processor.

[0030] The following describes in detail each level according to this application and its various implementations, such as Figure 5 As shown.

[0031] From stage A to stage E: A series of procedures can be used to control the acquisition system via software and decompose the raw image of 3 channels per pixel (RBG) acquired by the optical sensor (4) and the acquisition card (5) into three images (one image per channel): (1) height map, (2) 2D intensity image (2D photo summary); and (3) timestamp.

[0032] Stage F: Eliminating voids, noise, and edges. Using the input heightmap, the processor can optionally perform noise reduction operations, identifying discontinuities with area values ​​smaller than the area of ​​potential actual defects in the bar. Holes with areas smaller than defined areas can also be filled using the average value of the pixels surrounding the hole. Furthermore, edges can be smoothed, eroded, and the image made more uniform. While not mandatory, these steps have the advantage of eliminating future false defects and accelerating subsequent image processing and defect characterization.

[0033] Stage G: Smoothing and Contrast Enhancement. The processor will be able to smooth and / or flatten the projection of the heightmap, which, in the case of a cylinder, is projected as an equivalent of a circle. For this purpose, gamma correction is applied, which reduces curvature and increases contrast.

[0034] Stage H: Segmentation. Segmentation comprises the location of regions (sets of pixels, points, regions, or portions) of the image that have the same features or different features from their neighboring regions. With the heightmap corrected and contrast increased, the processor performs a thresholding process to detect discontinuities in the heightmap. It is necessary to apply local or adaptive thresholding because the background of the cylinder's heightmap is not constant (in this case, it does not have the same Z-axis because it is curved) and the contrast of the defects varies in the image. Therefore, standard thresholding works well for a portion of the image, but segmentation is unsatisfactory for the rest. Thus, it is necessary to examine the image intensity near each pixel (or group of pixels) and select a threshold by averaging the local intensity distribution obtained from the smoothed image. This yields "blob" regions, which will be the assumed defect regions of the bar. This segmentation step is important in this application because, through this step, the defect is extracted as a function of the defect depth, and defect detection and localization are the objectives of this application.

[0035] Phase I: Noise Elimination and Hole Filling. First, from the original area of ​​the defect, eliminate "spots" smaller than possible real defects, discard false defects, and fill the small internal holes that make up the "spots" in the area.

[0036] Stage J: Vibration Elimination. Advantageously, the processor will be able to eliminate vibrations of the rod (if any). This is accomplished by laterally connecting the regions obtained from the segments (if they are a few millimeters apart) and identifying those regions (or connections of regions) that are completely perpendicular to the rod and have a length equal to or similar to the diameter as vibrations.

[0037] Stage K: Reflection Elimination. Additionally, operations can be performed to eliminate discontinuities caused by laser reflection in the rod material. This involves identifying whether the average Z-value is higher than what the rod surface should have and eliminating these discontinuities, as defects such as indentations, pores, or notches consistently have lower Z-values ​​than the rod surface. This identifies clearly defined areas of rod defects.

[0038] Stage Y: A defect-free object. If the empty region of Stage K is obtained, the process will be complete, and the result will be a defect-free, good bar.

[0039] If no empty region is obtained in stage K, the process continues to stage L.

[0040] Stages L and M: Classification and characterization of region elements. Once those discontinuities that may correspond to actual defects in the rod have been identified, the processor can classify them as longitudinal, circular, or undefined orientations based on the internal and external geometries of the found discontinuities. Once the possible defects have been classified, their maximum length and width dimensions are measured, regardless of their orientation. In this way, the first defect characterization data is obtained: area, length, width, location, and orientation.

[0041] Stage N: Projection of the defect. The default axes are represented by the Y-axis (major axis of the circumscribed ellipse), the X-axis (minor axis of the ellipse), and the Z-axis (X and Y axes transverse to the bar surface). In this stage, the projection along the Y-axis is extracted. This step is fundamental in this application because the projection along the Y-axis represents the location of the defect with a greater depth. Figure 6A ).

[0042] Stage O: Transformation of the vertical projection. The equivalent function and its second derivative are obtained from the vertical projection. Therefore, the relative minimum of the horizontal projection (the location with the maximum defect depth) is given by the maximum value of the second derivative. Figure 6B ).

[0043] Stage P: The defect portion at the relative minimum. The region is sectioned using a plane perpendicular to the Y-axis at the minimum value (selecting 20 smaller regions) to obtain a cross-section of the defect in that plane, from which its depth can then be calculated. Figure 7 ).

[0044] Stage Q: Section transformation to obtain the maximum and average depth. In these sections, the depth of the defect is calculated using the height of the rectangle etched by the smoothed defect profile. Figure 7 Consider the maximum height and maximum depth, as well as the arithmetic mean of the rectangle's height and average depth.

[0045] In addition, the processor will be able to calculate the true speed of the stick using timestamps.

[0046] Finally, by independently analyzing each point cloud provided by each acquisition system or by unifying all point clouds in one or two point clouds, the processor will be able to output the following data based on the bar: the number of defects and the length, width, area, maximum depth, and location of each defect.

[0047] Based on the above description and accompanying drawings, those skilled in the art will understand that this application has been described according to some preferred embodiments, but various changes may be introduced in the preferred embodiments without departing from the purpose claimed in this application.

Claims

1. An inspection system for the surface of nuclear fuel rods, used for automatically detecting, locating, and characterizing defects, said inspection system comprising: A point cloud acquisition system (1), each of the multiple point cloud acquisition systems comprising: Linear laser beam (3). Optical sensor (4), and And point cloud acquisition card (5); Wherein, the number of acquisition systems N o The Adq.Syst. relation is given by the following formula: N o . Adq.Syst. = f (Ørod) = [ 3 + 0.1 Ørod ] Where Ørod is the diameter of the nuclear fuel rod to be tested, in millimeters, and [ ] indicates rounding to the nearest integer; A translation device for the nuclear fuel rods; wherein the inspection system further includes a processing device configured to: The point cloud is transformed into a Z-height map, where the Z coordinate is perpendicular to the Y-axis, which coincides with the vertical axis of the nuclear fuel rod. Local thresholding is performed on segments of the height map to obtain discontinuities in the height map; Discontinuities of different geometric shapes are classified to obtain their area, length, width, position, and orientation. The defect is projected onto the Y-axis to obtain the depth of the defect. Starting from the projection of the defect onto the Y-axis, calculate the equivalent function and its second derivative to obtain the point where the depth of the defect is at its maximum.

2. The inspection system according to claim 1, wherein, After transforming to a height map, the processing device eliminates gaps, noise, and / or edges.

3. The inspection system according to claim 1 or 2, wherein, The processing device eliminates the interruptions caused by vibration.

4. The inspection system according to claim 1, wherein, The processing device calculates whether the average value of Z is higher than the average value of the surface of the nuclear fuel rod, thereby identifying and eliminating reflections.

5. A method for inspecting the surface of nuclear fuel rods, used for automatically detecting, locating, and characterizing defects, the inspection method comprising: Point clouds are collected as the fuel rod moves continuously along its longitudinal axis Y. The point cloud is transformed into a Z-height map, where the Z coordinate is perpendicular to the Y-axis, which is the same as the vertical axis of the nuclear fuel rod. Local thresholding is performed on segments of the height map to obtain discontinuities in the height map; Discontinuities of different geometric shapes are classified to obtain their area, length, width, position, and orientation. The defect is projected onto the Y-axis to obtain the depth of the defect. Starting from the projection of the defect onto the Y-axis, calculate the equivalent function and its second derivative to obtain the point where the depth of the defect is at its maximum.

6. The inspection method according to claim 5 further includes: Eliminate gaps, noise, and / or edges.

7. The inspection method according to claim 5 or 6 further includes: Eliminate the interruptions caused by vibration.

8. The inspection method according to claim 5 further includes: The average value of Z is calculated to determine whether it is higher than the average value of the surface of the nuclear fuel rod, in order to identify and eliminate reflections.

Images