Neutron collimator flexible diaphragm measurement methods, program products and electronic equipment

Abstract

This application provides a method, program product, and electronic device for measuring the flexible diaphragm of a neutron collimator. The method includes: acquiring multi-angle images of the flexible diaphragm of the neutron collimator; mapping the multi-angle images to a three-dimensional spatial coordinate system based on pixel information and camera parameters to obtain multi-dimensional vectors corresponding to pixels in the multi-angle images; using the multi-dimensional vectors to characterize the spatial position and light direction of the pixels; generating a three-dimensional point cloud model using the multi-dimensional vectors and a trained neural radiation field model; generating a digital twin model of the flexible diaphragm based on the three-dimensional point cloud model; the digital twin model is an information model based on the virtual mapping of the flexible diaphragm; and measuring the flexible diaphragm using the digital twin model. The synergistic technology chain of the neural radiation field model and digital twin improves the accuracy and efficiency of angle measurement of the flexible diaphragm in the neutron collimator.

Description

Technical Field

[0001] This application relates to the field of instrument measurement technology, and more specifically, to a method, program product, and electronic equipment for measuring a flexible diaphragm of a neutron collimator. Background Technology

[0002] A neutron fine collimator consists of an array of hundreds to thousands of flexible neutron-absorbing films. The angular tolerance of each film layer must be strictly controlled to meet the precision requirements of neutron scattering experiments for resolving interatomic spacing. Current methods for measuring the angle of flexible films require manual calibration, which is inefficient and inaccurate. Summary of the Invention

[0003] The purpose of this application is to provide a method, program product, electronic device and storage medium for measuring a flexible diaphragm of a neutron collimator, in order to improve the above-mentioned problems.

[0004] In a first aspect, embodiments of this application provide a method for measuring a flexible diaphragm of a neutron collimator, comprising: acquiring multi-angle images of the flexible diaphragm of the neutron collimator; mapping the multi-angle images to a three-dimensional spatial coordinate system based on pixel information and camera parameters to obtain multi-dimensional vectors corresponding to pixels in the multi-angle images; the multi-dimensional vectors being used to characterize the spatial position and light direction corresponding to the pixels; generating a three-dimensional point cloud model using the multi-dimensional vectors and a trained neural radiation field model; the three-dimensional point cloud model being used to characterize the sampling point information corresponding to the light rays; generating a digital twin model of the flexible diaphragm based on the three-dimensional point cloud model; the digital twin model being an information model based on the virtual mapping of the flexible diaphragm; and measuring the flexible diaphragm using the digital twin model.

[0005] In the aforementioned implementation process, a high-precision 3D point cloud model is generated based on a multi-angle camera using a neural radiation field model. This 3D point cloud model can reconstruct the surface morphology of the diaphragm, improving the data loss problem caused by the poor light transmittance and reflectivity of the diaphragm in traditional optical methods. Digital twin technology maps the state of the real physical device to a virtual model, supporting real-time monitoring of the flexible diaphragm's angle changes. The synergistic technology chain of the neural radiation field model and digital twin not only improves the accuracy of the neutron fine collimator's flexible diaphragm angle measurement but also greatly enhances its efficiency.

[0006] Optionally, in this embodiment, the pixel information includes the pixel coordinates of the pixel points; based on the pixel information of the multi-angle image and the camera parameters, the multi-angle image is mapped to a three-dimensional spatial coordinate system to obtain the multi-dimensional vector corresponding to the pixel points in the multi-angle image, including: acquiring the camera parameters for acquiring the multi-angle image and the pixel coordinates of the pixel points in the multi-angle image; the camera parameters include camera intrinsic parameters and camera extrinsic parameters; calculating the three-dimensional direction vector of the light rays using the camera intrinsic parameters and pixel coordinates; determining the light ray direction of the pixel point based on the three-dimensional direction vector of the light rays; determining the spatial position corresponding to the pixel point based on the three-dimensional direction vector of the light rays and the camera extrinsic parameters; and obtaining the multi-dimensional vector corresponding to the pixel point based on the light ray direction and spatial position of the pixel point.

[0007] In the above implementation process, the processed and pre-processed multi-angle image data is transformed into multi-dimensional vectors that meet the requirements of the neural radiation field model. This allows the spatial location and viewpoint information in the images to be used as input to generate three-dimensional reconstruction data of the flexible membrane. This process provides accurate input data for the neural radiation field model, enabling it to accurately reconstruct the three-dimensional point cloud model of the flexible membrane based on two-dimensional image information from multiple viewpoints.

[0008] Optionally, in this embodiment of the application, a three-dimensional point cloud model is generated using multidimensional vectors and a trained neural radiation field model, including: inputting the multidimensional vectors corresponding to the pixels into the neural radiation field model to obtain the parameter information of the pixels; the parameter information includes color and density; and generating a three-dimensional point cloud model using volumetric rendering technology based on the parameter information of the pixels; the volumetric rendering technology is used to simulate the changes in light propagation in three-dimensional space.

[0009] In the aforementioned implementation process, volumetric rendering technology simulates the real light propagation process based on the parameter information of pixels. Multiple spatial points are sampled along each light path, and the transmittance and absorptivity of the light passing through each point are calculated based on density parameters, while simultaneously accumulating color contribution values. Through integration, the discrete sampled points are restored to a continuous spatial material distribution, ultimately constructing a 3D point cloud model that combines geometric accuracy and optical realism. This approach is more suitable for flexible, semi-transparent objects such as membranes, accurately simulating the light scattering path and energy deposition distribution under neutron radiation. It provides an atomically accurate spatial data base for subsequent digital twins, thereby improving the accuracy of flexible membrane measurements.

[0010] Optionally, in this embodiment, the neural radiation field model includes a vector input layer, a position encoding layer, a feature extraction layer, a ray projection and sampling layer, and an output layer. The vector input layer is used to convert multi-dimensional vectors into feature representations. The position encoding layer is used to encode the feature representations to obtain encoded spatial coordinates. The feature extraction layer is used to extract features from the encoded spatial coordinates to generate feature data. The feature extraction layer includes multiple fully connected layers. The fully connected layers include an occlusion attention module, which dynamically masks the features of the occluded area using the membrane spacing data. The ray projection and sampling layer is used to sample the ray path corresponding to the pixel multiple times to generate multiple sampling points. The output layer is used to output the parameter information of the sampling points. The parameter information of the sampling points is used to generate a three-dimensional point cloud model.

[0011] In the aforementioned implementation process, the neural radiation field model adopts a layered architecture design to achieve accurate digital reconstruction of the complex light propagation process in the real world. The vector input layer is compatible with the three-dimensional coordinates and light direction parameters of physical space, enabling the model to naturally interface with raw data collected by various sensors. The position encoding layer uses high-frequency signal conversion technology to transform smooth spatial position and direction information into detailed mathematical expressions, effectively improving the model's ability to perceive minute structural features, such as micron-level wrinkles on the membrane surface or the gradient effect of a light-transmitting coating. The feature extraction layer, as the core of the deep neural network, extracts the abstract laws of light-matter interaction from the encoded data through multi-layer nonlinear transformations, including implicit physical field information such as material density distribution and scattering characteristics. The light projection and sampling layer maps the abstract features of the preceding steps back to a physically interpretable dimension. This layer intelligently generates density-adaptive sampling points along each light path, automatically densifying the point cloud in the membrane entity area and sparsely sampling in open areas, significantly improving computational efficiency. The final output layer simultaneously generates dual-channel results for color and density, satisfying both visual rendering requirements and providing geometric existence criteria for subsequent point cloud reconstruction.

[0012] Optionally, in this embodiment, generating a digital twin model of the flexible diaphragm based on the three-dimensional point cloud model includes: geometrically repairing the three-dimensional point cloud model after format conversion to obtain a repaired model; geometrical repair includes light transmission compensation and curvature optimization; assigning physical properties to the repaired model to obtain a physical entity model; physical property assignment includes material modeling and dynamic property configuration; performing virtual environment modeling on the physical entity model to obtain the object to be simulated; virtual environment modeling includes neutron radiation field simulation and embedded radiation thermal effects; and performing multiphysics coupling based on the parameters of the object to be simulated to generate a digital twin model of the flexible diaphragm; multiphysics coupling includes mechanical-thermal coupling and boundary conditions.

[0013] In the aforementioned implementation process, format conversion resolves data compatibility issues, geometric repair strip model computability is ensured, physical property assignment imparts material realism, virtual environment modeling loads external stimuli, and multiphysics coupling integrates all elements for behavioral inference. For example, after loading a neutron radiation field, the repaired geometric model can predict edge warping of 0.1 mm through thermo-mechanical coupling simulation, thereby guiding the implementation of multi-density monitoring of this area in actual measurements. This transforms the traditional R&D model, which relies on physical trial and error, into an efficient iteration in digital space, shortening the R&D cycle and reducing trial and error costs. Furthermore, coupled simulation can predict the failure risk of the diaphragm under extreme conditions in advance.

[0014] Optionally, in this embodiment of the application, after generating a digital twin model of the flexible diaphragm based on the three-dimensional point cloud model, the method further includes: performing static performance simulation, dynamic performance simulation, and multiphysics coupling simulation on the digital twin model in sequence to generate a measurement strategy; the static performance simulation is used to simulate the deformation and stress distribution of the diaphragm under working conditions; the static performance simulation includes deformation analysis and stress concentration location analysis under neutron radiation; the dynamic performance simulation is used to analyze the vibration and fatigue behavior of the diaphragm under dynamic loads; the dynamic performance simulation includes vibration mode analysis and fatigue life prediction; the multiphysics coupling simulation is used to comprehensively analyze the interaction between heat, force, and radiation fields; the multiphysics coupling simulation includes thermo-mechanical coupling analysis and radiation material performance degradation; measuring the flexible diaphragm through the digital twin model includes: measuring the flexible diaphragm through the digital twin model according to the measurement strategy.

[0015] In the above implementation process, static performance simulation first provides a baseline snapshot in the spatial dimension, accurately locating high-stress areas and deformation over-limit points under fixed working conditions; dynamic performance simulation then introduces the time dimension, capturing transient response patterns through vibration mode and fatigue life analysis; multi-physics coupling simulation finally integrates the interactions of heat, force, radiation, etc., to accurately predict the performance limits of the diaphragm in extreme environments, providing data support and key decision support for subsequent measurements.

[0016] Optionally, in this embodiment of the application, measuring the flexible diaphragm using a digital twin model includes: simulating the geometry of the flexible diaphragm under different working conditions using the digital twin model, measuring the angle of the flexible diaphragm in virtual space, and obtaining measurement data; the measurement data includes the angle of the flexible diaphragm relative to a reference plane, and the relative angle changes between different parts of the flexible diaphragm; after measuring the flexible diaphragm using the digital twin model, the method further includes: obtaining the measurement data of the flexible diaphragm, analyzing the measurement data, generating a diaphragm angle correction scheme; and correcting the measurement data using the diaphragm angle correction scheme.

[0017] In the above implementation process, the angle deviation of the flexible diaphragm can be identified and corrected by comparing and analyzing the angle measurement of the digital twin with the actual data, so that the diaphragm can always maintain the predetermined angle accuracy during operation.

[0018] Optionally, in this embodiment of the application, the measurement of the flexible diaphragm using a digital twin model includes: real-time monitoring of the flexible diaphragm and uploading the detection results to the control system; if an angle shift of the flexible diaphragm is detected, the position or external force of the flexible diaphragm is automatically adjusted by the control system.

[0019] In the aforementioned implementation process, the angle correction scheme generated in the digital twin is fed back to the actual equipment in real time through the control system. This enables the flexible diaphragm to maintain an ideal angle during actual operation and makes dynamic corrections based on real-time data. This process improves the accuracy and efficiency of angle correction, ensuring long-term stable operation of the equipment.

[0020] Secondly, embodiments of this application also provide a measurement device for a flexible diaphragm of a neutron collimator, comprising: an image acquisition module for acquiring multi-angle images of the flexible diaphragm of the neutron collimator; a multi-dimensional vector conversion module for mapping the multi-angle images to a three-dimensional spatial coordinate system based on pixel information and camera parameters, thereby obtaining multi-dimensional vectors corresponding to pixels in the multi-angle images; the multi-dimensional vectors are used to characterize the spatial position and light direction corresponding to the pixels; a point cloud module for generating a three-dimensional point cloud model using the multi-dimensional vectors and a trained neural radiation field model; the three-dimensional point cloud model is used to characterize the sampling point information corresponding to the light rays; a digital twin module for generating a digital twin model of the flexible diaphragm based on the three-dimensional point cloud model; the digital twin model is an information model based on the virtual mapping of the flexible diaphragm; and a measurement module for measuring the flexible diaphragm using the digital twin model.

[0021] Thirdly, embodiments of this application also provide a computer program product, including computer program instructions, which are executed by a processor to perform the method provided in the first aspect or any implementation thereof.

[0022] Fourthly, embodiments of this application also provide an electronic device, including: a processor and a memory, the memory storing computer program instructions, which are executed by the processor to perform the method provided in the first aspect or any implementation thereof.

[0023] Fifthly, embodiments of this application also provide a computer-readable storage medium storing computer program instructions, which, when executed by a processor, perform the method provided in the first aspect or any implementation thereof.

[0024] This application discloses a method, program product, electronic device, and storage medium for measuring the flexible diaphragm of a neutron collimator. A high-precision three-dimensional point cloud model is generated based on a multi-angle camera using a neural radiation field model. This three-dimensional point cloud model can reconstruct the diaphragm surface morphology, improving upon the data loss problem caused by poor light transmission and reflectivity of the diaphragm in traditional optical methods. Digital twin technology maps the virtual model to the real-world physical device state, supporting real-time monitoring of the flexible diaphragm angle changes. The synergistic technology chain of the neural radiation field model and digital twin not only improves the accuracy of the neutron fine collimator flexible diaphragm angle measurement but also significantly enhances its efficiency. Attached Figure Description

[0025] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 A schematic flowchart illustrating a method for measuring a flexible diaphragm in a neutron collimator, provided in an embodiment of this application;

[0027] Figure 2 This is a schematic diagram of a multi-angle image acquisition process provided in an embodiment of this application;

[0028] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0029] The embodiments of the technical solution of this application will now be described in detail with reference to the accompanying drawings. These embodiments are only used to more clearly illustrate the technical solution of this application and are therefore merely examples, and should not be used to limit the scope of protection of this application.

[0030] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit this application.

[0031] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly defined.

[0032] As a core component of neutron scattering spectrometers (such as the China Spallation Neutron Source (CSNS), the UK Spallation Neutron Source (ISIS), the US Spallation Neutron Source (SNS), and the Japan Spallation Neutron Source (J-PARC), the neutron fine collimator, by constraining the neutron optical path (initial divergence angle greater than or equal to 1 degree, collimated and constrained to less than or equal to 0.1 degrees) and defining the sample measurement volume, improves the spectrometer resolution from the nanometer level (1-10 nanometers) to the sub-angstrom level, and improves the experimental signal-to-noise ratio by 2-3 orders of magnitude (from 10:1 to 300:1 in actual measurements). The device consists of an array of hundreds to thousands of flexible neutron absorbing films. Each film is coated on both sides with a polyimide organic substrate (thickness 20-75 micrometers, transmittance greater than or equal to 90%). They are densely assembled at a micro-angle of 0 to 5 degrees with a spacing of less than or equal to 0.1 millimeters. The angular tolerance must be strictly controlled within ±0.001 degrees to ±0.05 degrees (corresponding to a dimensional error of less than or equal to 1 micrometer) to meet the requirements of neutron scattering experiments for the precision of atomic spacing resolution.

[0033] However, the unique properties of flexible diaphragms present the following technical challenges:

[0034] 1. Dynamic deformation interference: The diaphragm is subjected to non-uniform deformation due to assembly preload and neutron radiation heat load, with an angle drift Δθ greater than or equal to 0.02 degrees (40% over tolerance), and the deformation rate cannot be captured by traditional static measurement methods;

[0035] 2. Dense occlusion effect: When the inter-membrane spacing is less than or equal to 0.1 mm, more than 80% of the area is occluded by adjacent membranes. The effective data coverage of laser triangulation method (such as Keyence LK-G5000) is less than 20%, and the occlusion shadows cause topological breaks in the reconstruction model (break rate greater than 65%).

[0036] 3. Material optical defects: The surface reflectivity of the material coating is less than 5% (wavelength 532 nm), the transmittance of the substrate is greater than or equal to 90%, the signal-to-noise ratio attenuation of the optical sensor (such as Basler ace 2) is greater than 60%, and the failure rate of feature point matching is greater than 70% (compared to a success rate of greater than 95% for rigid devices).

[0037] Traditional measurement methods have systemic flaws:

[0038] 1. Indirect calculation distortion: Laser trackers (such as Leica AT960) calculate the angle by measuring the clamping spacer (thickness tolerance ±2 micrometers). Due to the relaxation of assembly stress (creep rate 0.1 micrometers / h), the actual angle of the diaphragm deviates from the calculated value by more than or equal to ±0.1 degrees (more than 2 times the tolerance).

[0039] 2. Conflict between efficiency and accuracy: A single measurement requires manual calibration of 15-20 reference points (taking more than or equal to 30 minutes per piece), and manual calibration introduces an error of ±0.05 degrees (accounting for 50% of the total error), and it is impossible to achieve synchronous measurement of the entire array (1000 pieces of array require more than or equal to 500 hours).

[0040] 3. Environmental sensitivity defects: thermal disturbances in the neutron radiation field cause the optical sensor baseline to drift by more than 10 micrometers (corresponding to an angle error of ±0.03 degrees), and humidity fluctuations (greater than or equal to 30%) cause the lens distortion rate to increase by 15%.

[0041] In this embodiment, computer vision and digital twin technology are combined to provide a new approach for measuring the angle of flexible diaphragms. NeRF (Neural Radiance Fields) technology generates a high-precision three-dimensional model from multi-view images (such as 50-100 two-dimensional images taken by an industrial camera), which can reconstruct the surface morphology of the diaphragm (with an accuracy of ±0.01 mm), overcoming the data loss problem caused by the poor light transmittance and reflectivity of the diaphragm in traditional optical methods.

[0042] This collaborative technology chain of NeRF and digital twins not only improves the accuracy of angle measurement of the flexible diaphragm in the neutron fine collimator, but also realizes the automation and real-time feedback of the measurement process, greatly improving the efficiency and accuracy of angle measurement of the flexible diaphragm in the neutron fine collimator.

[0043] Please see Figure 1 The illustration shows a flowchart of a neutron collimator flexible diaphragm measurement method provided in an embodiment of this application. The neutron collimator flexible diaphragm measurement method provided in this application can be applied to electronic devices, which may include physical devices such as servers, PCs, tablets, or smartphones, or virtual devices such as virtual machines or containers. The electronic device can be a single device, a combination of multiple devices, or a cluster of a large number of devices. The neutron collimator flexible diaphragm measurement method may include:

[0044] Step S110: Acquire multi-angle images of the flexible diaphragm of the neutron collimator.

[0045] Step S120: Based on the pixel information of the multi-angle image and the camera parameters, map the multi-angle image to a three-dimensional spatial coordinate system to obtain the multi-dimensional vector corresponding to the pixel in the multi-angle image; the multi-dimensional vector is used to represent the spatial position and light direction of the pixel.

[0046] Step S130: Generate a three-dimensional point cloud model using multidimensional vectors and a trained neural radiation field model; the three-dimensional point cloud model is used to represent the sampling point information corresponding to the light rays.

[0047] Step S140: Generate a digital twin model of the flexible membrane based on the 3D point cloud model; the digital twin model is an information model based on the virtual mapping of the flexible membrane.

[0048] Step S150: Measure the flexible diaphragm using a digital twin model.

[0049] In step S110, the first step in the three-dimensional reconstruction of the flexible membrane is image acquisition and preparation. The goal of this step is to acquire sufficiently detailed two-dimensional image data from multiple angles to provide accurate input for subsequent three-dimensional reconstruction. The images from these multiple angles can be of the same size or different sizes, recording the size and shape of the flexible membrane at different angles.

[0050] Please see Figure 2 The illustration shows a schematic diagram of a multi-angle image acquisition process provided in an embodiment of this application.

[0051] As one implementation method, the camera can be an industrial camera (such as the Basler ace 2 industrial camera) equipped with a preset lens, so that the resolution of a single pixel is less than or equal to 0.05 mm; the auxiliary equipment can integrate a ring LED polarized light source to suppress light transmission interference of the film substrate (so that the light transmittance is reduced to less than or equal to 30%).

[0052] Cameras can be mounted and arranged in a circular array: several cameras are arranged in a circle (e.g., 8 cameras arranged with a radius of 500 mm), covering the entire circumference of the membrane at preset angles (e.g., 0.1 degrees), ensuring a preset overlap rate of adjacent images, such as 70%, to reduce data loss caused by dense occlusion. The viewing angle can be 40 degrees. Of course, other array layouts can also be used, such as placing cameras in the front, back, left, and right directions. Camera attitude calibration is then performed: a laser level (error ±0.01 degrees) is used to adjust the camera pitch angle, ensuring the optical axis is perpendicular to the membrane surface (tilt error less than or equal to 0.05 degrees).

[0053] Regarding the optimization of environment and light: Light transmission suppression can be used by laying a black light-absorbing curtain on the back of the membrane (making the reflectivity less than 1%), combined with a polarizing filter (extinction ratio 1000:1) to eliminate substrate light transmission noise; and dynamic supplementary lighting is adopted, which adaptively adjusts the light source intensity based on the reflectivity of the membrane coating (reflectivity less than 5%), and extracts effective feature points (extraction rate greater than 95%) through a brightness threshold segmentation algorithm (for example, the threshold can be set to 200 lux).

[0054] Image acquisition process, for example: Multispectral imaging: Simultaneous acquisition of visible and near-infrared images to enhance coating edge contrast (signal-to-noise ratio improved by 40%); Radiation interference resistance: Under neutron radiation fields (e.g., flux greater than or equal to 1×10⁻⁶), 6 ~ 8At a density of neutron flux density (per square centimeter per second), the camera is encased in a lead shield (5 mm thick) and a real-time noise filtering algorithm (based on wavelet transform) is embedded to ensure that the image signal-to-noise ratio (SNR) is greater than or equal to 35 dB. The basic unit of neutron flux density is "per square centimeter per second," written as per square centimeter per second.

[0055] It can also calibrate and correct the camera. Geometric calibration: Using a checkerboard calibration board with preset parameters (e.g., 12×9, square size 10 mm ± 0.001 mm), the camera's intrinsic parameters (focal length, distortion coefficient) and extrinsic parameters (rotation and translation matrix) are calculated using OpenCV (a cross-platform computer vision library). After calibration, the reprojection error is less than or equal to 0.01 mm. Transmission compensation: To address the light transmittance of the film substrate, a transmission compensation algorithm is introduced during the calibration stage. The transmission noise is separated by the background difference method (the background is modeled as a Gaussian mixture model), reducing the hole generation rate to less than 5%. Dynamic distortion correction: Based on temperature sensor data (accuracy ± 0.1 degrees Celsius), the distortion caused by lens thermal expansion is corrected in real time, ensuring that the image stitching error is less than or equal to 0.005 mm.

[0056] This process ensures that the acquired images possess high-quality resolution, uniform illumination, accurate viewing angles, and corrected distortion, providing ample data support for subsequent precise 3D reconstruction using NeRF technology. It can acquire n images of the flexible membrane, including image 1 to image n. The images vary in size, recording the size and shape of the flexible membrane from different viewing angles.

[0057] In step S120, according to the input requirements of the neural radiation field model, each pixel (point) in the image needs to be converted into a 5D vector (i.e., a multidimensional vector). Each 5D vector includes the following five elements:

[0058] Spatial location ( This represents the three-dimensional spatial position of an image pixel, determined by the camera's external parameters (camera position and viewpoint). The spatial position of each pixel is calculated, indicating the specific location of that point in three-dimensional space.

[0059] Light direction ( ): Indicates the direction of light from the camera's viewpoint to the object's surface. It is the pitch angle (vertical direction). It is the azimuth angle (horizontal direction). These angles are calculated based on the camera's angle of view and pixel position.

[0060] We need to obtain basic image information and camera parameters to calculate the position and orientation of each pixel in three-dimensional space. Pixel information can be represented by two-dimensional coordinates for each pixel, denoted as (u, v), indicating the pixel position in the image.

[0061] Ray direction (denoted as d): For each pixel, we need to calculate the unit vector from the camera to the object surface (i.e., the ray direction). This is because each pixel in the image represents a ray of light projected from the camera onto the object surface.

[0062] Calculate the 3D direction vector of the light rays using camera intrinsic parameters and pixel coordinates; determine the direction of the light rays at each pixel based on the 3D direction vector of the light rays. Based on the three-dimensional direction vector of the light and the camera's extrinsic parameters, the spatial position corresponding to the pixel is determined. This yields a multidimensional vector, also known as a 5D vector.

[0063] In step S130, the constructed multidimensional vector is input into the pre-trained neural radiation field model. This model is essentially a deep learning network that calculates the optical properties (including color, brightness, and material density) of each spatial point by analyzing the position and orientation information in the vector. For example, the model determines whether a point belongs to the surface of a membrane and outputs the corresponding translucent properties. The multidimensional vector corresponding to each pixel is input into the neural radiation field model to obtain the pixel's parameter information; the parameter information includes color and density. Based on the pixel's parameter information, a three-dimensional point cloud model is generated using volumetric rendering technology; volumetric rendering technology is used to simulate the changes in light propagation in three-dimensional space.

[0064] In step S140, after obtaining the three-dimensional point cloud model, the three-dimensional point cloud model is subjected to format conversion, geometric repair, physical attribute assignment, virtual environment modeling, and multiphysics coupling to generate a digital twin model of the flexible membrane.

[0065] Format conversion transforms 3D point cloud models into formats recognizable by computational tools. Geometric repair intelligently repairs model defects generated during scanning or conversion. Physical property assignment injects materials science principles into the geometric model, achieving a leap from "shape" to "physical entity." Virtual environment modeling recreates the extreme working conditions of the diaphragm in digital space. Multiphysics coupling is the core engine of the digital twin, simultaneously solving the physical equations of interaction. When the diaphragm is heated by neutron radiation, the heat conduction equation interacts with thermal expansion strain in real time: temperature field changes induce structural deformation, which in turn affects the thermal boundary conditions; simultaneously, radiation embrittlement causes dynamic degradation of material properties, further influencing the mechanical response.

[0066] In step S150, the flexible diaphragm is measured using a digital twin model, that is, angle measurement is performed in virtual space. The geometric shape of the flexible diaphragm under different working conditions can be simulated using a digital twin model, and the angle of the flexible diaphragm is measured in virtual space to obtain measurement data. The measurement data includes the angle of the flexible diaphragm relative to the reference plane, as well as the relative angle changes between different parts of the flexible diaphragm.

[0067] In the aforementioned implementation process, a high-precision 3D point cloud model is generated based on a multi-angle camera using a neural radiation field model. This 3D point cloud model can reconstruct the surface morphology of the diaphragm, improving the data loss problem caused by the poor light transmittance and reflectivity of the diaphragm in traditional optical methods. Digital twin technology maps the state of the real physical device to a virtual model, supporting real-time monitoring of the flexible diaphragm's angle changes. The synergistic technology chain of the neural radiation field model and digital twin not only improves the accuracy of the neutron fine collimator's flexible diaphragm angle measurement but also greatly enhances its efficiency.

[0068] In an optional embodiment, after acquiring multi-angle images, image data defects caused by membrane transmittance (making the transmittance greater than or equal to 90%) and dense occlusion (making the spacing less than or equal to 0.1 mm) can be improved through standardization processing and algorithm optimization, thereby reducing the accuracy error of 3D reconstruction (e.g., making the accuracy error less than or equal to 0.05 mm). The implementation method is as follows:

[0069] 1. Image data processing

[0070] Data Classification: Each acquired image data needs to be classified to ensure that it is not confused during processing. Therefore, the image naming convention should be clear and unambiguous, usually classified and stored according to the shooting angle (0 degrees, 45 degrees, 90 degrees) and neutron energy level (e.g., 0.025eV thermal neutron, 1eV hyperthermal neutron). The file naming rule is "viewpoint_energy_number" (e.g., "45_0.025eV_001").

[0071] 2. Image format conversion and standardization

[0072] Standardize the format: Use image software such as ImageMagick to batch convert RAW format to 16-bit TIFF (lossless compression) while preserving dynamic range;

[0073] Resolution calibration: The resolution was uniformly adjusted to 300 dpi (dots per inch) using an FFmpeg script, and the super-resolution areas (such as the edge of the membrane) were enhanced to 0.025 mm / pixel using the ESRGAN algorithm (magnification × 2).

[0074] 3. Noise reduction and contrast enhancement

[0075] Noise suppression: For high-energy particle noise in the neutron radiation field (SNR less than or equal to 35dB), nonlocal mean filtering (search window 21×21, similar block 7×7) combined with wavelet threshold denoising (Symlet 8, threshold 0.05) is used to improve the signal-to-noise ratio to greater than or equal to 42dB.

[0076] Edge enhancement: Apply limited contrast adaptive histogram equalization (CLAHE, 8×8 grid, contrast limit 2.0) to enhance the contrast of the coating edges (gradient value increased from 15 to 45).

[0077] Transmittance compensation: Based on the transmittance area mask (brightness greater than 200 lux area), Poisson image editing is used to fill holes (repair rate greater than 98%).

[0078] 4. Color correction and brightness adjustment

[0079] Color chart calibration: Using a 24-color chart (embedded during the shooting stage), color cast is corrected and color difference is reduced through a multinomial regression model (3rd degree term);

[0080] Brightness normalization: Based on the reflectivity of the neutron-absorbing coating, adjust the image grayscale histogram to the mean to ensure consistent brightness across different viewing angles (e.g., variance less than 5).

[0081] HDR Composition: For overexposed (greater than 240) and underexposed (less than 20) areas, 16-bit HDR images are synthesized by exposure bracketing, with a dynamic range covering 0-65535.

[0082] 5. Image cropping and edge correction

[0083] During image acquisition, unwanted background or edge areas may exist, affecting image quality. To ensure that the image contains only the actual data of the flexible membrane, it needs to be cropped to remove unnecessary parts. The cropped image should contain only key information about the membrane surface for subsequent processing and analysis.

[0084] 6. Image registration and alignment

[0085] Before inputting images from multiple viewpoints, it is necessary to ensure that these images are spatially aligned, that is, to ensure that images from different viewpoints have a consistent scale and alignment. Image registration algorithms (such as feature matching and affine transformation) are used to ensure that the feature points of all images are accurately aligned so that the subsequent 3D reconstruction process can proceed smoothly.

[0086] Through the above steps, the acquired image data will have high-quality resolution, uniform illumination, clear details, and a consistent scale, further providing reliable basic data for subsequent 3D reconstruction processing.

[0087] Optionally, in this embodiment, pixel information includes the pixel coordinates of the pixel points; based on the pixel information of the multi-angle image and camera parameters, the multi-angle image is mapped to a three-dimensional coordinate system to obtain the multi-dimensional vector corresponding to the pixel points in the multi-angle image, including: acquiring the camera parameters for acquiring the multi-angle image, and the pixel coordinates of the pixel points in the multi-angle image; the camera parameters include camera intrinsic parameters and camera extrinsic parameters; wherein, the camera intrinsic parameters include the focal length ( , ) and principal point coordinates ( , These parameters can be obtained through the camera calibration process. Camera extrinsic parameters include the camera's spatial position (…). The parameters (R) describe the camera's position and orientation relative to the world coordinate system.

[0088] First, let's introduce how to calculate the direction of light ( The process of ).

[0089] 3D direction vector (d): For each pixel, we need to calculate the unit vector from the camera to the surface of the object (flexible membrane) (i.e., the direction of the light). This is because each pixel of the image represents a ray of light projected from the camera onto the surface of the object.

[0090] Based on the camera's intrinsic parameters, the three-dimensional direction vector can be calculated using the following formula:

[0091]

[0092] in,( , ) are the pixel coordinates in the image, ( , ) are the principal point coordinates of the camera, ( , ( ) is the camera's focal length. (The result is...) It is the three-dimensional direction vector of light. The parameter representing the light ray in the x-direction. The parameter representing the ray in the y-direction. This represents the parameter of the light ray in the z-direction.

[0093] Then, the pitch angle of the light ray is calculated based on the three-dimensional direction vector. and azimuth :

[0094] Based on the three-dimensional direction vector We can calculate the pitch angle of the light. and azimuth These two angles describe the direction of the light.

[0095] Pitch angle ( ( ) represents the angle of a ray of light relative to the vertical direction (in the xy plane), and its calculation formula is:

[0096]

[0097] azimuth angle ( () represents the angle of the ray relative to the horizontal plane (along the z-axis), and its calculation formula is:

[0098]

[0099] In the NeRF model, each ray contains not only direction information but also its spatial position. Since images are two-dimensional, while rays propagate in three-dimensional space, we need to map the coordinates of each pixel to its position in three-dimensional space. To obtain the spatial coordinates (… We need to perform transformations using the camera's extrinsic parameters (camera position and rotation matrix). The following describes how to calculate the spatial position (…). The process of ).

[0100] Given the camera's extrinsic parameters and its spatial position ( Given the camera coordinate system and the rotation matrix (R), the transformation relationship between the camera coordinate system and the world coordinate system is as follows:

[0101]

[0102] in: It is the three-dimensional direction vector of the light ray obtained from the previous calculation.

[0103] By giving the direction d of the ray in this way, and combining it with the camera's rotation and position, we can map the ray from the camera coordinate system to the world coordinate system, thus obtaining the ray's three-dimensional spatial position. ).

[0104] The final 5D vector is: 5D = ( Through the above steps, the processed and pre-processed image data is transformed into a 5D vector format that meets the requirements of NeRF technology. This allows the spatial location and viewpoint information in the images to be used as input to generate 3D reconstruction data for the flexible diaphragm. This process provides accurate input data for the NeRF model, enabling it to accurately reconstruct the 3D geometry of the flexible diaphragm based on 2D image information from multiple viewpoints.

[0105] In the implementation of the above embodiments: by converting the sorted and preprocessed multi-angle image data into multi-dimensional vectors that meet the requirements of the neural radiation field model, the spatial position and viewpoint information in the images are used as input to generate three-dimensional reconstruction data of the flexible membrane. Through this process, accurate input data is provided to the neural radiation field model, enabling it to accurately reconstruct the three-dimensional point cloud model of the flexible membrane based on two-dimensional image information from multiple viewpoints.

[0106] Optionally, in this embodiment, the neural radiation field model includes a vector input layer, a position encoding layer, a feature extraction layer, a ray projection and sampling layer, and an output layer;

[0107] The vector input layer is used to transform multidimensional vectors into feature representations. For example, a multidimensional vector ( ) will be input into the vector input layer of the Neural Radiation Field Model (NeRF algorithm model), where the spatial coordinates ( Accuracy ±0.05 mm, angle ( With an accuracy of ±0.01 degrees, the vector input layer processes spatial position and light direction, transforming them into feature representations that the neural network can understand.

[0108] The positional encoding layer is used to encode the feature representation to obtain the encoded spatial coordinates. Since neural networks have limitations in understanding continuous positions, NeRF in this embodiment uses positional encoding technology. Positional encoding is achieved by encoding the spatial coordinates (… The network is processed using sine and cosine functions, thereby increasing its ability to sense changes in location.

[0109] The encoded spatial coordinates become a high-dimensional spatial representation, allowing the network to better capture subtle differences between locations during the learning process. This process is crucial for capturing complex spatial information, especially for modeling details and changes in lighting.

[0110] For each spatial coordinate, the position encoding function yes:

[0111]

[0112]

[0113] In an optional embodiment, positional encoding enhancement can also be performed: the number of spatial coordinate encoding layers L = 10, generating a dimension of 3 × 2 × 10 = 60 dimensions; the number of ray direction encoding layers L = 4, generating a dimension of 2 × 2 × 4 = 16 dimensions. Transmittance region suppression is also performed: for 5D vectors with a brightness greater than 200 lux, transparency weights are embedded to reduce transmission noise interference (signal-to-noise ratio improved by 40%).

[0114] The feature extraction layer is used to extract features from the encoded spatial coordinates to generate feature data; the feature extraction layer includes multiple fully connected layers; the fully connected layers include an occlusion attention module, which is used to dynamically mask the features of the occluded area using the membrane spacing data.

[0115] The location-encoded data is then fed into the model's feature extraction layer for further processing. This layer uses an 8-layer fully connected network (FCN) instead of a traditional convolutional neural network (CNN). It performs non-linear mapping and feature extraction on the location-encoded data to generate high-level features related to the object's surface details.

[0116] Anti-occlusion design: Occlusion-AwareAttention modules are inserted in the third and fifth layers of the feature extraction layer to dynamically mask features in occluded areas using the inter-membrane spacing data (≤0.1 mm). Feature fusion formula:

[0117]

[0118] in, The characteristics after fusion Original features The distance between adjacent membranes, This is a skip connection feature.

[0119] The ray projection and sampling layer is used to sample the ray path corresponding to the pixel multiple times to generate multiple sampling points.

[0120] For each ray (one ray per pixel), the neural network samples it multiple times along its path. Specifically, as the light rays travel from the camera's viewpoint into the scene, the network predicts the color and density of each 3D point it passes through. The collection of these points forms a ray propagation model of the 3D scene. In this way, NeRF technology can simulate the propagation of light and the illumination of object surfaces in the real world.

[0121] In one alternative embodiment, the ray sampling optimization strategy can be hierarchical sampling: uniformly sampling 64 points along each ray (step size Δt = 0.5 mm); it can also be importance sampling: adding 64 points in the region where the density gradient ▽σ > 0.1, for a total of 128 sampling points / ray; or transmission compensation sampling: using sparse sampling (step size Δt = 2 mm) for the transmission region (α < 0.3) to reduce invalid calculations (50% reduction in sampling points).

[0122] The output layer outputs the parameter information of the sampling points; this parameter information is used to generate a 3D point cloud model. For the output layer, the neural network fuses the illumination, density, and color of each sampling point through multiple fully connected layers, and finally generates the final result for each sampling point through the output layer. The output includes: color (c = (r, g, b)): representing the color of light passing through the object's surface, and density (σ): representing the transparency or light absorption of that point. This output data provides all the necessary parameter information for subsequent 3D rendering.

[0123] In the implementation of the above embodiments: the neural radiation field model adopts a layered architecture design to achieve accurate digital reconstruction of the complex light propagation process in the real world. The vector input layer is compatible with the three-dimensional coordinates and light direction parameters of physical space, enabling the model to naturally connect with the raw data collected by various sensors. The position encoding layer uses high-frequency signal conversion technology to transform smooth spatial position and direction information into detailed mathematical expressions, effectively improving the model's ability to perceive minute structural features, such as micron-level wrinkles on the membrane surface or the gradient effect of the light-transmitting coating. The feature extraction layer, as the core of the deep neural network, extracts the abstract laws of light-matter interaction from the encoded data through multi-layer nonlinear transformations, including implicit physical field information such as material density distribution and scattering characteristics. The light projection and sampling layer maps the abstract features of the preceding steps back to a physically interpretable dimension. This layer intelligently generates density-adaptive sampling points along each light path, automatically densifying the point cloud in the membrane entity area and sparsely sampling in open areas, significantly improving computational efficiency. The final output layer simultaneously generates dual-channel results of color and density, satisfying both visual rendering requirements and providing geometric existence criteria for subsequent point cloud reconstruction.

[0124] The training strategy for the neural radiation field model is described below. The hardware platform, optimizer, batch size, and training cycle parameters in the training parameter configuration can be set according to actual needs. For example, the hardware platform can be an NVIDIA A100 GPU (80GB VRAM), distributed training (4 nodes); the optimizer can be Adam (learning rate 5×10⁻⁴, exponential decay to 1×10⁻⁵); the batch size can be 128 sampling points per ray, with 4096 rays per batch; and the training cycle can be 200,000 iterations (approximately 48 hours).

[0125] Loss function: During training, the difference between the 3D scene generated by the model and the actual image is calculated using a loss function. A commonly used loss function is mean squared error (MSE), which aims to minimize the error between the network-generated image and the real image.

[0126] The loss function formula is:

[0127]

[0128] in, These are the pixel values ​​generated by the model. is the pixel value of the actual image, N is the total number of pixels in the image, and i represents the pixel.

[0129] Dynamic training strategies can be employed, such as progressive training: the first 50,000 iterations use low-resolution images (1368×912) with a coding layer L=6; the next 150,000 iterations use full-resolution images (5472×3648) with a coding layer L=10. Alternatively, occlusion-aware training can be used: the membrane spacing mask is updated every 10,000 iterations; the weights of sampling points in occluded areas are reduced to 0.1 × the normal value.

[0130] During training, anti-interference enhancement techniques can be incorporated, such as neutron radiation field adaptation: 20% high-noise images (SNR=30dB) are mixed into the training data to improve the model's resistance to radiation interference; the PSNR of the noisy images is consistently greater than or equal to 28 dB. Thermal deformation compensation can also be used: temperature deformation data (ΔT=5 degrees Celsius expansion) is injected every 60 minutes to dynamically update the extrinsic parameter matrices R and T; the reconstruction error fluctuation after deformation is less than 0.005 mm.

[0131] Optionally, in this embodiment of the application, a three-dimensional point cloud model is generated using multidimensional vectors and a trained neural radiation field model, including:

[0132] The multidimensional vectors corresponding to each pixel are input into the neural radiation field model to obtain the pixel's parameter information, including color and density. Based on the pixel's parameter information, a 3D point cloud model is generated using volumetric rendering technology, which is used to simulate the changes in light as it propagates in 3D space.

[0133] After inputting the multidimensional vectors corresponding to each pixel into the neural radiation field model, the model gradually analyzes the physical behavior of light in three-dimensional space through a hierarchical information processing mechanism. The multidimensional vectors first carry the two-dimensional position of the pixel and its corresponding three-dimensional light direction. After high-frequency signal enhancement by the position encoding layer, subtle spatial features, such as surface ripples on a film or gradual changes in coating thickness, are transformed into recognizable mathematical patterns. The feature extraction layer analyzes the interaction between light and matter through deep neural network weights. Finally, the output layer generates the physical parameters for each spatial sampling point: the color parameter records the visible spectrum reflectance characteristics of light at that point, visually presenting the material's appearance; the density parameter quantifies the degree of material aggregation at that location, determining the attenuation intensity of light penetration.

[0134] Volume rendering technology combines the color and density of each sample point to generate the color of each ray of light. Volume rendering is used to simulate the changes in light as it propagates in three-dimensional space, ultimately producing an image of the three-dimensional scene.

[0135] Rendering equations, for example:

[0136]

[0137] in: It is the final color of the light. It represents the density at each sampling point, indicating the degree to which light is absorbed. The color (RGB value) of the sampled point is predicted by the network. Transmittance is the percentage of light that is not absorbed or scattered when it passes through a certain area. This represents a point in time. Among them, transmittance... It can be calculated from the density of each sampling point, and is defined as:

[0138]

[0139] In each calculation, the color of the light is gradually accumulated. Assume we start from a starting point and sample multiple points along the direction of the light. Each sample point has a contribution value, determined by its color and density values. The accumulation process can be achieved through the following steps:

[0140] The light rays originating from the camera will pass through multiple sampling points, and the color and density of each point will affect the final color.

[0141] The color contribution of each sampling point is weighted, and points with higher transmittance (higher density) have a greater impact on the final result.

[0142] Through step-by-step calculations, the composite color of all the light rays along the light direction from the camera's viewpoint to the object's surface is finally obtained.

[0143] The following describes the specific implementation of the volumetric rendering formula: At each sampling point, the final ray's color (RGB) and density (σ) can be calculated using weighted averages. For example, NeRF uses the following cumulative calculation formula to process the rendering of each ray:

[0144]

[0145] in, It is the transmittance at the current sampling point. It is the density at that point. It is the color of that point. It is the transmittance of the next sampling point corresponding to the current sampling point, where i represents the sampling pixel.

[0146] This process gradually accumulates the contribution of each sampling point, eventually synthesizing the color of the light.

[0147] The process then involves post-processing to generate a 3D model (industrial-grade precision). The first step is point cloud generation: Depth map extraction: Based on volumetric rendering results, a 16-bit depth map is generated (resolution 5472×3648, precision 0.01 mm); Point cloud construction: The 3D coordinates of each pixel are calculated through back projection, generating approximately 20 million points per patch.

[0148] Next, mesh reconstruction and optimization were performed, including Poisson reconstruction: setting the octree depth, smoothing iterations 5 times, generating a closed mesh (approximately 5 million faces); mesh optimization: Laplacian smoothing: iterations 3 times, eliminating jagged edges (curvature radius error less than 0.02 mm); edge refinement: local subdivision of areas with curvature radius greater than or equal to 0.1 mm (subdivision level 3); hole repair: based on a light-compensating mask (repair rate greater than 95%), filling missing faces.

[0149] After mesh reconstruction and optimization, model verification and correction can be performed, such as accuracy benchmarking: the laser scanner comparison error is less than or equal to 0.05 mm; automated defect detection: model tortuosity is identified through a CNN classification network (ResNet-50) (detection rate greater than 99%); thermal deformation compensation: temperature deformation coefficient is injected to dynamically adjust the mesh vertex coordinates.

[0150] The model can then be experimentally validated, including aspects such as accuracy indicators: geometric error of the 3D model less than or equal to 0.05 mm (ISO 10360 standard); angle measurement error ±0.01 degrees (compared to laser interferometer); and / or efficiency indicators: single-diaphragm reconstruction time less than or equal to 8 minutes (including post-processing); point cloud to mesh conversion less than or equal to 2 minutes; and / or anti-interference capability: model integrity greater than 98% under neutron radiation field; artifact rate in the transparent area less than 1%.

[0151] In the implementation of the above embodiments: volumetric rendering technology simulates the real light propagation process based on the parameter information of pixels. Multiple spatial points are sampled along each light path, and the transmittance and absorptivity of light passing through each point are calculated based on density parameters, while simultaneously accumulating color contribution values. Through integration, the discrete sampling points are restored to a continuous spatial material distribution, ultimately constructing a 3D point cloud model that combines geometric accuracy and optical realism. This approach is more suitable for flexible, semi-transparent objects such as membranes, accurately simulating the light scattering path and energy deposition distribution under neutron radiation, providing an atomically accurate spatial data base for subsequent digital twins, thereby improving the accuracy of flexible membrane measurements.

[0152] Optionally, in this embodiment of the application, generating a digital twin model of the flexible membrane based on the three-dimensional point cloud model includes:

[0153] The geometric restoration of the 3D point cloud model after format conversion is performed to obtain the restored model; the geometric restoration includes light transmission compensation and curvature optimization.

[0154] The 3D mesh model (5 million faces) generated by NeRF was converted to STEP format and compressed to Parasolid format (the number of faces was reduced to 2 million, and the accuracy was retained to ±0.01 mm). The converted data was then verified: the point cloud and the mesh model were compared by the ICP algorithm (iterative nearest point), and the registration error was less than or equal to 0.02 mm.

[0155] Transmittance compensation: For holes in the transmittance area (e.g., brightness greater than 200 lux), a non-manifold topology repair algorithm is used to fill the missing surface (e.g., repair rate greater than 95%).

[0156] Curvature optimization: For bending regions with a curvature radius greater than or equal to 0.1 mm, the Catmull-Clark subdivision algorithm (subdivision level 3) is applied, resulting in smaller curvature continuity errors.

[0157] The repaired model is then assigned physical properties to obtain a physical entity model; the assignment of physical properties includes material modeling and dynamic property configuration.

[0158] Material modeling includes elastic modulus and neutron-absorbing coating. Elastic modulus: The anisotropic parameters of the flexible membrane substrate (polyimide) are determined according to requirements and calibrated through nanoindentation experiments. Neutron-absorbing coating is used to set the coating thickness, density, and coefficient of thermal expansion.

[0159] Dynamic property configuration includes temperature-dependent parameters and stress relaxation models: Temperature-dependent parameters: The elastic modulus is related to temperature through the Arrhenius equation; while the stress relaxation model adopts the generalized Maxwell model (three relaxation times can be set).

[0160] A virtual environment model is created by modeling the physical entity to obtain the object to be simulated. The virtual environment modeling includes neutron radiation field simulation and embedding radiative thermal effects. The virtual environment modeling includes constructing a neutron flux distribution field based on MCNP (Monte Carlo particle transport code), with an exemplary mesh accuracy of 0.1 mm and a flux gradient error of less than or equal to 5%. The virtual environment modeling also includes embedding radiative thermal effects to calculate the neutron energy deposition power.

[0161] Multiphysics coupling is performed based on the parameters of the object to be simulated to generate a digital twin model of the flexible diaphragm. Multiphysics coupling includes mechanical-thermal coupling and boundary conditions. Mechanical-thermal coupling refers to the synchronous calculation of thermal expansion and stress distribution through a two-way coupled solver. Boundary conditions include fixed boundaries, such as fixed edges of the diaphragm, and dynamic loads, such as neutron impact force and loading frequency.

[0162] It is understood that the specific parameters for generating the digital twin model can be set according to actual needs, and this application embodiment does not limit this.

[0163] In the implementation of the above embodiments: format conversion solves data compatibility issues, geometric repair strip model is computable, physical property assignment imparts material realism, virtual environment modeling loads external stimuli, and multiphysics coupling integrates all elements for behavior inference. For example, after loading a neutron radiation field, the repaired geometric model can predict edge warping of 0.1 mm through thermo-mechanical coupling simulation, thereby guiding the implementation of multi-density monitoring of this area in actual measurements. This transforms the traditional R&D model that relies on physical trial and error (such as a single neutron irradiation experiment costing over 500,000 yuan) into an efficient iteration in digital space, shortening the R&D cycle and reducing trial and error costs. It also allows for the prediction of diaphragm failure risks under extreme conditions through coupled simulation.

[0164] After creating the digital twin model, multiphysics coupling simulation technology was used to predict and optimize the behavior of the flexible diaphragm under complex environments such as neutron radiation and temperature changes. The simulation process is described below:

[0165] Optionally, in this embodiment of the application, after generating a digital twin model of the flexible membrane based on the three-dimensional point cloud model, the method further includes:

[0166] The digital twin model is subjected to static performance simulation, dynamic performance simulation, and multiphysics coupling simulation in sequence to generate a measurement strategy. Static performance simulation is used to simulate the deformation and stress distribution of the diaphragm under working conditions. Static performance simulation includes deformation analysis and stress concentration location analysis under neutron radiation. Dynamic performance simulation is used to analyze the vibration, fatigue, and other behaviors of the diaphragm under dynamic loads. Dynamic performance simulation includes vibration mode analysis and fatigue life prediction. Multiphysics coupling simulation is used to comprehensively analyze the interaction between thermal, mechanical, and radiation fields. Multiphysics coupling simulation includes thermo-mechanical coupling analysis and radiation material performance degradation.

[0167] The goal of static performance simulation is to predict the deformation and stress distribution of the diaphragm under fixed operating conditions.

[0168] Static performance simulation includes deformation analysis under neutron radiation: for example, simulating a neutron beam (flux 1×10⁻⁶). 6 ~ 8When irradiated at a rate of (per square centimeter per second), the membrane experiences thermal expansion due to the deposition of neutron energy. For example, if the coating absorbs neutrons and its temperature rises, causing the edge area of ​​the membrane to warp up by 0.1 mm, the system will mark high-risk areas with deformation exceeding 0.05 mm.

[0169] Static performance simulation also includes stress concentration location: for example, identifying stress peaks (e.g., 85 MPa) at diaphragm bends (e.g., creases with a radius of curvature of 0.1 mm), which are close to the material yield limit (90 MPa), indicating that structural design needs to be optimized.

[0170] The goal of dynamic response simulation is to analyze the vibration, fatigue, and other behaviors of a diaphragm under dynamic loads.

[0171] Dynamic response simulation includes vibration mode analysis: for example, simulating the vibration mode of a diaphragm under a 10Hz neutron impact force, it was found that the first natural frequency is 120Hz (which is prone to resonance with external loads). By increasing the stiffness of the supporting structure, the frequency is raised to 150Hz to avoid the risk of resonance.

[0172] Static performance simulation also includes fatigue life prediction: for example, based on 100 million cycles of load (simulating 10 years of operation), it is predicted that the interface between the coating and the substrate is the fatigue crack initiation zone, with a life of only 5 years; by thickening the coating to 60 micrometers, the life is extended to 8 years.

[0173] The goal of physical field coupling simulation is to integrate the mutual influence of thermal, mechanical, and radiation fields.

[0174] The physical field coupling simulation includes thermo-mechanical coupling analysis: when neutron radiation causes a local temperature rise of 50 degrees Celsius, the diaphragm deforms by 0.03 mm due to thermal expansion. At the same time, the high temperature reduces the elastic modulus of polyimide by 5%, further aggravating the deformation. The system corrects the deformation in reverse through a pre-compensation algorithm.

[0175] The physical field coupling simulation also includes radiation-material property degradation: under long-term neutron irradiation, the coating embrittlement leads to a 10% decrease in elastic modulus and an increase in deformation error by 0.02 mm, triggering an automatic coating thickening maintenance prompt.

[0176] Through the above simulation process, the digital twin can accurately predict the performance limits of the diaphragm under extreme environments and guide design optimization, ultimately achieving industrial-grade reliability with an error of less than or equal to 0.05 mm, providing key decision support for practical applications.

[0177] Measurement of flexible diaphragms using digital twin models includes: measuring the flexible diaphragm using a digital twin model according to a measurement strategy. For example, if simulation predicts that the stress concentration area is at the bending point, the measurement strategy could be to add strain gauges to the R0.1 mm area and perform measurements according to the strategy, which can improve the capture rate of key data. Another example is if simulation predicts that thermal expansion deformation follows a gradient distribution; the measurement strategy could be to use full-field laser scanning instead of single-point thickness measurement, thereby improving the accuracy of deformation measurement.

[0178] In the implementation process of the above embodiments: static performance simulation first provides a baseline snapshot in the spatial dimension to accurately locate high stress areas and deformation over-limit points under fixed working conditions; dynamic performance simulation then introduces the time dimension to capture transient response laws through vibration mode and fatigue life analysis; multi-physics coupling simulation finally integrates the interaction of heat, force, radiation and other factors to accurately predict the performance limits of the diaphragm in extreme environments, providing data support and key decision support for subsequent measurements.

[0179] The simulation process involves mapping the complexities of the physical world onto a virtual space for pre-simulation and decision-making, ultimately outputting quantitative results that can guide engineering practice—that is, a measurement strategy. This can improve the accuracy, reduce costs, and mitigate risks in subsequent actual measurements.

[0180] Optionally, in this embodiment of the application, measuring the flexible diaphragm using a digital twin model includes:

[0181] The geometry of the flexible diaphragm under different working conditions is simulated by a digital twin model. The angle of the flexible diaphragm is measured in virtual space to obtain measurement data. The measurement data includes the angle of the flexible diaphragm relative to the reference plane, as well as the relative angle changes between different parts of the flexible diaphragm.

[0182] After measuring the flexible diaphragm using a digital twin model, the method further includes: obtaining measurement data of the flexible diaphragm, analyzing the measurement data, generating a diaphragm angle correction scheme, and correcting the measurement data using the diaphragm angle correction scheme.

[0183] The objective of this embodiment is to perform precise measurements of the angles of a flexible diaphragm using a digital twin, and to compare the measurement results in virtual space with real-world angle standards to guide the actual angle adjustment of the diaphragm. This allows the angle data measured by the digital twin to provide a valid basis for diaphragm angle correction in the real world, thereby improving the accuracy and stability of the diaphragm during operation.

[0184] An angle measurement of a flexible diaphragm is performed in virtual space using a digital twin model. The digital twin simulates the diaphragm's geometry under different operating conditions to accurately calculate its angle data. This process considers not only the diaphragm's initial angle but also angle changes caused by external environmental variations, temperature fluctuations, and other factors during use.

[0185] Virtual angle measurement: By accurately calculating the angle changes of the diaphragm in virtual space, angle data of the diaphragm under different operating conditions can be obtained. Specifically, this includes the angle of the diaphragm relative to a reference plane, as well as the relative angle changes between different parts of the diaphragm.

[0186] Angle deviation calculation: Measure the angle in the virtual model and compare it with the preset standard angle to calculate the angle deviation and obtain error data.

[0187] The virtual measurement results are then compared with the actual diaphragm angle data to perform deviation analysis. This process helps the system identify and quantify the angle errors that occur during the actual use of the diaphragm.

[0188] Data Comparison: The measurement results in the digital twin are compared with the angle data obtained from sensors or other measuring tools in the actual device. Based on the comparison results, the system can calculate the difference between the actual diaphragm and the standard angle.

[0189] Error source analysis: The root cause of the error may be the angle deviation caused by external environmental factors (such as temperature changes), non-ideal properties of the diaphragm material, external forces, etc.

[0190] Based on the comparative analysis results, the system will generate an error correction plan and provide guidance for angle correction in actual equipment. This correction plan can be implemented through design optimization or operational adjustments.

[0191] Correction scheme generation: By analyzing the measurement results, a diaphragm angle correction scheme is generated, including design adjustments and operational optimization schemes.

[0192] Correction Execution: The actual equipment operates and adjusts according to the correction scheme provided by the digital twin to ensure that the diaphragm is restored to the standard angle.

[0193] In the implementation of the above embodiments: by comparing and analyzing the angle measurement of the digital twin with the actual data, the angle deviation of the flexible diaphragm can be identified and corrected, so that the diaphragm can always maintain the predetermined angle accuracy during operation.

[0194] Optionally, in this embodiment of the application, the measurement of the flexible diaphragm using a digital twin model includes: real-time monitoring of the flexible diaphragm and uploading the detection results to the control system; if an angle shift of the flexible diaphragm is detected, the position or external force of the flexible diaphragm is automatically adjusted by the control system.

[0195] The control system transmits the virtual correction results directly to the actual equipment through a real-time feedback mechanism, automatically adjusting the diaphragm angle. This process ensures that the diaphragm can react quickly and correct deviations during actual operation.

[0196] Real-time feedback mechanism: The control system automatically transmits the feedback results to the actual equipment by monitoring the angle change of the diaphragm in real time through sensors, so as to ensure that the angle of the diaphragm is maintained within the predetermined standard range.

[0197] Automatic correction operation: The system automatically adjusts the operating parameters or performs correction operations based on feedback data to restore the actual diaphragm to the standard angle and avoid errors caused by human intervention.

[0198] Through real-time monitoring, the control system continuously acquires diaphragm angle data and automatically adjusts it according to different working conditions. This dynamic adjustment ensures the diaphragm maintains an ideal angle at all times, correcting for changes in the working environment.

[0199] Real-time angle monitoring: The system continuously tracks the diaphragm angle and compares it with the standard angle. Any deviation will be detected immediately and fed back to the control system.

[0200] Dynamic adjustment: When the system detects a shift in the diaphragm angle, it will immediately adjust the position of the diaphragm or the application of external force through closed-loop control to ensure that it returns to the preset ideal angle.

[0201] In the implementation of the above embodiments: the angle correction scheme generated in the digital twin is fed back to the actual equipment in real time through the control system, enabling the flexible diaphragm to maintain an ideal angle during actual operation and to make dynamic corrections based on real-time data. This process improves the accuracy and efficiency of angle correction, enabling the equipment to operate stably for a long time.

[0202] This application provides a neutron collimator flexible diaphragm measurement device, including:

[0203] The image acquisition module is used to acquire multi-angle images of the flexible diaphragm of the neutron collimator;

[0204] The multi-dimensional vector transformation module is used to map multi-angle images to a three-dimensional spatial coordinate system based on pixel information and camera parameters, and obtain multi-dimensional vectors corresponding to pixels in the multi-angle images; the multi-dimensional vectors are used to represent the spatial position and light direction of the pixels;

[0205] The point cloud module is used to generate a three-dimensional point cloud model using multi-dimensional vectors and a trained neural radiation field model; the three-dimensional point cloud model is used to represent the sampling point information corresponding to light rays;

[0206] The digital twin module is used to generate a digital twin model of a flexible membrane based on a 3D point cloud model; the digital twin model is an information model based on the virtual mapping of the flexible membrane.

[0207] The measurement module is used to measure the flexible diaphragm using a digital twin model.

[0208] Optionally, in this embodiment of the application, the neutron collimator flexible diaphragm measuring device includes pixel information including pixel coordinates of pixel points; a multi-dimensional vector conversion module is used to acquire camera parameters for acquiring multi-angle images, and pixel coordinates of pixel points in the multi-angle images; the camera parameters include camera intrinsic parameters and camera extrinsic parameters; the three-dimensional direction vector of light is calculated using the camera intrinsic parameters and pixel coordinates; the light direction of pixel points is determined based on the three-dimensional direction vector of light; and the spatial position corresponding to pixel points is determined based on the three-dimensional direction vector of light and camera extrinsic parameters.

[0209] Optionally, in this embodiment of the application, the neutron collimator flexible diaphragm measurement device and the point cloud module are used to input the multi-dimensional vector corresponding to the pixel into the neural radiation field model to obtain the parameter information of the pixel; the parameter information includes color and density; based on the parameter information of the pixel, a three-dimensional point cloud model is generated by volume rendering technology; the volume rendering technology is used to simulate the changes of light propagating in three-dimensional space.

[0210] Optionally, in this embodiment, the neutron collimator flexible diaphragm measurement device includes a neural radiation field model comprising a vector input layer, a position encoding layer, a feature extraction layer, a light projection and sampling layer, and an output layer. The vector input layer is used to convert multidimensional vectors into feature representations. The position encoding layer is used to encode the feature representations to obtain encoded spatial coordinates. The feature extraction layer is used to extract features from the encoded spatial coordinates to generate feature data. The feature extraction layer includes multiple fully connected layers. The fully connected layers include an occlusion attention module, which dynamically masks the features of the occluded area using diaphragm spacing data. The light projection and sampling layer is used to sample the light path corresponding to each pixel multiple times to generate multiple sampling points. The output layer outputs the parameter information of the sampling points. The parameter information of the sampling points is used to generate a three-dimensional point cloud model.

[0211] Optionally, in this embodiment, the neutron collimator flexible diaphragm measurement device and digital twin module are used to perform geometric repair on the format-converted three-dimensional point cloud model to obtain a repaired model; the geometric repair includes light transmission compensation and curvature optimization; the repaired model is assigned physical properties to obtain a physical entity model; the physical property assignment includes material modeling and dynamic property configuration; the physical entity model is modeled in a virtual environment to obtain the object to be simulated; the virtual environment modeling includes neutron radiation field simulation and embedded radiation thermal effects; multiphysics coupling is performed according to the parameters of the object to be simulated to generate a digital twin model of the flexible diaphragm; the multiphysics coupling includes mechanical-thermal coupling and boundary conditions.

[0212] Optionally, in this embodiment, the neutron collimator flexible diaphragm measurement device further includes a simulation module for sequentially performing static performance simulation, dynamic performance simulation, and multiphysics coupling simulation on the digital twin model to generate a measurement strategy. The static performance simulation is used to simulate the deformation and stress distribution of the diaphragm under operating conditions. The static performance simulation includes deformation analysis and stress concentration location analysis under neutron radiation. The dynamic performance simulation is used to analyze the vibration, fatigue, and other behaviors of the diaphragm under dynamic loads. The dynamic performance simulation includes vibration mode analysis and fatigue life prediction. The multiphysics coupling simulation is used to comprehensively analyze the interaction between heat, force, and radiation fields. The multiphysics coupling simulation includes thermo-mechanical coupling analysis and radiation material performance degradation. Measuring the flexible diaphragm through the digital twin model includes: measuring the flexible diaphragm through the digital twin model according to the measurement strategy.

[0213] Optionally, in this embodiment of the application, the neutron collimator flexible diaphragm measurement device includes a measurement module, which is used to simulate the geometry of the flexible diaphragm under different operating conditions using a digital twin model, and to perform angle measurements of the flexible diaphragm in virtual space to obtain measurement data. The measurement data includes the angle of the flexible diaphragm relative to a reference plane, as well as the relative angle changes between different parts of the flexible diaphragm. After measuring the flexible diaphragm using the digital twin model, the method further includes: obtaining the measurement data of the flexible diaphragm, analyzing the measurement data, generating a diaphragm angle correction scheme, and correcting the measurement data using the diaphragm angle correction scheme.

[0214] Optionally, in this embodiment of the application, the neutron collimator flexible diaphragm measuring device and the control and adjustment module are used to measure the flexible diaphragm through a digital twin model, including: real-time monitoring of the flexible diaphragm and uploading the detection results to the control system; if the angle of the flexible diaphragm is detected to be deviated, the position or external force of the flexible diaphragm is automatically adjusted by the control system.

[0215] It should be understood that this device corresponds to the above-described embodiment of the neutron collimator flexible diaphragm measurement method and is capable of performing the various steps involved in the above method embodiment. The specific functions of this device can be found in the description above, and detailed descriptions are omitted here to avoid repetition. The device includes at least one software functional module that can be stored in memory or embedded in the device's operating system (OS) in the form of software or firmware.

[0216] Please see Figure 3 The diagram shows a structural schematic of an electronic device provided in an embodiment of this application. An electronic device 300 provided in this application includes a processor 310 and a memory 320. The memory 320 stores machine-readable instructions executable by the processor 310. When the machine-readable instructions are executed by the processor 310, the method described above is performed.

[0217] Figure 3 The components shown can be implemented using hardware, software, or a combination thereof. Electronic device 300 may be a physical device, such as a server or PC, or a virtual device, such as a virtual machine or virtualization container. Furthermore, electronic device 300 is not limited to a single device; it can be a combination of multiple devices or a cluster of numerous devices.

[0218] This application also provides a storage medium storing a computer program, which is executed by a processor to perform the above-described method.

[0219] The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0220] This application also provides a computer program product, including computer program instructions, which are executed by a processor to perform the method described above.

[0221] It should be understood that the disclosed apparatus and methods can also be implemented in other ways, given the several embodiments provided in this application. The apparatus embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, or they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0222] In addition, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0223] The above description is only an optional implementation of the embodiments of this application, but the protection scope of the embodiments of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the embodiments of this application should be covered within the protection scope of the embodiments of this application.

Claims

1. A method for measuring a flexible diaphragm in a neutron collimator, characterized in that, include: Acquire multi-angle images of the flexible diaphragm of the neutron collimator; Based on the pixel information and camera parameters of the multi-angle image, the multi-angle image is mapped to a three-dimensional spatial coordinate system to obtain the multi-dimensional vector corresponding to the pixel in the multi-angle image; The multidimensional vector is used to characterize the spatial position and light direction corresponding to the pixel; Using the multidimensional vector and the trained neural radiation field model, a three-dimensional point cloud model is generated; the three-dimensional point cloud model is used to represent the sampling point information corresponding to the light rays. Based on the three-dimensional point cloud model, a digital twin model of the flexible membrane is generated; the digital twin model is an information model based on the virtual mapping of the flexible membrane. The flexible diaphragm was measured using the digital twin model; The step of generating a digital twin model of the flexible membrane based on the three-dimensional point cloud model includes: The geometric restoration of the 3D point cloud model after format conversion is performed to obtain the restored model; the geometric restoration includes light transmission compensation and curvature optimization. The repaired model is then assigned physical properties to obtain a physical entity model; the assignment of physical properties includes material modeling and dynamic property configuration. The physical entity model is used to perform virtual environment modeling to obtain the object to be simulated; the virtual environment modeling includes neutron radiation field simulation and embedded radiative thermal effects. A digital twin model of the flexible diaphragm is generated by performing multiphysics coupling based on the parameters of the object to be simulated; the multiphysics coupling includes mechanical-thermal coupling and boundary conditions.

2. The method according to claim 1, characterized in that, The pixel information includes the pixel coordinates of the pixels; based on the pixel information of the multi-angle image and camera parameters, the multi-angle image is mapped to a three-dimensional coordinate system to obtain the multi-dimensional vectors corresponding to the pixels in the multi-angle image, including: The camera parameters for acquiring the multi-angle images and the pixel coordinates of the pixels in the multi-angle images are obtained; the camera parameters include camera intrinsic parameters and camera extrinsic parameters. The three-dimensional direction vector of the light is calculated using the camera intrinsic parameters and the pixel coordinates; The direction of the light rays at the pixel is determined based on the three-dimensional direction vector of the light rays; The spatial position of the pixel is determined based on the three-dimensional direction vector of the light and the camera extrinsic parameters. Based on the light direction and spatial position of the pixel, the multidimensional vector corresponding to the pixel is obtained.

3. The method according to claim 1, characterized in that, The process of generating a three-dimensional point cloud model using the multidimensional vectors and the trained neural radiation field model includes: The multidimensional vector corresponding to the pixel is input into the neural radiation field model to obtain the parameter information of the pixel; the parameter information includes color and density. Based on the parameter information of the pixels, the three-dimensional point cloud model is generated using volumetric rendering technology; the volumetric rendering technology is used to simulate the changes in light as it propagates in three-dimensional space.

4. The method according to claim 1, characterized in that, in, The neural radiation field model includes a vector input layer, a position encoding layer, a feature extraction layer, a ray projection and sampling layer, and an output layer. The vector input layer is used to transform the multidimensional vector into a feature representation; The location encoding layer is used to perform location encoding on the feature representation to obtain the encoded spatial coordinates; The feature extraction layer is used to extract features from the encoded spatial coordinates to generate feature data; wherein, the feature extraction layer includes multiple fully connected layers; the fully connected layers include an occlusion attention module, which is used to dynamically shield the features of the occluded area through the membrane spacing data; The light projection and sampling layer is used to sample the light path corresponding to the pixel multiple times to generate multiple sampling points; The output layer is used to output the parameter information of the sampling points; the parameter information of the sampling points is used to generate the three-dimensional point cloud model.

5. The method according to claim 1, characterized in that, After generating a digital twin model of the flexible membrane based on the three-dimensional point cloud model, the method further includes: The digital twin model is subjected to static performance simulation, dynamic performance simulation, and multiphysics coupling simulation in sequence to generate a measurement strategy. The static performance simulation is used to simulate the deformation and stress distribution of the diaphragm under operating conditions. The static performance simulation includes deformation analysis and stress concentration location analysis under neutron radiation. The dynamic performance simulation is used to analyze the vibration and fatigue behavior of the diaphragm under dynamic loads. The dynamic performance simulation includes vibration mode analysis and fatigue life prediction. The multiphysics coupling simulation is used to comprehensively analyze the interaction between thermal, mechanical, and radiation fields. The multiphysics coupling simulation includes thermo-mechanical coupling analysis and radiation material performance degradation. The measurement of the flexible membrane using the digital twin model includes: measuring the flexible membrane using the digital twin model according to the measurement strategy.

6. The method according to claim 1, characterized in that, The measurement of the flexible diaphragm using the digital twin model includes: The digital twin model is used to simulate the geometry of the flexible diaphragm under different working conditions, and the angle of the flexible diaphragm is measured in virtual space to obtain measurement data. The measurement data includes the angle of the flexible diaphragm relative to the reference plane, as well as the relative angle changes between different parts of the flexible diaphragm. After measuring the flexible diaphragm using the digital twin model, the method further includes: Obtain measurement data of the flexible diaphragm, analyze the measurement data, generate a diaphragm angle correction scheme, and use the diaphragm angle correction scheme to correct the measurement data.

7. The method according to claim 6, characterized in that, The measurement of the flexible diaphragm using the digital twin model includes: The flexible diaphragm is monitored in real time, and the detection results are uploaded to the control system. If the angle of the flexible diaphragm is detected to be off, the position of the flexible diaphragm or the application of external force will be automatically adjusted by the control system.

8. A computer program product, characterized in that, It includes computer program instructions that, when executed by a processor, perform the method as described in any one of claims 1 to 7.

9. An electronic device, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, perform the method as described in any one of claims 1 to 7.

Images