NNBI-oriented neutral beam tomography diagnostic system and prior fusion reconstruction algorithm

Abstract

The invention relates to the technical field of imaging diagnosis of a controlled nuclear fusion negative ion source, in particular to a neutral beam tomography diagnosis system facing NNBI and a prior fusion reconstruction algorithm. Aiming at the problems that the existing tomography diagnosis scheme cannot realize the collaborative optimization of structural design, time sequence control and reconstruction algorithm, the invention provides the following technical scheme: the system is used for realizing the three-dimensional tomography and state quantitative diagnosis of the NNBI neutral beam; comprising a multi-view imaging unit, a hardware synchronous acquisition unit and a data processing and reconstruction unit, the multi-view imaging units are arranged at preset positions of an NNBI negative ion source beam source flange at equal angles and collect neutral beam multi-view projection images. Through collaborative optimization of the structure, the time sequence and the algorithm, the problems of time sequence synchronization, interference resistance, calibration and reconstruction robustness in NNBI neutral beam chromatography diagnosis are effectively solved, the diagnosis precision and stability are improved, and the engineering implementability is high.

Description

Technical Field

[0001] This invention relates to the field of imaging diagnostic technology for controlled nuclear fusion negative ion sources, and particularly to a neutral beam tomography diagnostic system for NNBI and a priori fusion reconstruction algorithm. Background Technology

[0002] Neutral beam injection (NNBI) technology, as a key auxiliary heating and current-driven method in controlled nuclear fusion devices, relies heavily on the negative ion source, a core component, whose operational status directly determines the transmission efficiency and heating effect of the neutral beam. During the commissioning phase and long-term operation of the NNBI negative ion source, it is crucial to evaluate key parameters such as beam uniformity, divergence angle, beam waist position, and stability in real time and with high precision. This provides data support for parameter optimization, troubleshooting, and long-term reliable operation of the device.

[0003] Tomographic imaging technology, with its advantage of non-invasive three-dimensional measurement, has become the mainstream technology for beam diagnostics of NNBI negative ion sources. However, existing tomographic-based beam diagnostic solutions still face many technical bottlenecks in engineering applications, making it difficult to meet the requirements for high-precision and long-term stable diagnostics. Specific problems are as follows:

[0004] First, insufficient field of view coverage and poor temporal synchronization. Existing solutions mostly employ single-camera moving scan or multi-camera software-triggered acquisition modes. The former suffers from incomplete field of view coverage and long measurement cycles, while the latter is affected by software trigger delays, failing to meet the data homogeneity requirements of tomographic inversion and thus affecting reconstruction accuracy. Second, poor imaging quality under strong interference environments. The NNBI negative ion source operates with high-voltage pulse discharge and strong plasma radiation, creating a complex environment of strong electromagnetic interference and stray light, significantly reducing the image signal-to-noise ratio (SNR). Simultaneously, the conventional observation window structure has poor sealing reliability and unreasonable light-shielding design, easily leading to aging light leakage or window contamination, further exacerbating image quality degradation and affecting the long-term stable operation of the diagnostic system. Third, an imperfect calibration system leads to the accumulation of systematic errors. During long-term operation, factors such as camera mounting attitude drift, observation window contamination, and changes in the response characteristics of optical components can cause drift in the geometric extrapolation photometric characteristics. Existing solutions lack repeatable and automated geometric / photometric calibration procedures and have not established an effective background subtraction mechanism, leading to the continuous accumulation of systematic errors caused by these drifts, severely reducing the accuracy of beam parameter evaluation. Fourth, the robustness of reconstruction under finite angle and noise conditions is insufficient. Due to the structural space limitations of the NNBI device, the camera position of the tomographic imaging system cannot achieve full-angle coverage, resulting in a finite-angle projection scene. Simultaneously, Poisson noise and image sensor readout noise are unavoidable during beam imaging. Existing reconstruction algorithms (such as filtered back projection FBP, algebraic reconstruction ART, etc.) are prone to producing obvious fringe artifacts under the combined effects of finite angle and noise, leading to distortion in beam uniformity evaluation and large fluctuations in divergence angle quantization results, failing to provide a reliable basis for device parameter tuning.

[0005] In summary, existing tomographic imaging diagnostic schemes fail to achieve synergistic optimization of structural design, timing control, and reconstruction algorithms, making it difficult to overcome the aforementioned technical bottlenecks. Therefore, this invention proposes a neutral beam tomographic imaging diagnostic system and a priori fusion reconstruction algorithm for NNBI. Summary of the Invention

[0006] The purpose of this invention is to address the problems in the existing tomographic imaging diagnostic schemes in the background art, such as the failure to achieve synergistic optimization of structural design, timing control and reconstruction algorithms, and to propose a neutral beam tomographic imaging diagnostic system and a priori fusion reconstruction algorithm for NNBI.

[0007] In a first aspect, the present invention provides a neutral beam tomography diagnostic system for NNBI, which is used to realize three-dimensional tomography imaging and state quantification diagnosis of NNBI neutral beam.

[0008] It includes a multi-view imaging unit, a hardware synchronous acquisition unit, and a data processing and reconstruction unit;

[0009] The multi-view imaging unit is arranged at an equal angle at a preset position on the NNBI negative ion source beam source flange to acquire multi-view projection images of the neutral beam.

[0010] The hardware synchronous acquisition unit realizes multi-view image in-phase exposure acquisition based on the trigger signal of the NNBI device;

[0011] The data processing and reconstruction unit incorporates a priori fusion tomography reconstruction algorithm to complete image preprocessing, beam three-dimensional volume field reconstruction, and diagnostic index output.

[0012] Optionally, the multi-view imaging unit includes at least three cameras and matching imaging components, arranged at equal angles of 0°, 90°, 180°, 270° or 0°, 120°, 240°.

[0013] The supporting imaging components include an Hα filter and a labyrinth-style light shield, which are used to improve the imaging signal-to-noise ratio.

[0014] Optionally, the hardware synchronous acquisition unit uses the high-voltage trigger signal or current probe signal of the NNBI device as the time base and drives the camera's global shutter to expose in phase via a time synchronization distributor.

[0015] The exposure gate width is adjustable from 1 to 50 μs, the system end-to-end timing jitter is ≤100 ns, and the relative delay between channels is ≤50 ns.

[0016] Optionally, the data processing and reconstruction unit further includes a preprocessing module, which is used to complete flat field correction, subtract background signals using the bispectral scaling method, and perform exposure time normalization processing.

[0017] Secondly, the present invention provides a neutral beam prior fusion reconstruction algorithm for NNBI, which is applied to the system described in the first aspect. The algorithm is based on the SART iterative framework to fuse multiple types of prior constraints to achieve beam three-dimensional reconstruction.

[0018] This includes establishing a forward projection model based on geometrically calibrated extrinsic parameters, initializing the reconstructed volume field and setting SART iteration parameters, embedding TV regularization constraints and pluggable prior denoising substeps during the iteration process, and outputting the reconstructed volume field and beam diagnostic indicators after the convergence condition is met.

[0019] Optionally, the SART iteration count is 20–100 times, and the relaxation factor is 0.1–0.5;

[0020] The TV regularization parameter β has a value range of 1e-4–1e-2 and is used to suppress stripe artifacts.

[0021] Optionally, the pluggable prior denoising sub-step employs a pre-trained DnCNN or Noise2Inverse deep learning model to match the Poisson and readout noise models of beam imaging, and supports replacing the denoising model according to the noise intensity.

[0022] Optionally, the convergence condition is that the relative decrease of the projection residual is less than 1% or the preset maximum number of iterations is reached;

[0023] The beam diagnostic indicators include beam uniformity index, divergence angle, and beam waist position.

[0024] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein when the program is executed by a processor, the above-described prior fusion tomography reconstruction algorithm steps are performed.

[0025] In summary, this application includes at least one of the following beneficial technical effects:

[0026] This invention employs a hardware-level multi-view synchronous acquisition design, using the HV trigger / current probe signal of the NNBI device as the time base. It drives the global shutter of multiple cameras to expose in phase through a synchronization distributor, achieving high-precision timing control with an adjustable shutter width of 1-50μs, end-to-end jitter ≤100ns, and channel relative delay ≤50ns. Compared with the existing software trigger mode, it completely solves the problem of asynchronous exposure from different viewpoints, ensuring the spatiotemporal homogeneity of multi-view projection data and laying a data foundation for high-precision subsequent tomographic inversion.

[0027] The flange-type observation window and mounting components designed in this invention form a multi-layered anti-interference and protection mechanism: the fused silica observation window, combined with the elastomer O-ring sealing structure, achieves a helium leak detection rate of ≤1×10⁻⁶. 9 High vacuum sealing reliability at Pa·m³ / s; replaceable Hα filter and labyrinth-style light shield effectively suppress stray light; annular micro-purge channel reduces window contamination; full shielding of power supply and signal, 360° shielded grounding of cables, and single-point grounding design significantly reduce the impact of strong electromagnetic interference (EMI); greatly improves image signal-to-noise ratio, solves the problems of easy aging and light leakage of conventional observation windows and the large influence of environmental interference on imaging quality, and ensures long-term stable operation of the diagnostic system;

[0028] Furthermore, this invention constructs an integrated geometric / photometric calibration and background subtraction process. It obtains accurate multi-camera extrinsic parameters through pinhole array or checkerboard calibration benchmarks, and uses extrinsic parameter angle RMS≤0.2° and reprojection error≤0.3 pixels as qualified thresholds. If the limits are exceeded, recalibration is automatically triggered. Combined with dark field or beamless background acquisition and bispectral proportional background subtraction technology, flat field correction and exposure time normalization are completed. It effectively makes up for the deficiencies of existing calibration processes, and can correct parameter drift caused by camera attitude drift, window contamination, etc. in real time during long-term operation, eliminate systematic error accumulation from the root, and improve the accuracy of beam parameter evaluation.

[0029] Furthermore, the prior fusion tomography reconstruction algorithm proposed in this invention is based on the SART iterative framework, with built-in TV regularization constraints and the ability to insert PnP-DnCNN / Noise2Inverse denoising substeps, accurately matching the Poisson plus readout noise model of beam imaging. Compared with the traditional FBP / ART algorithm, it can effectively suppress the stripe artifacts caused by the combination of finite angle projection and noise, and significantly improve the repeatability and stability of core indicators such as beam uniformity index, divergence angle, and beam waist position, providing a reliable diagnostic basis for the debugging and operation of NNBI devices.

[0030] In summary, this invention effectively solves the problems of timing synchronization, anti-interference, calibration and reconstruction robustness in NNBI neutral beam tomography diagnosis through the synergistic optimization of structure, timing and algorithm, thereby improving diagnostic accuracy and stability, and has strong engineering feasibility. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of the acquisition and inversion process of a neutral beam tomography diagnostic system for NNBI. Detailed Implementation

[0032] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0033] The components of the embodiments of the invention described and shown in the accompanying drawings can typically be arranged and designed in a variety of different configurations. Therefore, the following detailed description of the embodiments of the invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention.

[0034] Example

[0035] like Figure 1 As shown, the neutral beam tomography diagnostic system for NNBI proposed in this invention is used to realize three-dimensional tomography imaging and state quantification diagnosis of NNBI neutral beams.

[0036] It includes a multi-view imaging unit, a hardware synchronous acquisition unit, and a data processing and reconstruction unit;

[0037] The multi-view imaging unit is arranged at a preset position on the NNBI negative ion source beam flange at an equal angle to acquire multi-view projection images of the neutral beam. The multi-view imaging unit includes at least 3 cameras and supporting imaging components, arranged at equal angles of 0°, 90°, 180°, 270° or 0°, 120°, 240°. The supporting imaging components include an Hα filter and a labyrinth-type light shield. The supporting imaging components are used to improve the imaging signal-to-noise ratio.

[0038] The hardware synchronous acquisition unit achieves in-phase exposure acquisition of multi-view images based on the trigger signal of the NNBI device. The hardware synchronous acquisition unit uses the high-voltage trigger signal or current probe signal of the NNBI device as the time base, and drives the camera's global shutter for in-phase exposure via a timing synchronization distributor. The exposure shutter width adjustment range is 1-50μs, the system end-to-end timing jitter is ≤100ns, and the relative delay between channels is ≤50ns. Specifically, timing synchronization begins with the high-voltage trigger or current rising edge signal extracted from the beam device. This signal first enters the isolated timing distributor, is converted into a synchronization pulse, and then sent to the trigger input port of each camera, with a phase fine-tuning resolution of 10ns. All cameras operate in global shutter mode. The data and power supply network is connected to an industrial-grade switch via fiber optic cable, ultimately converging to the main control computer. Camera power is delivered via independent shielded cables, using metal corrugated sheaths and IP65-rated glands to enter the equipment compartment. For electromagnetic compatibility, all cable shielding layers are grounded 360° around the ground. The equipment chassis and brackets are connected to the rack ground via short, thick grounding wires. Common-mode chokes and transient voltage suppressors are installed at the power inlet to effectively suppress interference and surges on the power lines. The main control computer is responsible for executing unified parameter distribution, including exposure gate width, gain, photosensitive area, and exposure delay, while simultaneously monitoring the system's health status, embedding precise timestamps and trigger counts for each frame, and implementing frame drop detection and retry mechanisms. Acquired high-speed image data is cached in real-time on an NVMe solid-state drive array, with a continuous write speed exceeding 1GB / s. Simultaneously, the software automatically generates experimental metadata files containing information such as geometric configuration, filter parameters, and calibration file version numbers, providing a complete context for subsequent data processing.

[0039] The data processing and reconstruction unit incorporates a priori fusion tomography reconstruction algorithm to complete image preprocessing, beam three-dimensional volume field reconstruction, and diagnostic index output. The data processing and reconstruction unit also includes a preprocessing module, which is used to complete flat field correction, subtract background signals using the bispectral scaling method, and perform exposure time normalization. Specifically, precise geometric calibration is required before data processing. During the system maintenance window, a pinhole array or checkerboard calibration board is placed in the beam channel. By capturing multiple frames of images, the intrinsic parameters of the camera, the lens distortion coefficient, and the extrinsic parameters of each camera relative to the global coordinate system are calculated. Cross-verification is performed using a beam interceptor or collimator to ensure that the extrinsic parameter angle RMS ≤ 0.2° and the reprojection error ≤ 0.3 pixels. If these thresholds are exceeded, the system will automatically trigger the recalibration process. Photometric calibration and background subtraction are performed by acquiring background images under dark field and no-beam conditions to complete pixel-level flat field correction. A bispectral scaling method is used, that is, the background channel signal is proportionally subtracted from the Hα channel image, or the bandwidth difference between the two channels is used for proportional subtraction. Finally, exposure time normalization is performed on all images.

[0040] In this embodiment, the algorithm applied to the above-mentioned neutral beam tomography diagnostic system for NNBI is a neutral beam prior fusion reconstruction algorithm for NNBI that integrates multiple types of prior constraints based on the SART iterative framework to achieve beam three-dimensional reconstruction.

[0041] The process includes establishing a forward projection model based on geometrically calibrated extrinsic parameters, initializing the reconstructed volume field and setting SART iteration parameters. The number of SART iterations is 20–100, and the relaxation factor is 0.1–0.5. During the iteration process, TV regularization constraints and pluggable prior denoising substeps are embedded. After the convergence condition is met, the reconstructed volume field and beam diagnostic indicators are output. The TV regularization parameter β takes a value range of 1e-4–1e-2 to suppress fringe artifacts.

[0042] The pluggable prior denoising sub-step employs a pre-trained DnCNN or Noise2Inverse deep learning model, matching the Poisson and readout noise models of beam imaging, and supports replacing the denoising model based on noise intensity; the convergence condition is that the relative decrease of the projection residual is less than 1% or the preset maximum number of iterations is reached; the beam diagnostic indicators include the beam uniformity index, divergence angle, and beam waist position. Specifically, in the projection generation and tomographic reconstruction stages, an accurate forward projection model is established based on the calibrated extrinsic parameters, mapping the corrected two-dimensional images from multiple perspectives to the integral intensity of the projection lines at the corresponding perspectives. To further improve reconstruction quality, the algorithm's pluggable prior module is embedded in a pre-trained deep learning model to perform a denoising sub-step after each SART iteration update. This plug-and-play design effectively improves the signal-to-noise ratio while maintaining physical consistency. The convergence criterion for the iterative process is set to a decrease in the current projection residual of <1% or reaching the preset maximum number of iterations. After reconstruction, the system outputs a series of quantitative indicators, including an exponent for characterizing uniformity, a divergence angle (RMS / full width at half maximum) obtained by Gaussian fitting of the reconstructed isosurface or radial profile, the waist position, and its ellipse fitting parameters. The entire reconstruction process supports GPU acceleration, meeting the requirements for near real-time diagnostics.

[0043] Furthermore, the present invention also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of the priori fusion tomography reconstruction algorithm described above.

[0044] Specifically, basic hardware deployment is completed on the flange of the NNBI negative ion source beam. High vacuum sealing and optical adaptation are achieved through a flange-type observation window assembly. 3-8 cameras are arranged at equal angles of 0°, 90°, 180°, and 270°. The cameras are fixed with adjustable brackets and grounded for interference suppression, ensuring stable positioning of the imaging hardware and meeting multi-view acquisition requirements. The system uses the high-voltage trigger signal or current probe signal of the NNBI device as the time base, generating synchronization pulses through a timing synchronizer to drive all cameras to operate in global shutter mode for in-phase exposure. The exposure shutter width can be adjusted within the range of 1-50μs.

[0045] After debugging, the system's end-to-end timing jitter is ≤100ns, and the relative delay between channels is ≤50ns, ensuring the spatiotemporal homogeneity of multi-view projection data.

[0046] The acquired image data is transmitted to an industrial switch via fiber optic cable, then aggregated and cached on the main controller's NVMe solid-state drive array to ensure stable high-speed data storage. Simultaneously, the main controller issues unified acquisition parameters, performs image timestamp embedding, frame loss detection, and experimental metadata generation, providing a complete data context for subsequent processing. The power and signal transmission links employ full shielding and 360° surround grounding designs to effectively suppress strong electromagnetic interference.

[0047] In this embodiment, before formal diagnosis and during operation and maintenance, the system initiates an integrated calibration process: by photographing a pinhole array or checkerboard calibration board, the camera's intrinsic parameters, distortion coefficients, and extrinsic parameters are calculated. After cross-verification by a beam interceptor, the extrinsic parameter angle RMS is ensured to be ≤0.2° and the reprojection error is ≤0.3 pixels. If the limits are exceeded, automatic recalibration is performed.

[0048] In the data preprocessing stage, the system acquires dark field images and beamless background images to complete flat field correction. The background channel signal is subtracted from the Hα channel image using the bispectral scaling method. Then, the exposure time is normalized, and finally, a clean beam projection image with systematic errors is output.

[0049] Implementation of the priori fusion tomography reconstruction algorithm:

[0050] Based on the calibrated extrinsic parameters, a mapping relationship between the three-dimensional beam fluid field and the two-dimensional projection is established, and the preprocessed multi-view two-dimensional image is converted into the projection line integral intensity data of the corresponding view.

[0051] After initializing the reconstructed volume field, the SART iterative framework is used for reconstruction, with the number of iterations set to 20-100 and the relaxation factor set to 0.1-0.5. In each iteration, TV regularization (β=1e-4-1e-2) is embedded to suppress fringe artifacts. Then, a pluggable prior denoising sub-step is inserted, and a pre-trained DnCNN or Noise2Inverse model is called for denoising, matching the Poisson plus readout noise model of beam imaging.

[0052] The iteration stops when the relative decrease of the projection residual is less than 1% or the preset maximum number of iterations is reached. The system performs post-processing on the converged 3D volume field, calculating and outputting quantitative indicators such as the beam uniformity index, divergence angle, and beam waist position. The entire reconstruction process supports GPU acceleration, meeting the requirements for near real-time diagnostics.

[0053] The above specific embodiments are merely several optional embodiments of the present invention. Based on the technical solutions of the present invention and the relevant teachings of the above embodiments, those skilled in the art can make various alternative improvements and combinations to the above specific embodiments.

Claims

1. A neutral beam tomography diagnostic system for NNBI, characterized in that, Used to achieve three-dimensional tomographic imaging and state quantification diagnosis of NNBI neutral beam; It includes a multi-view imaging unit, a hardware synchronous acquisition unit, and a data processing and reconstruction unit; The multi-view imaging unit is arranged at an equal angle at a preset position on the NNBI negative ion source beam source flange to acquire multi-view projection images of the neutral beam. The hardware synchronous acquisition unit realizes multi-view image in-phase exposure acquisition based on the trigger signal of the NNBI device; The data processing and reconstruction unit incorporates a priori fusion tomography reconstruction algorithm to complete image preprocessing, beam three-dimensional volume field reconstruction, and diagnostic index output.

2. The neutral beam tomography diagnostic system for NNBI according to claim 1, characterized in that, The multi-view imaging unit includes at least three cameras and matching imaging components, arranged at equal angles of 0°, 90°, 180°, 270° or 0°, 120°, 240°. The supporting imaging components include an Hα filter and a labyrinth-style light shield, which are used to improve the imaging signal-to-noise ratio.

3. The neutral beam tomography diagnostic system for NNBI according to claim 1, characterized in that, The hardware synchronous acquisition unit uses the high-voltage trigger signal or current probe signal of the NNBI device as the time base, and drives the camera's global shutter to expose in phase via a time synchronization distributor. The exposure gate width is adjustable from 1 to 50 μs, the system end-to-end timing jitter is ≤100 ns, and the relative delay between channels is ≤50 ns.

4. The neutral beam tomography diagnostic system for NNBI according to claim 1, characterized in that, The data processing and reconstruction unit also includes a preprocessing module, which is used to complete flat field correction, subtract background signals using the bispectral scaling method, and perform exposure time normalization.

5. A neutral bundle prior fusion reconstruction algorithm for NNBI, characterized in that, Applied to the system according to any one of claims 1-4, the algorithm is based on the SART iterative framework to fuse multiple types of prior constraints to achieve beam three-dimensional reconstruction; This includes establishing a forward projection model based on geometric calibration extrinsic parameters, initializing the reconstructed volume field and setting SART iteration parameters, embedding TV regularization constraints and pluggable prior denoising substeps during the iteration process, and outputting the reconstructed volume field and beam diagnostic indicators after the convergence condition is met.

6. The neutral bundle prior fusion reconstruction algorithm for NNBI according to claim 5, characterized in that, The SART iteration count is 20–100 times, and the relaxation factor is 0.1–0.

5. The TV regularization parameter β has a value range of 1e-4–1e-2 and is used to suppress stripe artifacts.

7. The neutral bundle prior fusion reconstruction algorithm for NNBI according to claim 6, characterized in that, The pluggable prior denoising sub-step employs a pre-trained DnCNN or Noise2Inverse deep learning model, matches the Poisson and readout noise models of beam imaging, and supports replacing the denoising model based on the noise intensity.

8. The neutral bundle prior fusion reconstruction algorithm for NNBI according to claim 7, characterized in that, The convergence condition is that the relative decrease of the projection residual is less than 1% or the preset maximum number of iterations is reached. The beam diagnostic indicators include beam uniformity index, divergence angle, and beam waist position.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it performs the steps of the priori fusion tomography reconstruction algorithm as described in any one of claims 5-8.

Images