Accelerator control method and system based on AI control

Abstract

The invention discloses an accelerator control method and system based on AI control, and relates to the field of accelerator control, and the method comprises the steps: carrying out the phase control array particle beam damage detection of a historical damage sample, and building an initial phase recognition model of a phase control array emission particle beam based on the feedback data of the detected sample; based on the physical characteristics of the multilayer material sample, calculating the particle beam energy distribution requirement of each layer of the multilayer material sample, and establishing a sequential control model of a particle beam phase control array emission unit; the method comprises the following steps: preliminarily determining a material damage position through ultrasonic detection, and performing initial phase setting on a phase control array particle beam based on an initial phase recognition model; based on the material characteristics of the detected material, the emission time sequence of the phase control array particle beam is controlled through a time sequence control model; accelerator parameters are manipulated in real time according to control requirements of initial phase setting and emission time sequence of a phase control array particle beam.

Description

Technical Field

[0001] The present invention relates to the field of accelerator control, and in particular to an accelerator control method and system based on AI control. Background Art

[0002] In industrial nondestructive testing, especially the testing of multilayer materials, composite materials and heterogeneous structures, traditional testing methods (such as ultrasonic, X-ray and laser testing) often fail to meet the requirements of high resolution and deep penetration. Dynamic phased beam control technology can perform efficient and accurate nondestructive testing of complex materials by adjusting the direction, intensity and focus of the particle beam. Therefore, it has become a key technology in the current field of nondestructive testing. However, traditional phased beam control mainly relies on manually set parameters and lacks intelligent phase control. In addition, in the testing of multilayer materials, due to the interaction between layers, the energy of the particle beam is easily attenuated during transmission, resulting in uneven penetration of each layer of material. Traditional phased systems are difficult to accurately distribute energy, resulting in unstable test results.

[0003] The present invention proposes an accelerator control method based on AI control. By analyzing historical detection data, a rapid identification model for the initial phase of particle beam emission in the damaged area is established. At the same time, based on the idea of ​​mutual interference and superposition of particle beam peaks, the particle beam injection intensity in multi-layer material detection is dynamically controlled according to the emission timing of the particle beam, thereby realizing an accelerator parameter control method that can meet the requirements of intelligent, rapid and accurate damage identification of multi-layer materials. Summary of the invention

[0004] In order to solve the above technical problems, an accelerator control method based on AI control is provided. This technical solution solves the problems raised in the above background technology.

[0005] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0006] An accelerator control method based on AI control, comprising:

[0007] Perform phased array particle beam damage detection on historical damaged samples, and establish an initial phase recognition model for the phased array particle beam based on the feedback data of the detected samples;

[0008] Based on the physical properties of multi-layer material samples, the particle beam energy distribution requirements of each layer of the multi-layer material sample are calculated, and the timing control model of the particle beam phased array emission unit is established;

[0009] The material damage location is preliminarily determined through ultrasonic testing, and the initial phase of the phased array particle beam is set based on the initial phase recognition model;

[0010] Based on the material properties of the material being tested, the emission timing of the phased array particle beam is controlled through a timing control model;

[0011] The initial phase setting and emission timing of the phased array particle beam are controlled to meet the requirements and manipulate the accelerator parameters in real time.

[0012] Preferably, the phased array particle beam damage detection is performed on the historical damage samples, and based on the feedback data of the detected samples, an initial phase recognition model of the phased array emission particle beam is established, which specifically includes:

[0013] Select at least one historical damage detection sample, and obtain the material type, material properties, and damage type of the detection sample;

[0014] The surface of the sample to be tested is covered and scanned by emitting a phased array particle beam, and backscattered particle imaging is performed according to the backscattered particle signal intensity generated on the surface of the sample to be tested during the scanning process;

[0015] Obtain the damaged area of ​​the imaging result image, mark the coordinates of the damaged area, and record the real-time phase angle of each phased array transmitting unit when the particle beam is emitted to the damaged area;

[0016] The material type, material properties, damage type and damage area coordinates of the test sample are processed as feature input. The real-time phase angle of each emission unit of the phased array when the particle beam is emitted to the damage area is used as feature output to establish a phase recognition data set for the test sample.

[0017] Select a machine learning algorithm, train the initial phase recognition model of the particle beam phased array transmitting unit based on the detection sample phase recognition data set, and deploy the trained AI model to the application system.

[0018] Preferably, the step of calculating the particle beam energy distribution requirements of each layer of the multilayer material sample based on the physical properties of the multilayer material sample and establishing the timing control model of the particle beam phased array emission unit specifically includes:

[0019] Setting a material detection stratification scale, marking samples with a thickness exceeding the material detection stratification scale as multi-layer material samples, stratifying the multi-layer material samples based on the material detection stratification scale, and detecting the physical properties of each layer of the multi-layer material samples at the same time;

[0020] Particle beam penetration analysis of multi-layer material samples;

[0021] The damaged area and the density and thickness of each layer of the current multi-layer material sample are obtained, and the particle beam is transmitted through the phased array transmitting unit to penetrate the damaged area, and the absorption coefficient of each layer of the multi-layer material sample is calculated. The specific calculation method is:

[0022]

[0023] In the formula, μ j is the absorption coefficient of the jth layer of the multilayer material sample, which is used to express the absorption characteristics of each layer of the multilayer material sample. j , j d j are the actual particle beam energy, density and thickness reaching the jth layer of the multilayer material sample, respectively;

[0024] Based on the absorption coefficient of each layer of the multilayer material sample and the actual particle beam intensity, an energy optimization allocation model is established to calculate the energy demand of each layer layer by layer. The energy optimization allocation model is:

[0025]

[0026] In the formula, I j is the particle beam intensity requirement for injecting into the jth layer of the multilayer material sample, I0 is the initial intensity of the injected particle beam, μ k d k , k are the absorption coefficient, thickness and density of the kth layer of the multilayer material sample, respectively;

[0027] Identify the initial phase of the particle beam phased array emission unit of the current multi-layer material sample through an initial phase recognition model of the particle beam emission;

[0028] The timing control model of the particle beam phased array emission unit is established through the phase-timing joint control algorithm, specifically:

[0029]

[0030] Where I(t,x) is the required particle beam intensity at time t for a multilayer material sample with damage coordinate x based on the time series corresponding to the set layering scale, Δt i is the transmission delay of the i-th transmitting unit of the phased array, which is used to dynamically adjust the transmission timing. i 、w i and φ i are the particle beam intensity coefficient, particle beam emission frequency, and initial phase identification result of the damage coordinate x of the i-th emission unit of the phased array, respectively, and N is the total number of units in the phased array.

[0031] Preferably, the method of preliminarily determining the material damage location by ultrasonic detection and performing initial phase setting on the phased array particle beam based on the initial phase recognition model specifically includes:

[0032] Obtain the material type and material properties of the material entity to be tested;

[0033] Through ultrasonic testing, the damaged area and damage type of the material to be tested are preliminarily detected and quickly calibrated;

[0034] The material type, material properties, damage type and damage area coordinate data of the material entity to be detected are processed and input into the initial phase recognition model to obtain the initial phase of the particle beam emission phased array emission unit of the material entity to be detected.

[0035] Preferably, the controlling the emission timing of the phased array particle beam by a timing control model based on the material properties of the detected material specifically includes:

[0036] Layer the material entity to be tested and test the physical properties of each layer of the material entity;

[0037] Calculate the absorption coefficient of each layer of the material entity to be tested, and bring the absorption coefficient of each layer into the energy optimization allocation model to calculate the energy demand of each layer of the material entity to be tested;

[0038] Based on the set material detection layer scale, the energy demand of each layer is mapped to the time series corresponding to the layer scale, and the particle beam intensity and particle beam emission frequency of each emission unit of the current phased array are obtained;

[0039] The time series energy demand of the current material entity to be detected, as well as the particle beam intensity, particle beam emission frequency and initial phase of each emission unit of the phased array are brought into the timing control model of the particle beam phased array emission unit to calculate the real-time emission delay of each emission unit of the phased array.

[0040] Preferably, the control requirement of setting the initial phase and emission timing of the phased array particle beam and manipulating the accelerator parameters in real time specifically includes:

[0041] Obtaining the initial phase of the particle beam emission phased array for the current material entity to be detected and the real-time emission delay of each emission unit of the phased array;

[0042] The accelerator operating parameters are dynamically controlled according to the initial phase and real-time emission delay of each emission unit of the phased array, so that the accelerator meets the emission requirements of the particle beam emission phased array.

[0043] Furthermore, an accelerator control system based on AI control is proposed, which is used to implement the above-mentioned accelerator control method based on AI control, including:

[0044] A phase recognition model building module, wherein the phase recognition model building module detects the surface of the detection sample by emitting a phased array particle beam, marks the coordinates of the damaged area based on the backscattered particle imaging results, records the real-time phase angle of each phased array emitting unit when the particle beam is emitted to the damaged area, and trains an initial phase recognition model of the particle beam phased array emitting unit;

[0045] A timing control model establishment module, wherein the timing control model establishment module performs particle beam penetration analysis on a multilayer material sample, calculates the absorption coefficient of each layer of the multilayer material sample, establishes an energy optimization allocation model based on the absorption coefficient of each layer of the multilayer material sample and the actual particle beam intensity injected, calculates the energy demand of each layer layer by layer, identifies the initial phase of the particle beam phased array emission unit of the current multilayer material sample through an initial phase recognition model of particle beam emission, and finally establishes a timing control model of the particle beam phased array emission unit through a phase-timing joint control algorithm;

[0046] An initial phase setting module, which performs preliminary detection and rapid calibration of the damage area and damage type of the material entity to be detected by ultrasonic detection, performs feature processing on the feature data corresponding to the material entity to be detected, and then inputs the feature data into the initial phase recognition model to obtain the initial phase of the particle beam emission phased array emission unit of the material entity to be detected;

[0047] A timing control module, wherein the timing control module brings the time series energy demand of the current material entity to be detected, the particle beam intensity of each emission unit of the phased array, the particle beam emission frequency and the initial phase of the particle beam emission phased array emission unit into the timing control model of the particle beam phased array emission unit, and calculates the real-time emission delay of each emission unit of the phased array;

[0048] An accelerator control module dynamically controls the accelerator operating parameters according to the initial phase and real-time transmission delay of each transmitting unit of the phased array.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] The material damage location is preliminarily determined through ultrasonic testing, and the initial phase of the phased array particle beam is set based on the initial phase recognition model; based on the material properties of the material being tested, the emission timing of the phased array particle beam is controlled through the timing control model; the initial phase setting and emission timing of the phased array particle beam are used to control the needs and manipulate the accelerator parameters in real time.

[0051] By analyzing historical detection data, a rapid identification model for the initial phase of particle beam emission in the damaged area is established. At the same time, based on the idea of ​​mutual interference and superposition of particle beam peaks, the particle beam injection intensity in multi-layer material detection is dynamically controlled according to the emission timing of the particle beam, realizing an accelerator parameter control method that can meet the requirements of intelligent, rapid and accurate damage identification of multi-layer materials. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is a flow chart of an accelerator control method based on AI control of the present invention;

[0053] Figure 2 A flow chart of an initial phase recognition model for establishing a phased array emitting a particle beam according to the present invention;

[0054] Figure 3 To calculate the particle beam energy distribution requirements of each layer of the multi-layer material sample of the present invention, a flow chart of the timing control model of the particle beam phased array emission unit is established;

[0055] Figure 4 This is a flow chart of setting the initial phase of a phased array particle beam based on an initial phase recognition model of the present invention;

[0056] Figure 5 This is a flow chart of controlling the emission timing of a phased array particle beam through a timing control model based on the material properties of the material being detected in the present invention;

[0057] Figure 6 A flow chart for real-time manipulation of accelerator parameters for the present invention to control the initial phase setting and emission timing of the phased array particle beam;

[0058] Figure 7 The figure is a schematic diagram of the structure of an accelerator control system based on AI control of the present invention. DETAILED DESCRIPTION

[0059] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art may think of other obvious variations.

[0060] Reference Figure 1 As shown, an accelerator control method based on AI control includes:

[0061] Perform phased array particle beam damage detection on historical damaged samples, and establish an initial phase recognition model for the phased array particle beam based on the feedback data of the detected samples;

[0062] Based on the physical properties of multi-layer material samples, the particle beam energy distribution requirements of each layer of the multi-layer material sample are calculated, and the timing control model of the particle beam phased array emission unit is established;

[0063] The material damage location is preliminarily determined through ultrasonic testing, and the initial phase of the phased array particle beam is set based on the initial phase recognition model;

[0064] Based on the material properties of the material being tested, the emission timing of the phased array particle beam is controlled through a timing control model;

[0065] The initial phase setting and emission timing of the phased array particle beam are controlled to meet the requirements and manipulate the accelerator parameters in real time.

[0066] Reference Figure 2 As shown, a phased array particle beam damage detection is performed on historical damaged samples, and an initial phase recognition model of the phased array emitted particle beam is established based on the feedback data of the detected samples.

[0067] Select at least one historical damage detection sample to obtain the material type, material properties and damage type of the detection sample. The number of historical damage detection samples to be selected needs to be set according to the actual detection accuracy requirements and material types. The damage area superposition of all samples must cover the entire detection plane. The more samples selected, the higher the model recognition accuracy.

[0068] The surface of the test sample is covered and scanned by emitting a phased array particle beam, and backscattered particle imaging is performed based on the backscattered particle signal intensity generated on the surface of the test sample during the scanning process. For example, when the electron beam irradiates the surface of the material, elastic scattering and inelastic scattering will occur, causing a portion of the electrons to be reflected back at different angles. The backscattered electrons can be used to analyze the composition and structure of the material.

[0069] The damaged area of ​​the imaging result image is obtained, the coordinates of the damaged area are marked, and the real-time phase angle of each phased array transmitting unit when the particle beam is emitted to the damaged area is recorded.

[0070] The material type, material properties, damage type and damage area coordinates of the test sample are processed as feature input. The real-time phase angle of each emission unit of the phased array when the particle beam is emitted to the damage area is used as feature output to establish a phase recognition data set for the test sample.

[0071] Select a machine learning algorithm, train the initial phase recognition model of the particle beam phased array transmitting unit based on the detection sample phase recognition data set, and deploy the trained AI model to the application system.

[0072] In the whole process of modeling, training, optimization and application, AI builds a set of intelligent control models for dynamic phased particle beams based on the characteristic data of damaged samples. This model significantly improves the detection efficiency and accuracy, enabling the system to achieve fast and accurate damage identification and positioning in complex industrial non-destructive testing scenarios.

[0073] Reference Figure 3 As shown, based on the physical properties of the multi-layer material sample, the particle beam energy distribution requirements of each layer of the multi-layer material sample are calculated, and the timing control model of the particle beam phased array emission unit is established, which specifically includes:

[0074] The main challenge of multi-layer material detection is that the density, thickness and absorption characteristics of each layer of material are different, and the energy requirements of the particle beam are different. The AI ​​system first analyzes the basic characteristics of each layer of material and establishes an energy requirement model to ensure that the particle beam can effectively penetrate each layer. Through the joint control of phase and timing, the particle beam is accurately superimposed in space and time, improving the specific intensity focus control of the target area.

[0075] Set the material detection stratification scale, mark the samples with thickness exceeding the material detection stratification scale as multi-layer material samples, stratify the multi-layer material samples based on the material detection stratification scale, and detect the physical properties of each layer of the multi-layer material samples at the same time, including density, thickness, absorbance, etc.;

[0076] Particle beam penetration analysis of multi-layer material samples;

[0077] The damaged area and the density and thickness of each layer of the current multi-layer material sample are obtained, and the particle beam is transmitted through the phased array transmitting unit to penetrate the damaged area, and the absorption coefficient of each layer of the multi-layer material sample is calculated. The specific calculation method is:

[0078]

[0079] In the formula, μ j is the absorption coefficient of the jth layer of the multilayer material sample, which is used to express the absorption characteristics of each layer of the multilayer material sample. j , j d j are the actual particle beam energy, density and thickness reaching the jth layer of the multilayer material sample, respectively;

[0080] Based on the absorption coefficient of each layer of the multilayer material sample and the actual particle beam intensity, an energy optimization allocation model is established to calculate the energy demand of each layer layer by layer. The energy optimization allocation model is:

[0081]

[0082] In the formula, I jis the particle beam intensity requirement for injecting into the jth layer of the multilayer material sample, I0 is the initial intensity of the injected particle beam, μ k d k , k are the absorption coefficient, thickness and density of the kth layer of the multilayer material sample, respectively;

[0083] Identify the initial phase of the particle beam phased array emission unit of the current multi-layer material sample through an initial phase recognition model of the particle beam emission;

[0084] The timing control model of the particle beam phased array emission unit is established through the phase-timing joint control algorithm, specifically:

[0085]

[0086] Where I(t,x) is the required particle beam intensity at time t for a multilayer material sample with damage coordinate x based on the time series corresponding to the set layering scale, Δt i is the transmission delay of the i-th transmitting unit of the phased array, which is used to dynamically adjust the transmission timing. i 、w i and φ i are the particle beam intensity coefficient, particle beam emission frequency, and initial phase identification result of the damage coordinate x of the i-th emission unit of the phased array, respectively, and N is the total number of units in the phased array.

[0087] Controlling the intensity of a particle beam by adjusting the timing of particle emission is an efficient method, especially in phased array systems, which can accurately control the intensity of a particle beam at a specific location. The core of emission timing control is to adjust the arrival and superposition of particles in time, thereby enhancing or weakening the overall intensity of the particle beam.

[0088] Reference Figure 4 As shown, the material damage location is preliminarily determined by ultrasonic testing, and the initial phase setting of the phased array particle beam is performed based on the initial phase recognition model, which specifically includes:

[0089] Obtain the material type and material properties of the material entity to be tested;

[0090] Ultrasonic testing is used to conduct preliminary detection and rapid calibration of the damaged area and damage type of the material entity to be tested. Based on the fast speed of ultrasonic waves and the portable detection characteristics of the equipment, it is suitable for use in scenarios where rapid preliminary screening is used.

[0091] However, in the detection of deep defects and high-density materials, particle beam detection (such as X-ray, proton, neutron beam) is usually more accurate than ultrasonic detection, especially when it comes to three-dimensional imaging, particle beam can provide higher resolution and clearer structural images. In surface or near-surface detection, electron beam has higher resolution and is more suitable for fine surface analysis.

[0092] The material type, material properties, damage type and damage area coordinate data of the material entity to be detected are processed and input into the initial phase recognition model to obtain the initial phase of the particle beam emission phased array emission unit of the material entity to be detected.

[0093] Reference Figure 5 As shown, based on the material properties of the material being tested, the emission timing of the phased array particle beam is controlled by the timing control model, specifically including:

[0094] Layer the material entity to be tested and test the physical properties of each layer of the material entity;

[0095] Calculate the absorption coefficient of each layer of the material entity to be tested, and bring the absorption coefficient of each layer into the energy optimization allocation model to calculate the energy demand of each layer of the material entity to be tested;

[0096] Based on the set material detection layer scale, the energy demand of each layer is mapped to the time series corresponding to the layer scale, and the particle beam intensity and particle beam emission frequency of each emission unit of the current phased array are obtained;

[0097] The time series energy demand of the current material entity to be detected, as well as the particle beam intensity, particle beam emission frequency and initial phase of each emission unit of the phased array are brought into the timing control model of the particle beam phased array emission unit to calculate the real-time emission delay of each emission unit of the phased array.

[0098] Reference Figure 6 As shown, the initial phase setting and emission timing of the phased array particle beam are used to control the requirements, and the real-time manipulation of accelerator parameters specifically includes:

[0099] Obtaining the initial phase of the particle beam emission phased array for the current material entity to be detected and the real-time emission delay of each emission unit of the phased array;

[0100] In a phased array system, the accelerator can control the phase of each acceleration unit or electromagnetic unit so that the particle beams of each unit are coherently superimposed at the target location. Phased array technology can achieve control of different directions and intensities by adjusting the phase offset of each unit.

[0101] In some accelerators, the arrival time of the particle beam can be controlled by adding delay lines or adjustable electromagnetic delays on different emission paths, so that different particle groups can form a desired time superposition at the target to achieve the control of the particle beam injection intensity.

[0102] The accelerator operating parameters are dynamically controlled according to the initial phase and real-time emission delay of each emission unit of the phased array, so that the accelerator meets the emission requirements of the particle beam emission phased array.

[0103] Reference Figure 7 As shown, an accelerator control system based on AI control includes:

[0104] A phase recognition model building module, wherein the phase recognition model building module detects the surface of the detection sample by emitting a phased array particle beam, marks the coordinates of the damaged area based on the backscattered particle imaging results, records the real-time phase angle of each phased array emitting unit when the particle beam is emitted to the damaged area, and trains an initial phase recognition model of the particle beam phased array emitting unit;

[0105] A timing control model establishment module, wherein the timing control model establishment module performs particle beam penetration analysis on a multilayer material sample, calculates the absorption coefficient of each layer of the multilayer material sample, establishes an energy optimization allocation model based on the absorption coefficient of each layer of the multilayer material sample and the actual particle beam intensity injected, calculates the energy demand of each layer layer by layer, identifies the initial phase of the particle beam phased array emission unit of the current multilayer material sample through an initial phase recognition model of particle beam emission, and finally establishes a timing control model of the particle beam phased array emission unit through a phase-timing joint control algorithm;

[0106] An initial phase setting module, which performs preliminary detection and rapid calibration of the damage area and damage type of the material entity to be detected by ultrasonic detection, performs feature processing on the feature data corresponding to the material entity to be detected, and then inputs the feature data into the initial phase recognition model to obtain the initial phase of the particle beam emission phased array emission unit of the material entity to be detected;

[0107] A timing control module, wherein the timing control module brings the time series energy demand of the current material entity to be detected, the particle beam intensity of each emission unit of the phased array, the particle beam emission frequency and the initial phase of the particle beam emission phased array emission unit into the timing control model of the particle beam phased array emission unit, and calculates the real-time emission delay of each emission unit of the phased array;

[0108] An accelerator control module dynamically controls the accelerator operating parameters according to the initial phase and real-time transmission delay of each transmitting unit of the phased array.

[0109] Furthermore, the present solution also proposes a storage medium for an accelerator control method based on AI control, on which a computer-readable program is stored. When the computer-readable program is called, the above-mentioned accelerator control method based on AI control is executed.

[0110] It is understandable that the storage medium may be a magnetic medium, such as a floppy disk, a hard disk, or a magnetic tape; an optical medium, such as a DVD; or a semiconductor medium, such as a solid state drive (SSD).

[0111] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the attached claims and their equivalents.

Claims

1. An accelerator control method based on AI control, characterized in that: include: Perform phased array particle beam damage detection on historical damaged samples, and establish an initial phase recognition model for the phased array particle beam based on the feedback data of the detected samples; Based on the physical properties of multi-layer material samples, the particle beam energy distribution requirements of each layer of the multi-layer material sample are calculated, and the timing control model of the particle beam phased array emission unit is established; The material damage location is preliminarily determined through ultrasonic testing, and the initial phase of the phased array particle beam is set based on the initial phase recognition model; Based on the material properties of the material being tested, the emission timing of the phased array particle beam is controlled through a timing control model; The initial phase setting and emission timing of the phased array particle beam are controlled to meet the requirements and manipulate the accelerator parameters in real time.

2. The accelerator control method based on AI control according to claim 1, characterized in that: The phased array particle beam damage detection is performed on the historical damage samples, and based on the feedback data of the detected samples, an initial phase recognition model of the phased array emission particle beam is established, which specifically includes: Select at least one historical damage detection sample, and obtain the material type, material properties, and damage type of the detection sample; The surface of the sample to be tested is covered and scanned by emitting a phased array particle beam, and backscattered particle imaging is performed according to the backscattered particle signal intensity generated on the surface of the sample to be tested during the scanning process; Obtain the damaged area of ​​the imaging result image, mark the coordinates of the damaged area, and record the real-time phase angle of each phased array transmitting unit when the particle beam is emitted to the damaged area; The material type, material properties, damage type and damage area coordinates of the test sample are processed as feature input. The real-time phase angle of each emission unit of the phased array when the particle beam is emitted to the damage area is used as feature output to establish a phase recognition data set for the test sample. Select a machine learning algorithm, train the initial phase recognition model of the particle beam phased array transmitting unit based on the detection sample phase recognition data set, and deploy the trained AI model to the application system.

3. The accelerator control method based on AI control according to claim 2, characterized in that: The method of calculating the particle beam energy distribution requirements of each layer of the multilayer material sample based on the physical properties of the multilayer material sample and establishing the timing control model of the particle beam phased array emission unit specifically includes: Setting a material detection stratification scale, marking samples with a thickness exceeding the material detection stratification scale as multi-layer material samples, stratifying the multi-layer material samples based on the material detection stratification scale, and detecting the physical properties of each layer of the multi-layer material samples at the same time; Particle beam penetration analysis of multi-layer material samples; The damaged area and the density and thickness of each layer of the current multi-layer material sample are obtained, and the particle beam is transmitted through the phased array transmitting unit to penetrate the damaged area, and the absorption coefficient of each layer of the multi-layer material sample is calculated. The specific calculation method is: In the formula, μ j is the absorption coefficient of the jth layer of the multilayer material sample, which is used to express the absorption characteristics of each layer of the multilayer material sample. j , j d j are the actual particle beam energy, density and thickness reaching the jth layer of the multilayer material sample, respectively; Based on the absorption coefficient of each layer of the multilayer material sample and the actual particle beam intensity, an energy optimization allocation model is established to calculate the energy demand of each layer layer by layer. The energy optimization allocation model is: In the formula, I j is the particle beam intensity requirement for injecting into the jth layer of the multilayer material sample, I0 is the initial intensity of the injected particle beam, μ k d k , k are the absorption coefficient, thickness and density of the kth layer of the multilayer material sample, respectively; Identifying the initial phase of the particle beam phased array transmitting unit of the current multilayer material sample through the initial phase recognition model of the phased array transmitting unit; The timing control model of the particle beam phased array emission unit is established through the phase-timing joint control algorithm, specifically: In the formula, I ( t,x) is the required particle beam intensity at time t for a multilayer material sample with damage coordinate x based on the time series corresponding to the set layer scale, Δt i is the transmission delay of the i-th transmitting unit of the phased array, which is used to dynamically adjust the transmission timing. i 、w i and φ i are the particle beam intensity coefficient, particle beam emission frequency, and initial phase identification result of the damage coordinate x of the i-th emission unit of the phased array, respectively, and N is the total number of units in the phased array.

4. The accelerator control method based on AI control according to claim 3, characterized in that: The method of preliminarily determining the material damage position by ultrasonic detection and performing initial phase setting for the phased array particle beam based on the initial phase recognition model specifically includes: Obtain the material type and material properties of the material entity to be tested; Through ultrasonic testing, the damaged area and damage type of the material to be tested are preliminarily detected and quickly calibrated; The material type, material properties, damage type and damage area coordinate data of the material entity to be detected are processed and input into the initial phase recognition model to obtain the initial phase of the particle beam emission phased array emission unit of the material entity to be detected.

5. The accelerator control method based on AI control according to claim 4, characterized in that: The controlling of the emission timing of the phased array particle beam by the timing control model based on the material properties of the detected material specifically includes: Layer the material entity to be tested and test the physical properties of each layer of the material entity; Calculate the absorption coefficient of each layer of the material entity to be tested, and bring the absorption coefficient of each layer into the energy optimization allocation model to calculate the energy demand of each layer of the material entity to be tested; Based on the set material detection layer scale, the energy demand of each layer is mapped to the time series corresponding to the layer scale, and the particle beam intensity and particle beam emission frequency of each emission unit of the current phased array are obtained; The time series energy demand of the current material entity to be detected, as well as the particle beam intensity, particle beam emission frequency and initial phase of each emission unit of the phased array are brought into the timing control model of the particle beam phased array emission unit to calculate the real-time emission delay of each emission unit of the phased array.

6. The accelerator control method based on AI control according to claim 5, characterized in that: The control requirements of the initial phase setting and emission timing of the phased array particle beam and the real-time manipulation of accelerator parameters specifically include: Obtaining the initial phase of the particle beam emission phased array for the current material entity to be detected and the real-time emission delay of each emission unit of the phased array; The accelerator operating parameters are dynamically controlled according to the initial phase and real-time emission delay of each emission unit of the phased array, so that the accelerator meets the emission requirements of the particle beam emission phased array.

7. An accelerator control system based on AI control, used to implement an accelerator control method based on AI control as claimed in any one of claims 1 to 6, characterized in that: include: A phase recognition model building module, wherein the phase recognition model building module detects the surface of the detection sample by emitting a phased array particle beam, marks the coordinates of the damaged area based on the backscattered particle imaging results, records the real-time phase angle of each phased array emitting unit when the particle beam is emitted to the damaged area, and trains an initial phase recognition model of the particle beam phased array emitting unit; A timing control model establishment module, wherein the timing control model establishment module performs particle beam penetration analysis on a multilayer material sample, calculates the absorption coefficient of each layer of the multilayer material sample, establishes an energy optimization allocation model based on the absorption coefficient of each layer of the multilayer material sample and the actual particle beam intensity injected, calculates the energy demand of each layer layer by layer, identifies the initial phase of the particle beam phased array emission unit of the current multilayer material sample through an initial phase recognition model of particle beam emission, and finally establishes a timing control model of the particle beam phased array emission unit through a phase-timing joint control algorithm; An initial phase setting module, which performs preliminary detection and rapid calibration of the damage area and damage type of the material entity to be detected by ultrasonic detection, performs feature processing on the feature data corresponding to the material entity to be detected, and then inputs the feature data into the initial phase recognition model to obtain the initial phase of the particle beam emission phased array emission unit of the material entity to be detected; A timing control module, wherein the timing control module brings the time series energy demand of the current material entity to be detected, the particle beam intensity of each emission unit of the phased array, the particle beam emission frequency and the initial phase of the particle beam emission phased array emission unit into the timing control model of the particle beam phased array emission unit, and calculates the real-time emission delay of each emission unit of the phased array; The accelerator control module dynamically controls the accelerator operating parameters according to the initial phase and real-time transmission delay of each transmitting unit of the phased array.

Images