System and method for online measurement and feedback calibration of electron irradiation dose

Abstract

This invention provides a system and method for online measurement and feedback calibration of electron irradiation dose, relating to the field of automation control technology. It includes: a processing module for converting the raw dose signal into a five-channel digital reading sequence, and performing timestamp alignment and spatial coordinate calibration on the five-channel digital reading sequences to generate a three-dimensional discrete coordinate dataset characterizing the spatial distribution of radiation field intensity; and an extraction module for identifying the global maximum radiation intensity point from the three-dimensional discrete coordinate dataset as the spatial registration origin, extracting two characteristic lines characterizing the dominant radiation scattering direction, and calculating the spatial azimuth angle between the two characteristic lines to determine the main influence area of ​​radiation scattering. This invention achieves accurate processing of the raw dose signal, scientific positioning of the radiation influence area, and construction of a dynamic compensation model through multi-module collaboration. It can output high-confidence dose values ​​and complete feedback calibration based on time-series drift trends, effectively ensuring the accuracy and reliability of online electron irradiation dose measurement.

Description

Technical Field

[0001] This invention relates to the field of automation control technology, and in particular to a system and method for online measurement and feedback calibration of electron irradiation dose. Background Technology

[0002] In production lines that use electron beams to irradiate and modify industrial materials (such as polymer films), multiple radiation detectors are typically installed at key locations around the irradiation area to monitor the process and ensure radiation safety. Real-time dose data generated by these detectors is collected synchronously, forming multiple concurrent data streams, which are then transmitted to a central monitoring computer for processing and recording. Existing monitoring systems in this category typically treat each detector channel as an independent data source. The system's data processing logic often focuses on storing, displaying, or comparing each independent data stream separately based on thresholds. However, the relevant algorithms may be insufficient in supporting in-depth spatial correlation modeling and collaborative fusion analysis of these concurrent data streams based on the fixed layout of multiple detectors in three-dimensional space. When a slow, trending shift occurs in the data stream corresponding to a certain spatial location, existing systems' data processing programs, due to the lack of sufficient multi-source data cross-validation mechanisms based on detector spatial topology, may struggle to reliably identify the nature of this shift at the algorithmic level. Specifically, they may determine whether the shift primarily reflects a physical change in the spatial distribution of the radiation field or is mainly due to a slow degradation of the detector's own response characteristics at that location.

[0003] For example, in a polymer film irradiation production line, a detector deployed on the side of the irradiation chamber to monitor the scattered field may continuously upload dose rate data that exhibits slow unidirectional drift. Because existing data processing systems typically perform relatively isolated historical trend analysis on this single data sequence, they fail to systematically combine its known spatial coordinate information and fully utilize synchronous data from detectors at other locations to construct spatial correlation models for collaborative calculation and cross-validation. Therefore, the existing algorithms may have limited ability to distinguish the source of such data drift. This situation may lead the system to make less accurate assessments of process status or equipment health based on potentially distorted data, which could have a certain impact on the intelligence and reliability of monitoring. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a system and method for online measurement and feedback calibration of electron irradiation dose, so as to realize high-precision online monitoring and automatic calibration of electron irradiation dose.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0006] The first aspect is a system for online measurement and feedback calibration of electron irradiation dose, including:

[0007] The processing module is used to convert the raw dose signal into a five-channel digital reading sequence, and to perform timestamp alignment and spatial coordinate calibration on the five-channel digital reading sequences to generate a three-dimensional discrete coordinate dataset characterizing the spatial distribution of radiation field intensity.

[0008] The extraction module is used to identify the point of global maximum radiation intensity from the three-dimensional discrete coordinate dataset as the origin of spatial registration, and to extract two characteristic lines that characterize the dominant direction of radiation scattering. The spatial azimuth angle between the two characteristic lines is calculated to determine the main influence area of ​​radiation scattering.

[0009] The deployment module is used to determine five influence points in the main influence area of ​​radiation scattering, located at the upper and lower edges of the axial direction of the outer wall of the beam pipe, the left and right sides of the normal direction of the surface of the primary shield, and the central axis of the entrance to the maze. The five influence points constitute a spatial non-coplanar five-vertex topology.

[0010] The module is used to construct a dynamic compensation model based on the geometric configuration of the non-coplanar five-vertex topology and the real-time dose readings corresponding to each vertex, and to obtain the adaptive compensation coefficients for each vertex.

[0011] The calibration module is used to perform spatial consistency calibration and degradation compensation on the five-channel digital reading sequence based on the adaptive compensation coefficient of each vertex, so as to obtain high confidence real-time dose values ​​of each affected point.

[0012] The feedback module is used to determine whether the overall measurement deviation exceeds the preset tolerance based on the population drift trend of the high-confidence real-time dose value in the continuous monitoring time series, so as to complete the calibration.

[0013] Secondly, methods for online measurement and feedback calibration of electron irradiation dose include:

[0014] Step 1: Convert the original dose signal into a five-channel digital reading sequence, and perform timestamp alignment and spatial coordinate calibration on the five-channel digital reading sequences to generate a three-dimensional discrete coordinate dataset characterizing the spatial distribution of radiation field intensity.

[0015] Step 2: Identify the point of global maximum radiation intensity from the three-dimensional discrete coordinate dataset as the origin of spatial registration, and extract two characteristic lines representing the dominant direction of radiation scattering. Calculate the spatial azimuth angle between the two characteristic lines to determine the main influence area of ​​radiation scattering.

[0016] Step 3: Within the main influence area of ​​radiation scattering, determine five influence points located at the upper and lower edges of the axial direction of the outer wall of the beam pipe, the left and right sides of the normal direction of the primary shield surface, and the central axis of the labyrinth entrance. These five influence points constitute a spatial non-coplanar five-vertex topology.

[0017] Step 4: Based on the geometric configuration of the non-coplanar five-vertex topology and the real-time dose readings corresponding to each vertex, construct a dynamic compensation model to obtain the adaptive compensation coefficients for each vertex.

[0018] Step 5: Based on the adaptive compensation coefficient of each vertex, perform spatial consistency calibration and degradation compensation on the five-channel digital reading sequence to obtain the high-confidence real-time dose value of each affected point.

[0019] Step 6: Based on the population drift trend of the high-confidence real-time dose value in the continuous monitoring time series, determine whether the overall measurement deviation exceeds the preset tolerance in order to complete the calibration.

[0020] Thirdly, a computer-readable storage medium storing a program that, when executed by a processor, implements the system.

[0021] The above-described solution of the present invention has at least the following beneficial effects:

[0022] By aligning the timestamps and calibrating the spatial coordinates of five digital reading sequences, a three-dimensional discrete coordinate dataset of radiation field intensity is constructed. Multi-source data fusion analysis is then performed using the geometric correlation of a spatially non-coplanar pentagonal topology, resolving the measurement bias problem caused by the lack of a multi-source data cross-validation mechanism and reducing the impact of single detector data distortion on the overall measurement results. By constructing a spatially non-coplanar pentagonal topology and establishing a dynamic compensation model based on its geometric configuration and real-time dose readings, the system acquires multi-dimensional data cross-validation capabilities. Compared to isolated trend analysis of a single data sequence, this approach effectively identifies the root cause of data drift through spatial correlation logic and dynamic compensation algorithms, clearly distinguishing between physical changes in the spatial distribution of the radiation field and deviations caused by the degradation of the detector's own response characteristics. This avoids interference with process status or equipment health assessments due to data misjudgment. The adaptive compensation coefficients generated by the dynamic compensation model can specifically achieve spatial consistency calibration and detector degradation compensation, effectively offsetting the effects of spatial measurement differences and detector performance degradation after long-term use.

[0023] By monitoring the temporal drift trend of high-confidence real-time dose values, the system automatically determines measurement deviations and performs self-calibration, reducing process risks or safety hazards caused by measurement deviations. By identifying the global maximum radiation intensity point, extracting the dominant scattering direction, and defining the main influence area of ​​radiation scattering, the deployment of the five influence points becomes more targeted, enabling comprehensive monitoring of key areas of the radiation field and avoiding the blind placement of monitoring points. The adaptive adjustment characteristics and feedback calibration mechanism of the dynamic compensation model allow the system to adapt to the dynamic changes in the radiation field and the evolution of equipment operating status in real time, improving the intelligence level of radiation monitoring and reducing the workload of maintenance personnel. High-confidence real-time dose values ​​provide accurate data support for monitoring the irradiation process, helping operators accurately grasp the process status and ensuring the consistency of irradiation modification effects. Simultaneously, the system's comprehensive monitoring and real-time deviation calibration of key areas of the radiation field can provide timely and accurate feedback on radiation leakage risks, providing reliable protection for radiation safety. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of an online electron irradiation dose measurement and feedback calibration system provided in an embodiment of the present invention.

[0025] Figure 2 This is a schematic flowchart of the method for online measurement and feedback calibration of electron irradiation dose provided in an embodiment of the present invention. Detailed Implementation

[0026] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0027] like Figure 1 As shown, embodiments of the present invention propose a system for online measurement and feedback calibration of electron irradiation dose, comprising:

[0028] The processing module is used to convert the raw dose signal into a five-channel digital reading sequence, and to perform timestamp alignment and spatial coordinate calibration on the five-channel digital reading sequences to generate a three-dimensional discrete coordinate dataset characterizing the spatial distribution of radiation field intensity.

[0029] The extraction module is used to identify the point of global maximum radiation intensity from the three-dimensional discrete coordinate dataset as the origin of spatial registration, and to extract two characteristic lines that characterize the dominant direction of radiation scattering. The spatial azimuth angle between the two characteristic lines is calculated to determine the main influence area of ​​radiation scattering.

[0030] The deployment module is used to determine five influence points in the main influence area of ​​radiation scattering, located at the upper and lower edges of the axial direction of the outer wall of the beam pipe, the left and right sides of the normal direction of the surface of the primary shield, and the central axis of the entrance to the maze. The five influence points constitute a spatial non-coplanar five-vertex topology.

[0031] The module is used to construct a dynamic compensation model based on the geometric configuration of the non-coplanar five-vertex topology and the real-time dose readings corresponding to each vertex, and to obtain the adaptive compensation coefficients for each vertex.

[0032] The calibration module is used to perform spatial consistency calibration and degradation compensation on the five-channel digital reading sequence based on the adaptive compensation coefficient of each vertex, so as to obtain high confidence real-time dose values ​​of each affected point.

[0033] The feedback module is used to determine whether the overall measurement deviation exceeds the preset tolerance based on the population drift trend of the high-confidence real-time dose value in the continuous monitoring time series, so as to complete the calibration.

[0034] In this embodiment of the invention, a three-dimensional discrete coordinate dataset of radiation field intensity is constructed by aligning the timestamps and calibrating the spatial coordinates of five digital reading sequences. Multi-source data collaborative fusion analysis is then performed by combining the geometric correlation of a spatially non-coplanar pentagonal topology. This solves the measurement deviation problem caused by the lack of a multi-source data cross-validation mechanism and reduces the impact of single detector data distortion on the overall measurement results. By constructing a spatially non-coplanar pentagonal topology and establishing a dynamic compensation model based on the geometric configuration of this topology and real-time dose readings, the system possesses multi-dimensional data cross-validation capabilities. Compared to isolated trend analysis of a single data sequence, this approach effectively identifies the root cause of data drift through spatial correlation logic and dynamic compensation algorithms. It clearly distinguishes between physical changes in the spatial distribution of the radiation field and deviations caused by the degradation of the detector's own response characteristics, avoiding interference with process status or equipment health assessments due to data misjudgment. The adaptive compensation coefficients generated by the dynamic compensation model can specifically achieve spatial consistency calibration and detector degradation compensation, effectively offsetting the effects of spatial measurement differences and detector performance degradation after long-term use.

[0035] By monitoring the temporal drift trend of high-confidence real-time dose values, the system automatically determines measurement deviations and performs self-calibration, reducing process risks or safety hazards caused by measurement deviations. By identifying the global maximum radiation intensity point, extracting the dominant scattering direction, and defining the main influence area of ​​radiation scattering, the deployment of the five influence points becomes more targeted, enabling comprehensive monitoring of key areas of the radiation field and avoiding the blind placement of monitoring points. The adaptive adjustment characteristics and feedback calibration mechanism of the dynamic compensation model allow the system to adapt to the dynamic changes in the radiation field and the evolution of equipment operating status in real time, improving the intelligence level of radiation monitoring and reducing the workload of maintenance personnel. High-confidence real-time dose values ​​provide accurate data support for monitoring the irradiation process, helping operators accurately grasp the process status and ensuring the consistency of irradiation modification effects. Simultaneously, the system's comprehensive monitoring and real-time deviation calibration of key areas of the radiation field can provide timely and accurate feedback on radiation leakage risks, providing reliable protection for radiation safety.

[0036] In a preferred embodiment of the present invention, the original dose signal is converted into a five-channel digital reading sequence, and the five-channel digital reading sequences are timestamped and spatially calibrated to generate a three-dimensional discrete coordinate dataset characterizing the spatial distribution of radiation field intensity, including:

[0037] Raw dose signals were collected from ionization chamber detectors located at the upper and lower sides of the outer wall of the beam pipe, the left and right sides of the primary shielding surface, and the central axis of the irradiation chamber labyrinth entrance. Specifically, this involved: precisely fixing five ionization chamber detectors to their designated target positions; the detectors on the upper and lower sides of the beam pipe outer wall were tightly fitted against the pipe wall using bolts and anti-vibration pads, axially fixed to ensure a seamless fit with the pipe surface; the detectors on the left and right sides of the primary shielding surface were installed perpendicular to the shielding surface using brackets, which were welded to the shielding to prevent loosening; the detector at the central axis of the irradiation chamber labyrinth entrance was fixed along the central axis using adjustable mounting brackets, and calibrated using a laser positioning device after installation to ensure that the deviation between the actual installation position and the design coordinates of each detector was within the allowable range and that the attitude remained stable without tilting; after installation, the detector preheating program was initiated. The detector is continuously powered on for more than 30 minutes, during which the output baseline signal of the detector is monitored in real time. When the fluctuation of the baseline signal does not exceed ±1% within 30 consecutive seconds, the detector is considered to have reached a stable operating state. Subsequently, each detector is connected to the corresponding channel of the data acquisition unit one by one through a dedicated transmission cable with a metal shielding layer. During the cable laying process, strong electromagnetic interference sources are avoided, and shielded connectors are used at the joints to further isolate external electromagnetic interference. At the same time, a unified level trigger signal is output by the data acquisition unit to control the five detectors to start signal acquisition at the same time. During the acquisition, the temperature and humidity of the detector's working area are monitored in real time through the environmental monitoring module. The temperature is controlled at 20℃±2℃, and the humidity is controlled at 40%RH to 60%RH. If it exceeds this range, the temperature and humidity control equipment is automatically activated to adjust, ensuring that the detector works under stable environmental conditions and that the acquired raw dose signal can truly and accurately reflect the actual state of the radiation field.

[0038] The raw dose signal output from each ionization chamber detector is converted from analog to digital to obtain the corresponding digital reading, forming a five-channel digital reading sequence. Each ionization chamber detector is assigned a three-dimensional spatial coordinate measured from its fixed installation position. Specifically, this involves: first, low-pass filtering preprocessing of the raw dose signal output from each ionization chamber detector, with the filter cutoff frequency set to 10Hz. This value matches the low-frequency characteristics of the radiation dose signal, effectively filtering out high-frequency noise from environmental electromagnetic interference while retaining the effective signal components; then, inputting the preprocessed analog signal into a high-precision analog-to-digital converter (ADC). The sampling rate of this converter is set according to the Nyquist sampling theorem to be more than twice the highest frequency of the original signal, and combined with the filter cutoff frequency, the sampling rate is determined to be no less than 20Hz to ensure complete reconstruction of the analog signal; during the analog-to-digital conversion process, the voltage range of the analog signal is first defined as -10V to +10V (this range covers the output voltage range of the ionization chamber detector under conventional irradiation scenarios), and then this range is evenly divided... Divided into 65,536 quantization levels (corresponding to 16-bit analog-to-digital conversion accuracy), the analog signal is converted into a corresponding digital reading by calculating the digital reading as: (preprocessed analog signal voltage value - lower limit voltage value) ÷ (upper limit voltage value - lower limit voltage value) × total number of quantization levels. Five detectors generate five consecutive digital reading sequences to ensure that the digital signal can accurately map the dose information of the original analog signal. After completing the analog-to-digital conversion, a spatial rectangular coordinate system is established with a fixed reference point inside the irradiation chamber (such as a fixed marker point at the edge of the labyrinth entrance) as the origin. The x-axis extends along the beam pipe axis, the y-axis is perpendicular to the beam pipe axis and parallel to the ground, and the z-axis is perpendicular to the ground and upward. A laser rangefinder is used to measure the x-axis, y-axis, and z-axis distances of each detector's installation position relative to the origin multiple times. The arithmetic mean of all measurement results is taken as the coordinate component of each axis. The three coordinate components are combined to form the spatial three-dimensional coordinates of each detector, and a fixed association is established between the coordinates and the corresponding detector.

[0039] A unified clock source is used to synchronize the five digital reading sequences, ensuring that the sampling points of each sequence have the same timestamp. At each synchronized sampling moment, the digital reading of each ionization chamber detector is bound to the corresponding three-dimensional spatial coordinates of the ionization chamber detector, forming a three-dimensional discrete coordinate dataset. Specifically, this involves: using a high-precision crystal oscillator as the unified clock source to provide a unified time reference for the sampling process of the five digital reading sequences; connecting the analog-to-digital converter (ADC) unit corresponding to each detector to this unified clock source; first calculating the inherent deviation of the timing module of each ADC unit (i.e., the difference between the timing value of the ADC unit itself and the timing value of the unified clock source); then correcting this inherent deviation to zero by adjusting the clock frequency of the timing module, ensuring that the timing of all ADC units is completely synchronized with the unified clock source; and setting a fixed sampling interval to unify... The clock source generates a synchronization pulse signal every sampling interval, triggering all analog-to-digital conversion units to simultaneously acquire a digital reading. The timestamp of each sampling point directly adopts the time value corresponding to the unified clock source, so that all sampling points of the five digital reading sequences have a consistent timestamp identifier. For each sampling time, each digital reading in the five digital reading sequences is associated and bound with the spatial three-dimensional coordinates pre-assigned to that detector, forming a complete data item containing the digital reading, x-axis coordinate, y-axis coordinate, z-axis coordinate, and timestamp. The complete data items corresponding to all sampling times are summarized in chronological order of timestamps, and finally a three-dimensional discrete coordinate dataset representing the spatial distribution of radiation field intensity is constructed. This dataset can synchronously reflect the radiation dose information of different spatial locations at the same time point.

[0040] This embodiment, through standardized detector installation, environmental control, and synchronous acquisition, reduces the distortion of raw signals caused by installation deviations, environmental interference, and asynchronous acquisition, ensuring that the acquired raw dose signals accurately reflect the actual state of the radiation field. Standardized preprocessing and analog-to-digital conversion processes guarantee the accuracy of digital readings and prevent information loss during signal conversion. Time synchronization processing using a unified clock source ensures that the sampling points of the five digital reading sequences have a unified time reference, eliminating the impact of time differences on the collaborative analysis of multi-source data. Precise assignment and data binding of spatial three-dimensional coordinates associate dispersed digital readings with specific spatial locations, forming a three-dimensional discrete coordinate dataset with both temporal and spatial attributes, solving the problems of isolated data from each channel and lack of spatiotemporal correlation. The three-dimensional discrete coordinate dataset integrates time, space, and dose information, fully presenting the distribution characteristics of the radiation field intensity in three-dimensional space and its changing patterns over time, avoiding analytical biases caused by missing data dimensions or spatiotemporal inconsistencies.

[0041] In a preferred embodiment of the present invention, the point of global maximum radiation intensity is identified from a three-dimensional discrete coordinate dataset as the origin for spatial registration, and two characteristic lines characterizing the dominant direction of radiation scattering are extracted. The spatial azimuth angle between the two characteristic lines is calculated to determine the main influence area of ​​radiation scattering, including:

[0042] The process involves traversing all data points in a 3D discrete coordinate dataset, comparing the dose intensity values ​​of each data point, and defining the spatial 3D coordinates corresponding to the data point with the highest dose intensity value as the spatial registration origin. A neighborhood radius is then set around this spatial registration origin, and all data points in the 3D discrete coordinate dataset whose spatial distance from the spatial registration origin is less than the neighborhood radius are selected to form a subset of neighborhood data points. Specifically, this includes: first, starting the data traversal program and sequentially reading each data point in the 3D discrete coordinate dataset. Each data point contains a dose intensity value and its corresponding spatial 3D coordinates. During the traversal, the maximum dose intensity value among the currently read data points is recorded in real time, and its corresponding spatial 3D coordinates are synchronously associated. After traversing all data points, the spatial 3D coordinates corresponding to the finally recorded maximum dose intensity value are formally defined as the spatial registration origin. Subsequently, a neighborhood radius of 5 cm is set, the determination of which is based on the physical properties of the electron irradiation field. Based on the attenuation characteristics, the intensity of the radiation field generated by electron irradiation decreases gradually from the core region outwards. Considering the typical propagation range of beam scattering in industrial irradiation scenarios, the beam scattering of industrial electron accelerators varies due to differences in equipment type and energy parameters. The core scattering energy is concentrated in the region within 10 cm of the radiation center axis, and the scattering dose at greater distances has significantly attenuated and its correlation has decreased. To ensure that the neighborhood contains a sufficient number of effective data points to support subsequent principal component analysis (avoiding distortion of analysis results due to insufficient data points), and to exclude interference data that is too far away or has a weak correlation with the core radiation field, the radius of the neighborhood was determined after comprehensive verification. For each data point in the three-dimensional discrete coordinate dataset, the three-dimensional Euclidean distance between it and the spatial registration origin was calculated. All data points with a spatial distance of less than 5 cm were selected and summarized to form a subset of neighborhood data points, ensuring that this subset can focus on the relevant data of the core region of the radiation field intensity.

[0043] Principal component analysis is performed on the spatial three-dimensional coordinates of all data points in the neighborhood data point subset to calculate the directions of the first and second principal components. The directions of the first and second principal components are then fitted to the first characteristic line through the spatial registration origin, and the directions of the second principal components are fitted to the second characteristic line through the spatial registration origin. Specifically, this includes: firstly, preprocessing the neighborhood data point subset by calculating the mean of the x-axis, y-axis, and z-axis coordinates of all data points within the subset; for each data point, subtracting the x-axis mean from its x-axis coordinate yields the x-axis deviation, and subtracting the y-axis mean from its y-axis coordinate yields the y-axis deviation. The z-axis deviation value is obtained by subtracting the z-axis mean from the z-axis coordinate, forming a three-dimensional deviation vector for each data point. A covariance matrix is ​​calculated based on the three-dimensional deviation vectors of all data points. The diagonal elements of the covariance matrix represent the variances of the x-axis, y-axis, and z-axis deviations (variance is calculated by summing the squares of the deviations for all data points divided by the total number of data points). The off-diagonal elements represent the covariances of the x-axis and y-axis deviations, the x-axis and z-axis deviations, and the y-axis and z-axis deviations (covariance is calculated by summing the products of the deviations for each of the two dimensions divided by the total number of data points).

[0044] The calculated 3×3 covariance matrix is ​​subjected to eigenvalue decomposition. The specific process is as follows: First, it is determined that the covariance matrix is ​​a third-order symmetric matrix, whose elements consist of the variances of each coordinate dimension and the covariances between dimensions. Based on the principles of linear algebra, a characteristic equation is constructed. The third-order identity matrix involved is a 3×3 matrix with all diagonal elements being 1 and all off-diagonal elements being 0. Specifically, the elements in the first row and first column, the second row and second column, and the third row and third column are 1, while the elements in the remaining positions (first row and second column, first row and third column, second row and first column, second row and third column, third row and first column, third row and second column) are 0. The characteristic equation is constructed by calculating the difference matrix between the covariance matrix and λ (eigenvalues) multiplied by the third-order identity matrix, then calculating the determinant of this difference matrix and setting it equal to zero. Through algebraic expansion of the determinant, the equation is obtained regarding λ. The equation is a cubic polynomial in one variable. Numerical methods (such as the Jacobi method) are used to solve the cubic equation, yielding three eigenvalues ​​(λ1, λ2, λ3). For each eigenvalue, it is substituted into a homogeneous linear system of equations (the covariance matrix minus the eigenvalue multiplied by the third-order identity matrix, multiplied by the three-dimensional column vectors, equals the zero vector). Solving this system yields a non-zero solution vector, which is the eigenvector corresponding to the eigenvalue. This results in three eigenvalues ​​and their unique corresponding eigenvectors. The three eigenvalues ​​are sorted in descending order, with the eigenvector corresponding to the largest eigenvalue defined as the first principal component direction, and the eigenvector corresponding to the second largest eigenvalue defined as the second principal component direction. Both eigenvectors are normalized (each component is divided by the vector's magnitude to make the magnitude equal to 1) to ensure the uniqueness of the direction representation.

[0045] Starting from the three-dimensional coordinates of the spatial registration origin, the extension direction of the straight line is determined based on the normalized eigenvector of the first principal component direction. The coordinates of the spatial registration origin are then multiplied by k (where k is any non-zero real number within the range of -10 cm to 10 cm, which fully covers the spatial scale of the main influence area of ​​radiation scattering, ensuring the straight line completely represents the principal component direction). This yields the coordinates of a series of continuous points on the straight line. These points are then fitted to form the first characteristic straight line, whose direction is completely consistent with the first principal component direction. Similarly, starting from the three-dimensional coordinates of the spatial registration origin, the coordinates of the spatial registration origin are then multiplied by k (where k remains within the range of -10 cm to 10 cm) based on the normalized eigenvector of the second principal component direction. This process is repeated to form the second characteristic straight line. Both characteristic straight lines share the spatial registration origin as their common starting point, and their extension directions are consistent with the corresponding principal component directions.

[0046] The angle between the direction vectors of the first and second characteristic lines is calculated to obtain the spatial azimuth angle. Using the spatial registration origin as the vertex and the first and second characteristic lines as boundaries, the conical spatial region formed by sweeping the spatial azimuth angle around the spatial registration origin in three-dimensional space is determined as the main influence area for radiation scattering. Specifically, this involves: firstly, extracting the direction vectors of the first and second characteristic lines, which are the obtained first and second principal component eigenvectors; then calculating the spatial azimuth angle between the two direction vectors, i.e., first calculating the dot product of the two direction vectors, and then calculating... Calculate the magnitudes of the two direction vectors, then divide the dot product by the product of the two magnitudes to obtain the cosine of the included angle. Finally, perform an inverse cosine operation on the cosine to obtain the spatial azimuth angle between the two direction vectors. Using the spatial registration origin as the vertex of the cone-shaped region, and the first and second characteristic lines as the two generatrices (boundaries) of the cone, sweep the spatial azimuth angle around the spatial registration origin in three-dimensional space in a 360-degree motion. During the sweeping process, the two generatrices always maintain a fixed spatial azimuth angle. The resulting closed cone-shaped spatial region is the main influence area of ​​radiation scattering, which can accurately cover the core area where radiation scattering energy is concentrated.

[0047] This embodiment accurately locates the core position of the radiation field by traversing the dataset and selecting the point with the maximum dose intensity as the spatial registration origin. Combined with the neighborhood data point subset formed by neighborhood radius selection, it can focus on the effective data in the core area and avoid interference from irrelevant data. By extracting the first and second principal component directions from the neighborhood data through principal component analysis, it can objectively reflect the main propagation direction of radiation scattering. Combined with the calculation of the angle between the direction vectors and the cone-shaped region formed by three-dimensional sweep, it can define the main influence area of ​​radiation scattering, solving the problem of not being able to clearly define the influence range of the scattering core. The determination of the main influence area of ​​radiation scattering avoids the blindness of point layout, and the spatial characteristics of this area provide important spatial correlation basis for the construction of dynamic compensation model, which helps to improve the accuracy and pertinence of multi-source data fusion analysis.

[0048] In a preferred embodiment of the present invention, within the main influence region of radiation scattering, five influence points are identified located at the upper and lower edges of the axial direction of the outer wall of the beam pipe, the left and right sides of the surface normal of the primary shield, and the central axis of the labyrinth entrance. These five influence points constitute a spatially non-coplanar five-vertex topology, including:

[0049] Within the conical spatial region defined by the main radiation scattering influence zone, based on the known equation of the cylindrical surface of the beam pipe's outer wall, the intersection line between the equation and the main radiation scattering influence zone is calculated. The highest and lowest points on this intersection line are then taken as the upper and lower influence points along the axial direction of the beam pipe's outer wall. Specifically, this involves: first, clarifying the known equation of the cylindrical surface of the beam pipe's outer wall, which is established based on the structural parameters of the beam pipe (the axis of the cylinder is aligned with the beam pipe's axial direction, i.e., parallel to the x-axis, and the radius of the cylinder is the same as the radius of the pipe's outer wall). The specific formula is as follows: in, For the axis of the beam channel axis, Given the fixed coordinates of the axes (known), R is the radius of the outer wall of the beam pipe (known), and considering the defined main influence area of ​​radiation scattering (conical spatial region), its boundary equation is derived from the spatial registration origin (vertex). ), two characteristic straight lines (genealogies) and the included angle of spatial orientation Confirmed, the specific formula is as follows: ,in, For any point With vertex The connecting vector, Let be the direction vector of the first characteristic line (known). The physical meaning of this formula is that, with the vertex as the starting point, the angle between the line connecting any point to the vertex and the first characteristic line does not exceed the spatial azimuth angle. Simultaneously, the equations of the cylindrical surface of the outer wall of the beam pipe and the boundary equations of the main influence zone of radiation scattering are combined to construct a system of equations. Solving this system of equations using algebraic elimination eliminates one coordinate variable (such as the x-axis), resulting in a bivariate equation containing the remaining two coordinate variables (y-axis and z-axis). This equation is the projection equation of the intersection line onto the yz plane. Combined with the cylindrical surface equation, the spatial trajectory of the intersection line can be completely characterized. (A spatial curve on a cylindrical surface); determine the upper and lower edge influence points, i.e., traverse the spatial three-dimensional coordinates of all points on the intersection line, and extract the z-axis coordinates of each point (since the z-axis is perpendicular to the ground and pointing upwards, the z-axis coordinates directly correspond to the spatial height); by comparing the values ​​of all z-axis coordinates, select the data point with the largest z-axis coordinate, and define it as the upper edge influence point of the beam pipe outer wall axis; select the data point with the smallest z-axis coordinate, and define it as the lower edge influence point of the beam pipe outer wall axis. Both points are located on the intersection line and cover the key scattering influence range of the pipe outer wall axis.

[0050] Based on the known plane equation of the outer surface of the primary shield, the intersection of the plane equation of the outer surface of the primary shield and the main influence area of ​​radiation scattering is calculated. Within this intersection area, the leftmost and rightmost points are selected as the left and right influence points of the primary shield surface normal. Specifically, this includes: the known plane equation of the outer surface of the primary shield, established based on the installation position and orientation of the shield (the plane is perpendicular to the y-axis, which is perpendicular to the beam axis and parallel to the ground; this plane is the outer surface of the shield). The specific formula is as follows: ,in, Given the fixed y-axis coordinates of the outer surface of the primary shielding body, a system of equations is constructed by simultaneously solving the plane equations of the outer surface of the primary shielding body and the boundary equations of the main radiation scattering influence zone. When solving this system of equations, the plane equations are substituted into the boundary equations of the conical region to obtain the set of points in the plane that satisfy the conical region constraints. The closed region formed by this set of points is the intersection (an elliptical or circular region in the plane). The left and right influence points are determined by traversing the spatial three-dimensional coordinates of all points in the intersection region and extracting the y-axis coordinates of each point (since the y-axis is perpendicular to the beam axis and parallel to the ground, the y-axis coordinates directly correspond to the left and right positions in the plane). By comparing the values ​​of all y-axis coordinates, the data point with the smallest y-axis coordinate is selected and defined as the left influence point of the primary shielding body surface normal; the data point with the largest y-axis coordinate is selected and defined as the right influence point of the primary shielding body surface normal. Both points are located in the intersection region and cover the key scattering influence range of the shielding body surface normal.

[0051] Based on the known spatial equation of the maze entrance centerline, the intersection point of the spatial equation of the maze entrance centerline and the boundary of the main influence zone of radiation scattering is calculated as the influence point of the maze entrance centerline. Specifically, this includes: the known spatial equation of the maze entrance centerline, established based on the design parameters of the maze (the straight line is along the direction of the maze entrance centerline, i.e., the reference direction for detector installation), specifically in parametric form. ,in, The starting coordinates of the center line of the maze entrance are fixed (known). The direction vector of the centerline is known. Let be any real number parameter. Simultaneously solve the equations of the spatial straight line at the center line of the maze entrance and the boundary equations of the main influence zone of radiation scattering. The specific boundary equations are as follows: The equation of the straight line is expressed in parametric form (e.g., taking a fixed point on the straight line as the starting point and taking the parameter t along the direction vector of the axis). Substituting this into the boundary equation of the conical region (the angle between the line connecting any point to the vertex and the first characteristic line is equal to the spatial orientation angle), we obtain a univariate equation about the parameter t. Solving this equation yields the effective value of the parameter t (it is necessary to verify whether the corresponding point is within the main influence area of ​​radiation scattering and to exclude invalid solutions that are outside the conical range). Substituting the effective value into the straight line equation, we calculate the corresponding three-dimensional spatial coordinates. These coordinates are the intersection of the center line of the maze entrance and the boundary of the main influence area of ​​radiation scattering, and are defined as the influence point of the central axis of the maze entrance.

[0052] The spatial three-dimensional coordinates of the points influencing the upper edge of the beam duct outer wall, the lower edge of the beam duct outer wall, the left side of the primary shield surface, the right side of the primary shield surface, and the central axis of the maze entrance are not on the same plane, and are connected sequentially to form a spatial polygon containing five vertices. This spatial polygon is a non-coplanar five-vertex topological structure. Specifically, this includes: firstly, performing a non-coplanarity verification of the five influencing points. The specific process is as follows: From the influence points along the upper edge of the outer wall of the beam duct, the lower edge of the outer wall of the beam duct, the left side of the primary shield surface, and the right side of the primary shield surface, select three non-collinear points as references: the influence point along the upper edge of the outer wall of the beam duct (denoted as point P1), the influence point along the lower edge of the outer wall of the beam duct (denoted as point P2), and the influence point on the left side of the primary shield surface (denoted as point P3). Based on the principle of constructing the three-point plane equation, first calculate vectors P1P2 and P1P3, and then obtain the result through the cross product of the two vectors. The normal vector of the plane (normal vector components A, B, and C correspond to the x, y, and z-axis components of the cross product result, respectively); substituting the three-dimensional coordinates of the reference point P1 into the plane equation Ax + By + Cz + D = 0, and obtaining the constant term D through rearranging terms, the plane equation uniquely determined by the three points P1, P2, and P3 is finally determined; extract the three-dimensional coordinates of the fifth point, i.e., the point affected by the central axis of the maze entrance (denoted as point P5), and substitute its x, y, and z-axis coordinates one by one into the previously determined plane equation to calculate the numerical result of Ax + By + Cz + D; since the plane The equation Ax+By+Cz+D=0 is a necessary and sufficient condition that the coordinates of all points located on this plane must satisfy. If the calculation result is not equal to 0, it indicates that the coordinates of point P5 do not satisfy the plane equation, that is, point P5 is not on the plane determined by points P1, P2, and P3. Combined with the fact that the remaining influence point on the right side of the primary shield surface (denoted as point P4) has been confirmed to be coplanar with P1, P2, and P3 through geometric position analysis (because they are all located in the near-field region formed by the beam pipe and the shield), it can be inferred that the five influence points do not have a common bearing plane, that is, they satisfy the non-coplanar characteristic.

[0053] Starting from the influence point along the upper edge of the outer wall of the beam duct, and following the principle of being close to the radiation propagation path and covering the critical structural boundary, the influence points along the lower edge of the outer wall of the beam duct, the influence points on the left side of the primary shield surface, the influence points on the right side of the primary shield surface, and the influence points along the central axis of the maze entrance are connected in sequence. Finally, the connection is traced back to the starting point, forming a closed spatial polygon. Because the five vertices of this polygon are not coplanar, it constitutes a spatial non-coplanar five-vertex topological structure. The distribution of its vertices corresponds to the three core scattering propagation directions: the beam duct axis, the primary shield normal, and the maze entrance axis, which can completely encompass the critical spatial range of the main influence area of ​​radiation scattering.

[0054] This embodiment calculates the intersection lines, intersections, and intersection points by solving simultaneous geometric equations, ensuring that all five influencing points are located within the main radiation scattering influence area and distributed at key positions in the core structures such as the beam pipe, primary shield, and labyrinth entrance. This avoids the randomness of point placement and achieves precise coverage of the scattering influence area. The non-coplanar five-vertex topological structure fully characterizes the spatial morphology of the main radiation scattering influence area, providing a clear spatial correlation basis for the subsequent dynamic compensation model. This enables multi-source detection data to achieve precise matching and fusion based on the topological structure, improving the logic and effectiveness of data processing. The determination of all influencing points is achieved by solving simultaneous geometric equations for cylindrical, planar, straight, and conical regions, avoiding errors from subjective human judgment and ensuring the accuracy and consistency of point coordinates. The five points correspond to three key directions: the beam pipe axis, the primary shield normal, and the labyrinth entrance axis, comprehensively covering the main propagation paths and influence areas of radiation scattering in industrial irradiation scenarios, enhancing the integrity of the overall technical solution.

[0055] In a preferred embodiment of the present invention, a dynamic compensation model is constructed based on the geometric configuration of the spatially non-coplanar five-vertex topology and the real-time dose readings corresponding to each vertex, to obtain the adaptive compensation coefficients for each vertex, including:

[0056] Based on the spatial distance and relative azimuth between any two vertices in the non-coplanar five-vertex topology of the spatial non-coplanar five-vertex topology, including the influence points along the upper and lower edges of the beam duct outer wall, the left and right edges of the primary shield surface, and the central axis of the maze entrance, a geometric weight matrix describing the spatial correlation of the five vertices is calculated. Specifically, this includes: first, defining the five vertices in the non-coplanar five-vertex topology, and defining them in a fixed order as vertex 1 (influence point along the upper edge of the beam duct outer wall), vertex 2 (influence point along the lower edge of the beam duct outer wall), vertex 3, vertex 4, vertex 5, vertex 6, vertex 7, vertex 8, vertex 9, vertex 10, vertex 11, vertex 12, vertex 13, vertex 14, vertex 15, vertex 16, vertex 17, vertex 18, vertex 1 ... (Points affecting the left side of the primary shielding surface), Vertex 4 (points affecting the right side of the primary shielding surface), Vertex 5 (points affecting the central axis of the maze entrance). The geometric weight matrix is ​​a 5x5 square matrix. The matrix arrangement rule is that the row index and column index correspond to the fixed order of the above five vertices, that is, the element in the i-th row and j-th column uniquely corresponds to the spatial correlation weight between the i-th vertex and the j-th vertex (i and j are both integers from 1 to 5), ensuring the uniqueness of the correspondence between matrix elements and vertices; the spatial distance and relative azimuth angle between any two vertices are calculated as follows: for any two vertices (the i-th... Extract the three-dimensional coordinates (x-axis, y-axis, z-axis coordinates) of each vertex (i-th vertex and j-th vertex). Calculate the differences between the x-axis coordinates of the i-th vertex and the j-th vertex, the differences between the y-axis coordinates of the i-th vertex and the j-th vertex, and the differences between the z-axis coordinates of the i-th vertex and the j-th vertex. Square each of these three coordinate differences, add the three squared results, and finally take the square root of the sum to obtain the spatial straight-line distance between the two vertices. The relative azimuth angle is divided into horizontal azimuth angle and vertical azimuth angle. The horizontal azimuth angle is the projection of the line connecting the two vertices onto the xy-plane and the x-axis (beam). The horizontal azimuth angle is the angle between the line connecting the two vertices and the xy-plane. It is calculated by dividing the difference in the y-axis coordinates of the two vertices by the difference in the x-axis coordinates to obtain the tangent value. Then, the arctangent operation is performed on this tangent value to obtain the horizontal azimuth angle (ranging from -180 degrees to 180 degrees). The vertical azimuth angle is the angle between the line connecting the two vertices and the xy-plane. It is calculated by first calculating the horizontal distance between the two vertices (the square root of the sum of the squares of the differences in the x-axis and y-axis coordinates), then dividing the difference in the z-axis coordinates by this horizontal distance to obtain the tangent value. The arctangent operation is then performed on this tangent value to obtain the vertical azimuth angle (ranging from -90 degrees to 90 degrees).

[0057] The geometric weight matrix is ​​constructed according to the following weight calculation rule: the closer the spatial distance and the closer the relative azimuth angle is to the main radiation propagation direction (beam axis or dominant scattering direction), the greater the weight. The specific calculation process is as follows: first, calculate the reciprocal of the spatial distance (the closer the distance, the larger the reciprocal, and the higher the basic weight value); then calculate the cosine values ​​of the horizontal and vertical azimuth angles (the closer the angle is to 0 degrees, the closer the cosine value is to 1, indicating that the direction is more consistent with the dominant radiation propagation direction, and the higher the weight proportion); multiply the reciprocal of the distance by the cosine values ​​of the two azimuth angles to obtain the original weight value corresponding to the set of vertices; sum all the original weight values ​​in each row of the matrix, and then divide each original weight value in that row by the total weight of that row to obtain the normalized weight value (the sum of the weight values ​​in each row is 1 to ensure the normalization and additivity of the weights); after calculating the elements of all rows and columns according to the above rules, a 5×5 geometric weight matrix is ​​finally formed.

[0058] Real-time dose readings are acquired at the following points: the upper edge of the outer wall of the beam duct, the lower edge of the outer wall of the beam duct, the left side of the primary shielding surface, the right side of the primary shielding surface, and the central axis of the labyrinth entrance. These points form a real-time dose reading vector. Specifically, this involves deploying dose detection devices at each of the five points, ensuring that the detection center of the device is precisely aligned with the corresponding three-dimensional spatial coordinates (deviation not exceeding a preset spatial error threshold), and that the detection range of the device completely covers the radiation scattering influence area of ​​that point. Real-time dose readings from the five detection devices are collected synchronously, with consistent timestamps, and the error controlled within a preset small time range (10 milliseconds) to avoid data deviations caused by time asynchrony. The five real-time dose readings are then arranged in a fixed order according to a defined vertex (vertex 1 to vertex 5) to form a five-dimensional real-time dose reading vector, where each element corresponds to the real-time dose measurement value of a point.

[0059] The geometric weight matrix is ​​used as a spatial correlation constraint. Real-time dose reading vectors and historical cumulative dose data are input into a pre-defined dynamic compensation model. In each iteration of the dynamic compensation model, the compensation coefficient vector output from the previous iteration is used in conjunction with the geometric weight matrix to calculate the predicted dose reading vectors for the five influencing points. Specifically, the pre-defined dynamic compensation model is built on an adaptive iteration principle, primarily used to correct spatial interference errors and equipment system errors in dose readings. Specifically, the obtained geometric weight matrix is ​​used as a spatial correlation constraint (constraining the mutual influence of dose data from different points), the obtained real-time dose reading vector is used as the current measurement input, and historical cumulative dose data is used as a trend reference input, all of which are input into the dynamic compensation model. The historical cumulative dose data is a set of dose readings for each point within a pre-defined time period (e.g., the most recent month), preprocessed using a moving average method (with a moving window length of 24 hours), and the dose change trend value for each point (i.e., the average rate of change of dose readings within that time period, in millisieverts / hour / day) is extracted. The core of the model's iterative calculation is the compensation coefficient vector. The dynamic update of the model involves obtaining the initial compensation coefficient vector as follows: During the initial model run, it is determined based on historical standard calibration data for each location. Specifically, the dose readings for each location during the preset standard calibration period (e.g., 24 hours after calibration) are extracted, the average dose value (standard reference value) for that period is calculated, and the average value of the actual measured readings (actual measured value) within that period is calculated. The ratio of the standard reference value to the actual measured value is used as the initial compensation coefficient for that location, arranged in a fixed order of vertices to form the initial five-dimensional compensation coefficient vector. In each iteration of the model, the five-dimensional compensation coefficient vector output from the previous iteration (the initial compensation coefficient vector is used in the first iteration) is called. This vector is then multiplied by the geometric weight matrix (specifically, each element of the compensation coefficient vector is multiplied by each element of the corresponding row of the weight matrix, and the product results of each row are summed to obtain a five-dimensional intermediate vector). Each element of the intermediate vector is then added to the historical dose change trend value of the corresponding location to obtain the dose reading prediction vector for the five affected locations in the current iteration. Each element of the prediction vector corresponds to the dose prediction value for one location.

[0060] The predicted dose reading vector is compared with the real-time dose reading vector to calculate the predicted residual vector. Based on historical cumulative dose data, a degradation factor reflecting the current response capability of each detector is calculated. Specifically, this involves subtracting the predicted dose reading vector from the real-time dose reading vector element-wise, i.e., subtracting the first element of the predicted vector from the first element of the real-time dose reading vector, subtracting the second element from the second element of the predicted vector, and so on, to obtain five residual values. These residual values ​​are arranged in a fixed order of vertices to form the predicted residual vector. The magnitude of the residual value directly reflects the degree of deviation between the predicted value and the actual measured value (a positive residual value indicates that the predicted value is lower than the actual value, and a negative residual value indicates that the predicted value is higher than the actual value). The degradation factor is used to quantify the current response capability of each detector (i.e., the actual response capability of the device). The deviation between the detected value and the standard dose value is calculated based on historical cumulative dose data. For each detection device, the standard calibration period reading (24 hours after the device is calibrated, when the device's response capability is in an ideal state) and the current period reading (24 hours before the current acquisition time) are extracted from the historical cumulative dose data. The average value of the standard calibration period reading (standard average dose value) and the average value of the current period reading (current average dose value) are calculated respectively. The ratio of the current average dose value to the standard average dose value is determined as the degradation factor of the device. If the ratio is close to 1, it indicates that the device's response capability is good; if the ratio deviates from 1 more, it indicates that the device's response capability is more severely degraded. Five points correspond to five degradation factors, which are arranged in a fixed order of vertices to form a degradation factor vector.

[0061] Combining the predicted residual vector and the degradation factor, a compensation coefficient adjustment is generated. This adjustment is then used to correct the compensation coefficient vector used in the current iteration, resulting in an updated compensation coefficient vector. Specifically, this involves: First, determining the core vectors involved in the calculation and their correspondences. Based on the obtained predicted residual vector (arranged in a fixed order by vertices, each element reflecting the deviation between the predicted dose value and the real-time measurement value at the corresponding point) and degradation factor vector (arranged in a fixed order by the same vertices, each element quantifying the degree of response degradation of the corresponding detection device), ensuring a one-to-one correspondence between the elements of the two vectors (i.e., the first element of the predicted residual vector corresponds to the first element of the degradation factor vector, and so on, all consistent with the fixed order of the five influencing points). Second, calculating the product term of the residual and the degradation factor. Each element of the predicted residual vector is multiplied by the corresponding degradation factor in the degradation factor vector. The core logic of this operation is that the greater the deviation of the degradation factor from 1, the more severe the degradation of the response capability of the corresponding detection device, and the greater the deviation weight of its measurement data needs to be. Therefore, the adjustment magnitude corresponding to the residual is adjusted according to the degree of device degradation through the product operation. The degree of synchronous increase or decrease enables precise and targeted correction of degraded equipment. For example, if the degradation factor at a certain point is 1.2 (indicating that the equipment response is 20% too high), and the corresponding residual value is 0.02 mSv / h, then the product result is 0.024 mSv / h. Compared with the case where the degradation factor is not considered, the adjustment range is reasonably amplified. In the third step, each of the above product results is multiplied by a preset iterative learning rate (set to 0.05). The value of this learning rate is determined based on the balance between model stability and convergence efficiency. The value of 0.05 can avoid excessively large single adjustment ranges that could damage the model. The oscillation (i.e., the compensation coefficient fluctuates greatly during iteration and cannot converge stably) can ensure that each iteration produces effective adjustments, ensuring that the model reaches a convergent state within a reasonable number of iterations; in the fourth step, after the above two steps of calculation, a total of five adjustment quantities are obtained (one for each point), which are arranged in a fixed order of the five influencing points (influencing point at the upper edge of the outer wall of the beam pipe, influencing point at the lower edge of the outer wall of the beam pipe, influencing point at the left side of the surface of the primary shield, influencing point at the right side of the surface of the primary shield, and influencing point at the central axis of the labyrinth entrance), forming a five-dimensional compensation coefficient adjustment vector.

[0062] Subsequently, the compensation coefficient vector is corrected. The source of the compensation coefficient vector used in the current iteration is clarified (if it is the first iteration, the initial compensation coefficient vector is used; if it is not the first iteration, the updated compensation coefficient vector output from the previous iteration is used). The vector is then added element-wise to the compensation coefficient adjustment vector constructed above. That is, the first element of the current compensation coefficient vector is added to the first element of the adjustment vector, the second element is added to the second element of the adjustment vector, and so on, to complete the update of the compensation coefficient for each point. Through this correction process, the compensation coefficient vector can fully absorb the real-time measurement deviation and equipment status information, and better match the current dose measurement situation and equipment operating status.

[0063] The iterative process of calculating the predicted vector, comparing the obtained residual vector, and generating adjustment and correction coefficients is repeated until the predicted residual vector meets the preset convergence condition. The compensation coefficient vector finally output by the dynamic compensation model at this point is then determined as the adaptive compensation coefficient corresponding to the influence points along the upper edge of the outer wall of the beam pipe, the lower edge of the outer wall of the beam pipe, the left side of the primary shield surface, the right side of the primary shield surface, and the central axis of the labyrinth entrance. Specifically, this includes repeatedly calculating the dose reading predicted vector, calculating the predicted residual vector and degradation factor, generating adjustment and correcting the compensation coefficient vector. After each iteration, it is determined whether the predicted residual vector meets the preset convergence condition. The convergence condition is that the absolute value of all elements in the predicted residual vector is less than the preset residual threshold (0.01 mSv / h), and the change in the residual vector between two consecutive iterations (the sum of the absolute values ​​of the differences between the corresponding elements in the two residual vectors) is less than the preset change threshold (0.005 mSv / h). When the iteration process meets the above convergence condition, the iteration operation stops. At this time, the updated compensation coefficient vector output by the dynamic compensation model is the final adaptive compensation coefficient vector. The five elements of this vector correspond one-to-one with the five affected points (vertex 1 to vertex 5), and can be directly used for accurate compensation correction of the dose readings at each point. That is, the final corrected dose value at this point = real-time dose measurement value × corresponding adaptive compensation coefficient.

[0064] In this embodiment, a geometric weight matrix constructed from spatial distance and relative azimuth angle quantifies the spatial correlation of five points into calculable weight parameters. This allows the dynamic compensation model to fully utilize spatial topological information, avoid errors caused by isolated processing of individual point data, and improve the spatial adaptability of compensation. The dynamic iteration process combines real-time dose data with historical accumulated data, while simultaneously quantifying equipment performance degradation using degradation factors. This enables the compensation coefficient to respond in real-time to changes in the radiation environment and equipment status, continuously correcting errors and ensuring the stability of compensation accuracy during long-term use. Through iterative convergence control of the predicted residuals, the final adaptive compensation coefficient is ensured to maximize the offsetting of the impact of spatial interference, equipment errors, and other factors on dose readings, outputting corrected data that is closer to the true radiation dose, providing reliable data support for radiation safety assessments and equipment optimization adjustments.

[0065] In a preferred embodiment of the present invention, based on the adaptive compensation coefficients of each vertex, spatial consistency calibration and degradation compensation are performed on the five-channel digital reading sequences to obtain high-confidence real-time dose values ​​for each affected point, including:

[0066] Using corresponding adaptive compensation coefficients, weighted corrections are applied to the real-time acquired first, second, third, fourth, and fifth digital reading sequences to obtain preliminary corrected dose values ​​for the influence points at the upper and lower edges of the beam duct outer wall, the left and right sides of the primary shield surface, and the central axis of the labyrinth entrance. Specifically, this involves: firstly establishing a unique correspondence between the five digital reading sequences and the five influence points to ensure unambiguous data association; secondly, strictly adhering to the established fixed vertex order, with the first digital reading sequence specifically corresponding to the influence point at the upper edge of the beam duct outer wall, the second to the lower edge, the third to the left side of the primary shield surface, and the fourth to the fifth... The sequence specifically corresponds to the influence points on the right side of the primary shield surface, and the fifth digital reading sequence specifically corresponds to the influence points on the central axis of the labyrinth entrance. This correspondence is maintained throughout the entire process to avoid data confusion. The adaptive compensation coefficient is essentially the ratio of the actual dose to the measurement reading of the detection equipment. It is used to quantitatively correct the systematic deviation of the detection equipment. Due to the degradation of the equipment's response after long-term use and the influence of spatial scattering interference on the measurement, there will be a fixed proportional deviation between the real-time digital reading output by the equipment and the actual dose at the point. Therefore, by multiplying the measured reading by the compensation coefficient, the dose data close to the actual value can be derived. For example, if the actual dose at a certain point is 0.54 mSv, but the equipment only measures 0.5 mSv due to response degradation, then the adaptive compensation coefficient obtained through iterative calculation is 1.08 (0.54 ÷ 0.5). Multiplying the measured reading of 0.5 by 1.08 can restore the value close to the actual value of 0.54 millisieverts (mSv) for deviation correction; each real-time acquired digital reading sequence is a continuous time-series data, with each sequence containing multiple synchronously acquired digital readings. The timestamps of all sequences are highly synchronized, with synchronization errors controlled within 10 milliseconds, ensuring comparability of dose data from different locations at the same time. For each digital reading sequence, the adaptive compensation coefficient corresponding to the specified location is extracted from the output adaptive compensation coefficient vector. This coefficient is obtained by combining historical equipment calibration data and real-time residual feedback iteration, and already includes quantitative information on the degree of equipment response degradation and spatial interference intensity at that location. For example, the adaptive compensation coefficient for the location affected by the beam duct outer wall along the upper edge of the first digital reading sequence is specifically matched to the detection equipment at that location. The deviation characteristics are then analyzed. Next, the digital readings corresponding to each timestamp in the sequence are traversed, and the digital reading of a single timestamp is multiplied by the extracted adaptive compensation coefficient. That is, the real-time digital reading is multiplied by the corresponding adaptive compensation coefficient, and the result is the initial corrected dose instantaneous value at the corresponding point under that timestamp. If the compensation coefficient is greater than 1, it means that the measured value of the equipment is less than the actual dose (such as the reading being too low due to decreased equipment sensitivity). The multiplication operation will proportionally amplify the measured value to be close to the actual value. If the compensation coefficient is less than 1, it means that the measured value of the equipment is greater than the actual dose (such as the reading being too high due to spatial scattering superposition). The multiplication operation will proportionally reduce the measured value to be close to the actual value. If the compensation coefficient is close to 1, it means that the equipment measurement deviation is extremely small, and the reading does not need to be adjusted significantly. Following the above method, multiplication operations for all timestamps were performed on each of the five digital reading sequences. The results of each operation were then organized according to the timestamp order, forming five independent preliminary corrected dose value sequences for each affected point. Each value in each sequence uniquely corresponds to the preliminary corrected dose value for that point at a given timestamp. This process initially offsets the equipment response degradation error and spatial scattering interference error contained in the original digital readings.

[0067] Based on the spatial correlation constraints provided by the geometric weight matrix, the five preliminary corrected dose values ​​are subjected to inverse distance weighted interpolation and moving average smoothing based on spatial distance. This yields high-confidence real-time dose values ​​for the following locations: the upper edge of the beam duct outer wall, the lower edge of the beam duct outer wall, the left side of the primary shield surface, the right side of the primary shield surface, and the central axis of the labyrinth entrance. Specifically, to further improve the accuracy, spatial consistency, and stability of the dose data, subsequent optimization is performed based on the generated geometric weight matrix. The normalized spatial distance weight in this matrix is ​​a core constraint parameter, and its value directly reflects the spatial correlation strength between different influence points (the larger the weight, the closer the spatial correlation between the two points, and the higher the mutual reference value of the dose data). The first step is to perform inverse distance weighted interpolation based on spatial distance, taking a single target influence point as the core (processing the five influence points as target points sequentially). First, a fixed timestamp is locked, and the preliminary corrected dose value of this target point, along with the other four non-target points, is extracted synchronously. The initial corrected dose value is obtained (ensuring all extracted data corresponds to the same timestamp to avoid spatial correlation distortion caused by time asynchrony); simultaneously, five weight values ​​are extracted from the geometric weight matrix corresponding to the target point row (these are the spatial distance weights between the target point and itself, and between the target point and four other non-target points, and the sum of the weights in this row is 1 to ensure the rationality of weighted fusion); the initial corrected dose value of each point (including the target point itself) is multiplied by the corresponding spatial distance weight. The closer the spatial correlation of the point, the higher the weight of its dose data and the greater its contribution, thus obtaining five weighted dose data; then, these five weighted dose data are summed, and the summation result is the corrected dose value of the target point at the current timestamp after inverse distance weighted interpolation. Through this interpolation process, the dose information of surrounding related points is effectively fused, and spatial correlation is used to compensate for the random error of individual point measurement (such as reading deviation caused by instantaneous equipment fluctuations), improve the spatial coordination of dose data of each point, and avoid the situation where single point anomalies deviate from the overall trend.

[0068] The second step involves performing a moving average smoothing process. For each point, after inverse distance weighted interpolation to form a corrected dose value sequence, a specific moving window length is first set. Considering the gradual changes in radiation dose in industrial irradiation scenarios (usually without instantaneous drastic fluctuations) and the time resolution of conventional dose detection equipment (1 second / time), the moving window length is set to 5 data points, corresponding to a time span of 5 seconds. This value can filter out short-term random fluctuations from a single measurement by covering data from 5 consecutive timestamps, while also avoiding masking the true slow trend of radiation dose changes (such as dose fluctuations caused by gradual changes in beam intensity) due to the moderate time span. The specific calculation process of the moving average is as follows: taking a certain data point in the sequence as the center, select 5 consecutive data points within the window. For example, when the 3rd data point is the center, the window covers the 1st to 5th data points. The values ​​of these 5 data points are summed, and then the sum is divided by 5 (the number of data points in the window). The resulting arithmetic mean is the smoothed value of the current center data point. This value can more objectively reflect the average dose level of the surrounding area over a short period of time at that moment, according to the timestamp. The above-mentioned central window averaging operation is performed sequentially on each data point in the sequence. For data points at the beginning and end of the sequence where a complete 5-point window cannot be constructed with itself as the center, a single-sided window processing method is used to ensure the integrity of the sequence. For the first data point at the beginning of the sequence, only the arithmetic mean of the 5 consecutive data points following it (the 1st to the 5th) is calculated; for the second data point, only the arithmetic mean of the 5 consecutive data points following it (the 2nd to the 6th) is calculated; for the last data point at the end of the sequence, only the arithmetic mean of the 5 consecutive data points preceding it (the 5th to the 6th) is calculated. The arithmetic mean of the first data point is calculated; for the second-to-last data point, the arithmetic mean of the five consecutive data points preceding it (from the sixth to the second-to-last) is calculated; after two progressive processing steps of inverse distance weighted interpolation (optimizing spatial consistency and offsetting random errors) and moving average smoothing (filtering short-term fluctuations and ensuring data stability), the dose value sequence of each affected point is comprehensively optimized, and the sequence is finally determined as the high-confidence real-time dose value of the corresponding point. The value corresponding to each timestamp in the sequence can reflect the actual radiation dose of the point at that moment.

[0069] This embodiment effectively offsets systematic errors caused by detection equipment degradation and spatial interference by multiplying the adaptive compensation coefficient with the corresponding digital readings. Inverse distance weighted interpolation based on the geometric weight matrix fully utilizes the spatial correlation of the five points, integrating dose information from multiple points, avoiding the bias of individual point measurements, reducing the impact of random errors on the results, and improving the spatial consistency of dose data from each point. Moving average processing effectively filters out short-term random fluctuations in dose readings, preserving the true trend of radiation dose changes, making the final high-confidence real-time dose value more stable and valuable for reference. From initial correction to interpolation optimization and then to smoothing, the process is logically coherent and progressive, ensuring both data accuracy and stability. The final high-confidence real-time dose value provides reliable data support for radiation safety monitoring, equipment operation adjustment, and protection scheme optimization.

[0070] In a preferred embodiment of the present invention, the calibration is completed by determining whether the overall measurement deviation exceeds a preset tolerance based on the population drift trend of the high-confidence real-time dose value in the continuous monitoring time series, including:

[0071] Time series analysis was performed on five continuously acquired high-confidence real-time dose values ​​to extract the long-term trend components for each point. The slope of the long-term trend component for each point was compared with its respective preset drift threshold. When the slope of the trend component for any point exceeded the corresponding drift threshold, measurement drift was determined to have occurred. Specifically, the monitoring object was identified as the continuously output high-confidence real-time dose value sequence from five influencing points: the upper edge of the outer wall of the beam duct, the lower edge of the outer wall of the beam duct, the left side of the surface of the primary shield, the right side of the surface of the primary shield, and the central axis of the labyrinth entrance. The dose value sequence for each point was arranged in order according to the timestamp. Time series analysis was performed on the dose value sequence for each point separately, and the long-term trend component was extracted using the trend-period-random decomposition method. The core of this method is to gradually remove interference components and ultimately retain the components that reflect the long-term trend. The core component of the change is first processed using a moving average method on the original sequence. The sliding window length is set to 5 data points (corresponding to a 5-second time span). Specifically, starting from the first data point of the sequence, the arithmetic mean is calculated for each group of 5 consecutive data points. The average value is used to replace the original value of the center data point in the group. The first and last data points are processed according to the one-sided window rule. This operation filters out short-term random fluctuations such as instantaneous noise and electromagnetic interference from the equipment, resulting in a smoothed sequence. The periodic component is identified and removed based on Fourier transform. The core is to convert the dose data (time-domain signal) that changes with time into a signal distributed according to frequency (frequency-domain signal) by using fixed operation rules. The conversion core is the combination of dose value and trigonometric function. The real and imaginary parts are the basic components of the frequency-domain signal. The specific process is as follows: First, it is clear that the object of processing is the dose value sequence smoothed by moving average, which is a time-domain signal arranged in time. Assuming data is collected once per second for a total of N times, the time-domain signal can be described as dose value x(0) at second 0, dose value x(1) at second 1, ... dose value x(N-1) at second N-1, where x (time point) represents the dose value at the corresponding moment, and the time point is represented by an integer from 0 to N-1, corresponding to the actual collection timestamp.Our goal is to identify hidden periodicities in the data, such as 24-hour or 1-hour cycles. These periodicities are essentially the recurring occurrences of signals at specific frequencies. Therefore, we need to convert the time-domain signal into a frequency-domain signal arranged by frequency. Each frequency corresponds to a frequency-domain signal component described by both real and imaginary parts, which together reflect the contribution of dose values ​​at all time points to that frequency. Taking the calculation of the frequency-domain signal component at a certain frequency K as an example (this process needs to be repeated for all frequencies), the steps are interconnected. The first step is to determine the trigonometric function parameters corresponding to the frequency. Each frequency K corresponds to a fixed set of cosine functions cos(2π×K×n / N) and sine functions sin(2π×K×n / N). The rate of function fluctuation is determined by K; the higher the frequency, the more frequent the fluctuation. Here, n is the time point of the time-domain signal (0 to N-1), K is the current frequency point being calculated (0 to N-1, each K representing a different frequency), and N is the total number of data points in the time-domain signal. Next, calculate the product of the dose value and the trigonometric function at each time point. For each time point n, multiply the dose value x(n) at that point by the cosine function value and the sine function value at the corresponding time to obtain the cosine product term and the sine product term. For example, when K=1 and n=0, the cosine function value is cos(0)=1 and the sine function value is sin(0)=0. If x(0)=0.5 mSv, then the cosine product term is 0.5×1=0.5 and the sine product term is 0.5×0=0. When n=1, multiply the corresponding function value by x(1) again. This completes the calculation for all time points. Finally, sum the products of all time points to obtain the real part and the imaginary part. That is, add all the cosine product terms together. The sum is the real part of the frequency domain component corresponding to the frequency K. After adding all the sine product terms, multiply by -1 (Fourier transform). (Fixed rules) The result is the imaginary part of the component corresponding to frequency K. The real part is the total contribution of the dose value to the cosine fluctuation of the frequency, and the imaginary part is the total contribution to the sine fluctuation. The two are like the horizontal and vertical axes, jointly determining the signal characteristics of a frequency. Perform the above operation on all frequency points K (0 to N-1) to obtain the frequency domain signal components (all containing real and imaginary parts) corresponding to all frequencies, completing the conversion from the time domain to the frequency domain. Since the real and imaginary parts cannot directly determine the strength of periodicity, amplitude quantization is required. First, the real and imaginary parts of the frequency component are squared separately, then the squared results are added together to obtain the sum of squares. Finally, the square root of the sum of squares is the value obtained, which is the amplitude of the frequency. The larger the amplitude, the stronger the presence of the frequency signal in the original time domain data, and the more significant the corresponding periodicity.

[0072] Based on the known periodic characteristics of industrial irradiation scenarios (equipment 24-hour operating cycle, beam current 1-hour adjustment cycle), the characteristic frequency is obtained by converting "frequency = 1 / cycle" (24-hour cycle corresponds to 1 / 24). A 1-hour cycle corresponds to 1 Then, combined with a 1-second acquisition interval (i.e., 1 / 3600 hours), the peak amplitude corresponding to these characteristic frequencies in the frequency domain is located. Subsequently, a filtering rule is constructed to set the amplitude of characteristic frequency components whose amplitude exceeds 1.5 times the average amplitude of all amplitudes to 0, while retaining the frequency domain signals of non-periodic and weakly periodic components. Finally, through inverse Fourier transform, the processed frequency domain signal is converted back to the time domain signal to obtain a sequence with periodic fluctuations removed. The consistency of the sequence with periodic fluctuations removed is checked with the smoothed sequence obtained in the first step. The numerical difference between the corresponding timestamps of the two sequences is calculated point by point. If the absolute value of all differences is less than 0.0001 millisieverts, it indicates that there is no obvious residual interference. This sequence is the long-term trend component reflecting the long-term change law of dose. This component only reflects slow-changing factors such as equipment drift and long-term beam attenuation, and is not affected by short-term fluctuations and periodic interference.

[0073] The slope of the long-term trend component at each point is calculated using linear regression. The slope reflects the rate of trend change. The specific calculation process is as follows: the timestamps of the trend components are converted into continuous time variables. Taking the collection time of the first data point in the sequence as the baseline (denoted as 0 hours), each subsequent data point is accumulated at a collection interval of 1 second. The accumulated result is then converted to hourly units (1 second = 1 / 3600 hours). Let the time variable sequence of the trend component at a certain point be t1, t2, ..., tn (n is the total number of data points in the sequence), and the corresponding trend component values ​​be y1, y2, ..., yn. The average value of the time variables and the average value of the trend component values ​​are calculated respectively, i.e., the average value t of the time variables. - Trend component average ; Compare each time variable with t - The difference is multiplied by the corresponding trend component value and The difference is then summed, i.e., the numerator S1 = (t1 - t2) / ( ... - )×(y1- ) + (t2 - t - )×(y2- )+…+(tn-t - )×(yn- ); Compare each time variable with t - Square the differences separately, then add all the squared results together, that is, the denominator S2 = (t1 - t2) / (t1 - t2) / (t2 ... - )²+(t2-t - )²+…+(tn-t -Slope k = numerator S1 ÷ denominator S2. A positive slope indicates a long-term upward trend in dose; a negative slope indicates a long-term downward trend. The larger the absolute value, the faster the dose change rate and the higher the risk of equipment drift. Considering the differences in the working environment of the five locations, independent drift thresholds are preset. The threshold values ​​are determined based on the degree of radiation influence at each location. The influence thresholds at the upper and lower edges of the beam duct outer wall are 0.002 mSv / h (directly close to the beam, with large dose base and fluctuations); the influence thresholds at the left and right sides of the primary shield surface are 0.001 mSv / h (radiation is weakened by shielding, and the dose is relatively stable). The influence point of the central axis at the entrance of the maze is 0.0005 mSv / h (the end of radiation diffusion, with the lowest and most stable dose). The absolute value of the slope of the trend component at each point is compared with the corresponding drift threshold one by one. If the absolute value of the slope of any point exceeds its corresponding drift threshold, for example, the slope of the point at the upper edge of the beam channel is 0.0025 mSv / h, which exceeds the threshold of 0.002, the system is immediately determined to have measurement drift, triggering the subsequent calibration process. If the absolute value of the slope of all points is within the corresponding threshold range, the system returns to the monitoring stage to continue to collect and analyze the high-confidence real-time dose value sequence of each point in real time.

[0074] After determining that a measurement drift has occurred, a reference radiation source with a known dose rate is controlled and sequentially moved to the standard measurement positions of five affected points. The radiation dose response readings of the five ionization chamber detectors at their respective standard measurement positions to the reference radiation source are recorded, resulting in five standard comparison readings. The difference between the standard comparison reading for each affected point and the arithmetic mean of the high-confidence real-time dose values ​​for the corresponding point within a preset time period before the measurement drift is determined is calculated, yielding the reading deviation values ​​for the five points. Specifically, after determining that a measurement drift has occurred, the reference radiation source calibration process is initiated. The reference radiation source is a standard radiation source with a known and stable dose rate (its dose rate is calibrated, and the error is controlled within a preset range of ±1%). An automated robotic arm controls its movement trajectory, precisely moving it sequentially to the standard measurement positions of the five affected points. The standard measurement positions completely coincide with the detection centers of the detection equipment at each point. The preset spatial accuracy is set to a dynamic positioning error of no more than ±0.5 mm. After the reference radiation source reaches the standard measurement position of a certain point, it maintains a stable output and controls the ionization chamber detector corresponding to that point to perform multiple continuous readings (e.g., 10 consecutive readings with a 1-second interval). After removing the maximum and minimum values ​​from the readings, the arithmetic mean of the remaining 8 readings is calculated as the standard comparison reading for that point. The readings of five points are collected sequentially according to the above process to obtain five standard comparison readings corresponding to each point. The high-confidence real-time dose values ​​of each point within a preset time period (e.g., the previous 24 hours) before the measurement drift occurs are extracted. The arithmetic mean of all dose values ​​at each point during this period is calculated. The standard comparison reading of each point is subtracted from the average value corresponding to that point to obtain the reading deviation value of the five points. A positive deviation value indicates that the measured value after drift is lower than the standard value; a negative deviation value indicates that the measured value after drift is higher than the standard value.

[0075] Based on the reading deviations at five points, the parameter correction amount used to correct the dynamic compensation model is calculated. This parameter correction amount is then used to update the geometric weight matrix and historical cumulative dose data used in the dynamic compensation model, completing the feedback calibration. Specifically, this includes: first, calculating the parameter correction amount based on the reading deviations at five points. The calculation of the correction amount needs to consider the spatial correlation and drift influence of each point. The drift at different points has different impacts on the overall measurement accuracy of the system, and there is spatial correlation between points. Therefore, this difference needs to be balanced by the normalized weights of the geometric weight matrix. The specific calculation logic is as follows: The generated geometric weight matrix is ​​called, and the normalized weights corresponding to each point in the matrix are extracted (…). The weights reflect the importance of each location in spatial correlation. The reading deviation of each location is multiplied by its corresponding weight to obtain a weighted deviation value (the higher the weight of a location, the greater its contribution to the overall correction). The sum of the five weighted deviation values ​​yields the overall correction coefficient of the geometric weight matrix. The core function of the overall correction coefficient is to synthesize the drift deviation of each location with the strength of spatial correlation, generating a unified adjustment parameter. This ensures that after the geometric weight matrix is ​​updated, it can both offset the systematic deviation caused by drift and maintain the original reasonable spatial correlation between locations, avoiding matrix distortion due to single-point drift. Simultaneously, the reading deviation of each location is compared with the average change of the historical cumulative dose data for that location. The correction increments for the historical data at that location are obtained by multiplying the correction rates. Five correction increments, arranged in location order, form the historical data correction vector. The parameter correction consists of two parts: a geometric weight matrix correction and a historical cumulative dose data correction. The geometric weight matrix is ​​updated by multiplying each element of the original matrix by its corresponding overall correction coefficient, ensuring that the matrix still satisfies row normalization (the sum of weights in each row is 1), guaranteeing the rationality of subsequent spatial weighting calculations. The historical cumulative dose data is updated by adding the corresponding location's correction increment to the dose value at each timestamp in the original historical data sequence. If the updated data exceeds a reasonable range (the reasonable range is...), the correction increment will be applied. The range is 0.01 to 0.5 mSv / h, covering the normal environmental background dose and normal working dose range in industrial irradiation scenarios. A reasonable boundary value (0.01 mSv / h for doses below 0.01 mSv / h and 0.5 mSv / h for doses above 0.5 mSv / h) is taken as the final update value to avoid abnormal data affecting model calculations. The updated geometric weight matrix and historical cumulative dose data are substituted into the dynamic compensation model to replace the original parameters, completing the model's feedback calibration. After calibration, the model will recalculate the adaptive compensation coefficients based on the new parameters to ensure that subsequent output dose values ​​can offset the errors caused by drift and restore measurement accuracy.

[0076] This embodiment extracts long-term trend components and calculates slopes through time series analysis, and determines drift by combining it with a location-specific drift threshold. This avoids false calibration triggered by short-term fluctuations, enabling early warning and accurate identification of measurement drift and reducing dose data distortion caused by drift. Standard comparison readings are obtained based on a reference radiation source with a known dose rate. Random errors are reduced by averaging multiple readings, ensuring the accuracy of deviation calculation. Precise positioning of the standard measurement location further ensures the matching between calibration data and the location, providing a reliable basis for model correction. Targeted parameter correction amounts are calculated based on the reading deviation values, enabling dynamic updates of the geometric weight matrix and historical cumulative dose data. This allows the dynamic compensation model to adapt to drift caused by factors such as equipment aging and environmental changes, avoiding accuracy degradation due to fixed model parameters.

[0077] like Figure 2 As shown, embodiments of the present invention also provide a method for online measurement and feedback calibration of electron irradiation dose, including:

[0078] Step 1: Convert the original dose signal into a five-channel digital reading sequence, and perform timestamp alignment and spatial coordinate calibration on the five-channel digital reading sequences to generate a three-dimensional discrete coordinate dataset characterizing the spatial distribution of radiation field intensity.

[0079] Step 2: Identify the point of global maximum radiation intensity from the three-dimensional discrete coordinate dataset as the origin of spatial registration, and extract two characteristic lines representing the dominant direction of radiation scattering. Calculate the spatial azimuth angle between the two characteristic lines to determine the main influence area of ​​radiation scattering.

[0080] Step 3: Within the main influence area of ​​radiation scattering, determine five influence points located at the upper and lower edges of the axial direction of the outer wall of the beam pipe, the left and right sides of the normal direction of the primary shield surface, and the central axis of the labyrinth entrance. These five influence points constitute a spatial non-coplanar five-vertex topology.

[0081] Step 4: Based on the geometric configuration of the non-coplanar five-vertex topology and the real-time dose readings corresponding to each vertex, construct a dynamic compensation model to obtain the adaptive compensation coefficients for each vertex.

[0082] Step 5: Based on the adaptive compensation coefficient of each vertex, perform spatial consistency calibration and degradation compensation on the five-channel digital reading sequence to obtain the high-confidence real-time dose value of each affected point.

[0083] Step 6: Based on the population drift trend of the high-confidence real-time dose value in the continuous monitoring time series, determine whether the overall measurement deviation exceeds the preset tolerance in order to complete the calibration.

[0084] It should be noted that this method is the same as the method described above for the system. All implementation methods in the above system embodiments are applicable to this embodiment and can achieve the same technical effect.

[0085] Embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the system as described above. All implementations in the above system embodiments are applicable to this embodiment and can achieve the same technical effects.

[0086] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A system for online measurement and feedback calibration of electron irradiation dose, characterized in that, The method comprises the following steps: A processing module is used to collect raw dose signals of ionization chamber detectors arranged at an upper position of an outer wall of a beam tube, a lower position of the outer wall of the beam tube, a left position of a surface of a primary shielding body, a right position of the surface of the primary shielding body, and a central axis position of a maze entrance of an irradiation chamber, convert the raw dose signals into five digital reading sequences, and perform time stamp alignment and spatial coordinate calibration on the five digital reading sequences to generate a three-dimensional discrete coordinate data set representing the spatial distribution of the intensity of a radiation field; An extraction module is used to identify a global maximum value point of radiation intensity from the three-dimensional discrete coordinate data set as a spatial registration origin, extract two characteristic straight lines representing dominant directions of radiation scattering, calculate a spatial orientation angle between the two characteristic straight lines to determine a main influence area of radiation scattering; A deployment module is used to determine five influence points at the upper and lower positions along the axial direction of the outer wall of the beam tube, the left and right positions along the normal direction of the surface of the primary shielding body, and the central axis of the maze entrance within the main influence area of radiation scattering, and the five influence points form a spatial non-coplanar five-vertex topological structure; A construction module is used to construct a dynamic compensation model according to the geometric configuration of the spatial non-coplanar five-vertex topological structure and real-time dose readings corresponding to each vertex to obtain adaptive compensation coefficients of each vertex; A calibration module is used to perform spatial consistency calibration and degradation compensation on the five digital reading sequences based on the adaptive compensation coefficients of each vertex to obtain high-confidence real-time dose values of each influence point; A feedback module is used to determine whether the overall measurement deviation exceeds a preset tolerance according to the group drift trend of the high-confidence real-time dose values in the continuous monitoring time sequence to complete calibration.

2. The system for online measurement and feedback calibration of electron irradiation dose according to claim 1, characterized in that, The raw dose signals are converted into five digital reading sequences, and the five digital reading sequences are time stamped and aligned and calibrated in spatial coordinates to generate a three-dimensional discrete coordinate data set representing the spatial distribution of the intensity of a radiation field, including: Analog-digital conversion is performed on the raw dose signals output by each ionization chamber detector to obtain corresponding digital readings, forming five digital reading sequences, and each ionization chamber detector is assigned a spatial three-dimensional coordinate measured from the fixed installation position of the corresponding ionization chamber detector; A unified clock source is used to time synchronize the five digital reading sequences, so that the sampling points of each sequence have the same time stamp, and the digital readings of each ionization chamber detector at each sampling time after synchronization are bound with the spatial three-dimensional coordinates of the corresponding ionization chamber detector to form a three-dimensional discrete coordinate data set.

3. The system for online measurement and feedback calibration of electron irradiation dose according to claim 2, characterized in that, A global maximum value point of radiation intensity is identified from the three-dimensional discrete coordinate data set as a spatial registration origin, two characteristic straight lines representing dominant directions of radiation scattering are extracted, a spatial orientation angle between the two characteristic straight lines is calculated to determine a main influence area of radiation scattering, including: Traverse all data points in the three-dimensional discrete coordinate data set, compare the dose intensity values of each data point, define the spatial three-dimensional coordinates corresponding to the data point with the maximum dose intensity value as the spatial registration origin, and set a neighborhood radius with the spatial registration origin as the center, screen all data points in the three-dimensional discrete coordinate data set whose spatial distance from the spatial registration origin is less than the neighborhood radius to form a neighborhood data point subset; Perform principal component analysis on the spatial three-dimensional coordinates of all data points in the neighborhood data point subset to calculate the first principal component direction and the second principal component direction, and fit the first principal component direction as a first characteristic straight line through the spatial registration origin in space, and fit the second principal component direction as a second characteristic straight line through the spatial registration origin in space; Calculate the included angle between the direction vector of the first characteristic straight line and the direction vector of the second characteristic straight line to obtain the spatial orientation angle, and sweep the conical space region formed by the spatial orientation angle around the spatial registration origin in three-dimensional space to determine the main influence area of radiation scattering.

4. The system for online measurement and feedback calibration of electron irradiation dose according to claim 3, characterized in that, In the main influence area of radiation scattering, five influence points located at the upper and lower edges of the beam pipe outer wall axis, the left and right sides of the primary shielding body surface normal, and the center axis of the maze entrance are determined, and the five influence points form a spatial non-coplanar five-vertex topological structure, including: In the conical space region defined by the main influence area of radiation scattering, the intersection line of the beam pipe outer wall cylindrical surface equation and the main influence area of radiation scattering is calculated according to the known beam pipe outer wall cylindrical surface equation, and the highest point and the lowest point on the intersection line are taken as the upper edge influence point and the lower edge influence point of the beam pipe outer wall axis, respectively. According to the known primary shielding body outer surface plane equation, the intersection of the primary shielding body outer surface plane equation and the main influence area of radiation scattering is calculated, and the leftmost point and the rightmost point in the intersection region are taken as the left side influence point and the right side influence point of the primary shielding body surface normal, respectively. According to the known maze entrance center line space straight line equation, the intersection point of the maze entrance center line space straight line equation and the boundary of the main influence area of radiation scattering is calculated as the center axis influence point of the maze entrance. The spatial three-dimensional coordinates of the upper edge influence point of the beam pipe outer wall, the lower edge influence point of the beam pipe outer wall, the left side influence point of the primary shielding body surface, the right side influence point of the primary shielding body surface, and the center axis influence point of the maze entrance are not in the same plane, and the lines connecting them in turn form a spatial polygon containing five vertices, which is a spatial non-coplanar five-vertex topological structure containing five vertices.

5. The system for online measurement and feedback calibration of electron irradiation dose according to claim 4, characterized in that, According to the geometric configuration of the spatial non-coplanar five-vertex topological structure and the real-time dose readings corresponding to each vertex, a dynamic compensation model is constructed to obtain adaptive compensation coefficients for each vertex, including: According to the spatial distance and relative azimuth angle between any two vertices of the spatial non-coplanar five-vertex topological structure, a geometric weight matrix describing the spatial correlation of the five vertices is calculated; Real-time dose readings of the upper edge influence point on the outer wall of the beam tube, the lower edge influence point on the outer wall of the beam tube, the left side influence point on the surface of the primary shielding body, the right side influence point on the surface of the primary shielding body, and the influence point of the center axis of the maze entrance are obtained to form a real-time dose reading vector; The geometric weight matrix is used as a spatial correlation constraint, and the instantaneous radiation field distribution information contained in the real-time dose reading vector is used as an observation reference. Combined with the response degradation trend represented by the historical cumulative dose data, the input data is iteratively fitted through a recursive learning algorithm to establish and update the dynamic compensation model. After each iteration operation of the dynamic compensation model, a compensation coefficient vector containing five components is output. The five coefficients in the compensation coefficient vector are respectively adaptive compensation coefficients corresponding to the upper edge influence point on the outer wall of the beam tube, the lower edge influence point on the outer wall of the beam tube, the left side influence point on the surface of the primary shielding body, the right side influence point on the surface of the primary shielding body, and the influence point of the center axis of the maze entrance.

6. The system for online measurement and feedback calibration of electron irradiation dose according to claim 5, characterized in that, The geometric weight matrix is used as a spatial correlation constraint, and the instantaneous radiation field distribution information contained in the real-time dose reading vector is used as an observation reference. Combined with the response degradation trend represented by the historical cumulative dose data, the input data is iteratively fitted through a recursive learning algorithm to establish and update the dynamic compensation model. After each iteration operation of the dynamic compensation model, a compensation coefficient vector containing five components is output. The five coefficients in the compensation coefficient vector are respectively adaptive compensation coefficients corresponding to the upper edge influence point on the outer wall of the beam tube, the lower edge influence point on the outer wall of the beam tube, the left side influence point on the surface of the primary shielding body, the right side influence point on the surface of the primary shielding body, and the influence point of the center axis of the maze entrance, including: The geometric weight matrix is used as a spatial correlation constraint, and the instantaneous radiation field distribution information contained in the real-time dose reading vector is used as an observation reference. Combined with the response degradation trend represented by the historical cumulative dose data, the input data is iteratively fitted through a recursive learning algorithm to establish and update the dynamic compensation model. After each iteration operation of the dynamic compensation model, a compensation coefficient vector containing five components is output. The five coefficients in the compensation coefficient vector are respectively adaptive compensation coefficients corresponding to the upper edge influence point on the outer wall of the beam tube, the lower edge influence point on the outer wall of the beam tube, the left side influence point on the surface of the primary shielding body, the right side influence point on the surface of the primary shielding body, and the influence point of the center axis of the maze entrance. The geometric weight matrix is used as a spatial correlation constraint, and the instantaneous radiation field distribution information contained in the real-time dose reading vector is used as an observation reference. Combined with the response degradation trend represented by the historical cumulative dose data, the input data is iteratively fitted through a recursive learning algorithm to establish and update the dynamic compensation model. After each iteration operation of the dynamic compensation model, a compensation coefficient vector containing five components is output. The five coefficients in the compensation coefficient vector are respectively adaptive compensation coefficients corresponding to the upper edge influence point on the outer wall of the beam tube, the lower edge influence point on the outer wall of the beam tube, the left side influence point on the surface of the primary shielding body, the right side influence point on the surface of the primary shielding body, and the influence point of the center axis of the maze entrance. The geometric weight matrix is used as a spatial correlation constraint, and the instantaneous radiation field distribution information contained in the real-time dose reading vector is used as an observation reference. Combined with the response degradation trend represented by the historical cumulative dose data, the input data is iteratively fitted through a recursive learning algorithm to establish and update the dynamic compensation model. After each iteration operation of the dynamic compensation model, a compensation coefficient vector containing five components is output. The five coefficients in the compensation coefficient vector are respectively adaptive compensation coefficients corresponding to the upper edge influence point on the outer wall of the beam tube, the lower edge influence point on the outer wall of the beam tube, the left side influence point on the surface of the primary shielding body, the right side influence point on the surface of the primary shielding body, and the influence point of the center axis of the maze entrance. The geometric weight matrix is used as a spatial correlation constraint, and the instantaneous radiation field distribution information contained in the real-time dose reading vector is used as an observation reference. Combined with the response degradation trend represented by the historical cumulative dose data, the input data is iteratively fitted through a recursive learning algorithm to establish and update the dynamic compensation model. After each iteration operation of the dynamic compensation model, a compensation coefficient vector containing five components is output. The five coefficients in the compensation coefficient vector are respectively adaptive compensation coefficients corresponding to the upper edge influence point on the outer wall of the beam tube, the lower edge influence point on the outer wall of the beam tube, the left side influence point on the surface of the primary shielding body, the right side influence point on the surface of the primary shielding body, and the influence point of the center axis of the maze entrance. The iterative process of calculating the prediction vector, comparing the residual vector, generating the adjustment amount and the correction coefficient is repeatedly performed until the prediction residual vector meets the preset convergence condition. The compensation coefficient vector finally output by the dynamic compensation model at this time is determined as the adaptive compensation coefficient corresponding to the upper edge influence point of the beam tube outer wall, the lower edge influence point of the beam tube outer wall, the left side influence point of the primary shielding body surface, the right side influence point of the primary shielding body surface and the influence point of the center axis of the maze entrance.

7. The system for online measurement and feedback calibration of electron irradiation dose according to claim 6, characterized in that, Based on the adaptive compensation coefficients of the vertices, the five digital reading sequences are calibrated and compensated for spatial consistency to obtain high-confidence real-time dose values at each influence point, including: Using the corresponding adaptive compensation coefficients, the first, second, third, fourth and fifth digital reading sequences collected in real time are weighted and corrected to obtain preliminary corrected dose values at the upper edge influence point of the beam tube outer wall, the lower edge influence point of the beam tube outer wall, the left side influence point of the primary shielding body surface, the right side influence point of the primary shielding body surface and the influence point of the center axis of the maze entrance. According to the spatial correlation constraints provided by the geometric weight matrix, the five preliminary corrected dose values are inversely distance weighted and interpolated based on spatial distance and smoothed by sliding average to finally obtain high-confidence real-time dose values at the upper edge influence point of the beam tube outer wall, the lower edge influence point of the beam tube outer wall, the left side influence point of the primary shielding body surface, the right side influence point of the primary shielding body surface and the influence point of the center axis of the maze entrance.

8. The system for online measurement and feedback calibration of electron irradiation dose according to claim 7, characterized in that, According to the group drift trend of the high-confidence real-time dose values in the continuous monitoring time sequence, it is determined whether the overall measurement deviation exceeds the preset tolerance to complete calibration, including: The five high-confidence real-time dose values obtained continuously are analyzed in time sequence, and the long-term trend component of each point is extracted. The slope of the long-term trend component of each point is compared with the respective preset drift threshold. When the slope of the trend component of any one point exceeds the corresponding drift threshold, it is determined that measurement drift occurs. After determining that measurement drift occurs, a reference radiation source with a known dose rate is moved to the standard measurement position of the five influence points in turn, and the radiation dose response readings of the five ionization chamber detectors at their respective standard measurement positions are recorded to obtain five standard comparison readings. The difference between the standard comparison reading corresponding to each influence point and the arithmetic mean of the high-confidence real-time dose values of the corresponding point within a preset time period before the measurement drift is determined to be a reading deviation value of the five points. According to the reading deviation values of the five points, a parameter correction amount for correcting the dynamic compensation model is calculated, and the geometric weight matrix and the historical cumulative dose data used in the dynamic compensation model are updated using the parameter correction amount to complete feedback calibration.

9. A method for online measurement and feedback calibration of electron irradiation dose, using a system according to any one of claims 1 to 8, characterized in that, including: Step 1, collect the original dose signals of ionization chamber detectors arranged at the upper side of the outer wall of the beam tube, the lower side of the outer wall of the beam tube, the left side of the surface of the primary shielding body, the right side of the surface of the primary shielding body, and the central axis position of the entrance of the maze of the irradiation chamber, convert the original dose signals into a five-way digital reading sequence, and time stamp align and space coordinate calibrate the five-way digital reading sequence to generate a three-dimensional discrete coordinate data set representing the spatial distribution of the radiation field intensity; Step 2, identify the global maximum point of the radiation intensity from the three-dimensional discrete coordinate data set as the spatial registration origin, extract two characteristic straight lines representing the dominant directions of radiation scattering, calculate the spatial azimuth angle between the two characteristic straight lines to determine the main influence area of radiation scattering; Step 3, within the main influence area of radiation scattering, determine five influence points located at the upper and lower sides of the axial direction of the outer wall of the beam tube, the left and right sides of the normal direction of the surface of the primary shielding body, and the central axis of the entrance of the maze, and the five influence points form a spatial non-coplanar five-vertex topological structure; Step 4, according to the geometric configuration of the spatial non-coplanar five-vertex topological structure and the real-time dose readings of each vertex, a dynamic compensation model is constructed to obtain adaptive compensation coefficients for each vertex; Step 5, based on the adaptive compensation coefficients of each vertex, the spatial consistency calibration and degradation compensation of the five-way digital reading sequence is carried out to obtain high-confidence real-time dose values of each influence point; Step 6, according to the group drift trend of the high-confidence real-time dose values in the continuous monitoring time sequence, it is determined whether the overall measurement deviation exceeds the preset tolerance to complete the calibration.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a program which is executed by the processor to realize the system of any one of claims 1 to 8.

Images