A wearable nuclear radiation protection device optimization configuration method and device

Abstract

The present invention proposes an optimization configuration method and device for wearable nuclear radiation protection equipment, which belongs to the field of nuclear radiation detection technology. The method includes: establishing a three-dimensional geometric model of the radiation shielding layer inside the wearable nuclear radiation protection equipment, and calculating the shielding coefficients at different positions; simulating the propagation process of nuclear radiation inside the wearable nuclear radiation protection equipment, and identifying the hot spots of radiation distribution; arranging detection units in the radiation hot spots, collecting accurate time domain data of radiation dose; converting the time domain data of radiation dose into frequency domain feature data of radiation dose distribution; establishing the frequency domain features of the radiation dose distribution and a renal injury prediction model based on deep learning, and integrating them into the wearable nuclear radiation protection equipment to evaluate the risk of renal injury of personnel in real time. In the present invention, by using a machine learning method, physical characteristics and statistical characteristics are integrated to more accurately predict the degree of radiation damage to the kidneys.

Description

Technical Field

[0001] The present invention belongs to the technical field of nuclear radiation detection, and in particular relates to an optimization configuration method and device for wearable nuclear radiation protection equipment. Background Art

[0002] Radiation monitoring is an important task in the military industry and other fields. In battlefields or nuclear power plants, personnel or workers may be exposed to intense radiation, which can cause serious damage to their health, especially to kidney function. Therefore, accurate prediction and effective warning of radiation-induced kidney damage are critical to protecting the lives of relevant personnel.

[0003] Existing radiation safety monitoring technologies mainly focus on the detection and evaluation of radiation doses. By deploying various radiation detectors, the radiation intensity can be monitored in real time, and the health risks of radiation-exposed personnel can be evaluated based on the cumulative dose. However, this evaluation method based on radiation dose has certain limitations:

[0004] Radiation dose cannot directly reflect the propagation and deposition process of radiation in the human body, and cannot accurately predict the damage to specific organs.

[0005] Different types and energies of radiation cause different degrees of damage to various organs in the human body, and a single dose indicator cannot fully reflect this difference.

[0006] Radiation damage often has a certain lag, and dose assessment cannot provide timely warning of possible damage.

[0007] Therefore, there is an urgent need for a new optimal configuration method for wearable nuclear radiation protection equipment that can more accurately and timely assess the risk of radiation-induced kidney damage and provide effective protection measures for relevant personnel. Summary of the invention

[0008] In view of this, the present invention provides a method for optimizing the configuration of a wearable nuclear radiation protection device.

[0009] Step 1, establishing a three-dimensional geometric model of the internal radiation shielding layer of the wearable nuclear radiation protection device, and calculating the overall shielding coefficient at different positions;

[0010] Step 2, simulating the propagation process of nuclear radiation inside the wearable nuclear radiation protection device to identify the hot spots of radiation distribution;

[0011] Step 3, arranging detection units in the radiation hotspot areas to collect accurate time domain data of radiation dose; converting the time domain data of radiation dose into frequency domain characteristic data of radiation dose distribution;

[0012] Step 4: Based on the frequency domain characteristic data of the radiation dose distribution, a deep learning-based kidney injury prediction model is established, and the model is integrated into the wearable nuclear radiation protection device to evaluate the risk of kidney injury of personnel in real time, and to promptly warn and inform the medical measures that need to be taken.

[0013] In particular, the step 1 specifically includes:

[0014] Establishing the overall shielding factor inside wearable nuclear radiation protection equipment The calculation model is:

[0015]

[0016] in: The energy of the i-th material is , the incident angle is The time-varying mass attenuation coefficient of the radiation is expressed as:

[0017] ;

[0018] in: It is i The energy of the material is The mass attenuation coefficient of radiation, a Indicates attenuation; It is i The energy of the material is , the incident angle is The mass scattering coefficient of the radiation, s Indicates scattering; r i It is i The reference density of the material; g ( t ) is the time correction factor, which describes the material t The influence of aging and loss; It is i The material is at the position coordinate ( ) time-varying thickness at ; It is i The time-varying density correction function of the material is expressed as: = r i (t) / p i，0 ;in, r i (t) It is i Materials at time t density; r i，0 It is i The reference density of the material.

[0019] In particular, the step 2 specifically includes: randomly sampling the initial energy and incident angle of the radiation particles using the Monte Carlo method according to the energy spectrum and angular distribution of the radiation source;

[0020] For each radiation particle, query the overall shielding coefficient based on the current position coordinates, initial energy and incident angle;

[0021] According to the overall shielding coefficient, the mean free path of the particle in the current medium is calculated, and the actual flight distance is determined by a random number;

[0022] Determine whether the particle undergoes elastic scattering or other interactions with the atomic nuclei in the medium. If so, randomly determine the new energy, scattering angle, and azimuth according to the corresponding cross section and scattering phase function, and update the particle state.

[0023] According to the particle energy loss, the overall shielding coefficient is updated to reflect the time-varying nature of the medium. When the particle energy is lower than the set threshold, the particle process is terminated. The energy deposition in each grid unit is counted to obtain the absorbed dose of the grid unit. According to the set dose threshold, the grid units above the dose threshold are marked as the hot spots of the radiation distribution.

[0024] In particular, the acquisition of accurate time domain data of radiation dose in step 3 specifically includes: selecting a suitable detector according to the expected radiation type and energy range, arranging more detection units in and near the hot spot area of ​​radiation distribution, improving the sampling density, using a standard radiation source to perform energy calibration and efficiency calibration on the detection unit, setting a reasonable time resolution and sampling frequency, and continuously acquiring the count rate or current response of each detection unit at different time points;

[0025] The time domain radiation dose data is stored and used Indicates, where k represents the detection unit number, m represents the time index, and n represents the spatial index.

[0026] In particular, in step 3, converting the time domain data of the radiation dose into frequency domain characteristic data of the radiation dose distribution specifically includes:

[0027] Frequency domain feature extraction is achieved through the following expression, which is used to extract radiation dose data from the time domain Extract frequency domain features from ;

[0028] ;

[0029] in, u , vis a frequency domain variable, corresponding to the frequency in the horizontal and vertical directions, M represents the number of sampling points of the time domain data; N represents the number of sampling points of the spatial data;

[0030] Indicates the kth detection unit in the time domain m and spatial coordinates n Time domain radiation dose data collected at Indicates that the kth detection unit is in the frequency domain ( u , v ) reflects the spectral characteristics of the radiation signal.

[0031] In particular, in step 4, the renal injury prediction model specifically includes:

[0032]

[0033] In the renal injury prediction model, the characteristic energy E, the incident angle θ, the position coordinates (x, y, z) and the frequency domain feature As input, calculate the predicted probability P of renal injury risk; where σ 2 represents the bandwidth parameter of the Gaussian kernel function, which is used to control the similarity between the input and reference features; represents the reference energy characteristic, represents the reference angle feature, ( ) represents the reference position coordinates; Represents the distance between the input feature and the reference feature; represents the coefficients of the Gaussian kernel function; J represents the number of linear characteristic functions, βj Indicates j The coefficients of the characteristic functions reflect the importance of each characteristic function in the entire prediction model; in the characteristic function part, Indicates k Frequency domain characteristics of each detection unit; The skewness of is used to describe the asymmetry of the signal; Indicates k Frequency domain characteristics of a detection unit The kurtosis is used to describe the peak characteristics of the signal.

[0034] In particular, in step 4, real-time assessment of a person's renal injury risk and timely warning specifically include: determining a suitable renal injury risk warning threshold based on actual needs and previous historical data, collecting measurement data in real time, inputting these into the characteristic renal injury prediction model, and calculating the current renal injury risk prediction probability P; if the current renal injury risk prediction probability P is greater than the warning threshold, a risk warning is triggered.

[0035] In particular, in step 4, the current predicted probability P of kidney injury risk is divided into different risk levels; when the risk level changes, the system triggers a warning signal of the corresponding level; the warning signal includes automatic recommendation of corresponding medical measures.

[0036] In particular, the method also includes a training process for the renal injury prediction model, which includes: collecting radiation measurement data samples; recording the actual renal injury degree corresponding to each radiation measurement data sample as a supervisory signal; calculating the high-order statistical moment features of each sample, including skewness and kurtosis; combining these statistical features with physical input features into a complete feature vector; using a gradient descent optimization algorithm to train the parameters of the prediction model; through iterative optimization, learning the optimal model parameters to minimize the loss function between the predicted probability and the actual renal injury degree; and using an independent test set to evaluate the performance of the trained prediction model on new data.

[0037] The present invention also discloses an optimization configuration device for wearable nuclear radiation protection equipment.

[0038] A model building module, used to build a three-dimensional geometric model of the internal radiation shielding layer of the wearable nuclear radiation protection device and calculate the overall shielding coefficient at different positions;

[0039] A hot spot area identification module is used to simulate the propagation process of nuclear radiation inside the wearable nuclear radiation protection device and identify the hot spot area of ​​radiation distribution;

[0040] A radiation detection and conversion module is used to arrange detection units in the hot spot area of ​​radiation distribution to collect accurate time domain data of radiation dose; and convert the time domain data of radiation dose into frequency domain characteristic data of radiation dose distribution;

[0041] The damage assessment and alarm module is used to establish a deep learning kidney injury prediction model based on the frequency domain characteristic data of the radiation dose distribution, and integrate it into the wearable nuclear radiation protection equipment to evaluate the kidney injury risk of personnel in real time and provide timely warnings.

[0042] Beneficial effects:

[0043] The technical solution proposed by the present invention can achieve the following technical effects:

[0044] 1. Improve the accuracy and timeliness of predicting renal damage in personnel under radiation environment.

[0045] The present invention adopts a machine learning method and integrates physical characteristics and statistical characteristics to more accurately predict the degree of radiation damage to the kidneys.

[0046] 2. Compared with the traditional single dose assessment method, the prediction model of the present invention can better reflect the complex propagation and deposition process of radiation in the human body.

[0047] 3. Realize real-time monitoring and graded early warning of people’s kidney injury risk.

[0048] The prediction model of the present invention can continuously evaluate the radiation risk of the environment in which personnel are located and classify it into different levels.

[0049] 4. When the risk level changes, the system will promptly issue warning signals of the corresponding level to help take targeted medical measures.

[0050] 5. Improve the pertinence and response speed of radiation safety protection.

[0051] According to different risk levels, the system of the present invention will automatically recommend appropriate protection suggestions, such as adjusting the work process, changing protective equipment, etc.

[0052] These graded medical measures can help on-site personnel take protective actions more promptly and effectively to reduce the occurrence of kidney damage.

[0053] 6. Provide intelligent decision-making support for radiation safety management and control.

[0054] The prediction model of the present invention can make full use of various radiation monitoring data and realize automated risk analysis and early warning through machine learning.

[0055] This intelligent early warning system can provide a reliable basis for command decisions and improve the overall level of radiation safety management and control. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 A schematic flow chart of the optimization configuration method of the wearable nuclear radiation protection device proposed by the present invention;

[0057] Figure 2 A schematic diagram of an optimized configuration device for a wearable nuclear radiation protection device proposed in the present invention. DETAILED DESCRIPTION

[0058] The present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0059] The present invention provides a method for optimizing the configuration of a wearable nuclear radiation protection device, such as Figure 1 As shown, the nuclear radiation protection equipment includes a radiation shielding layer and a detection device, and the method includes:

[0060] Step 1, establishing a three-dimensional geometric model of the internal radiation shielding layer of the wearable nuclear radiation protection device, and calculating the overall shielding coefficient at different positions;

[0061] Establishing the overall shielding factor inside wearable nuclear radiation protection equipment The calculation model is:

[0062]

[0063] in: is the energy of the i-th material , the incident angle is The time-varying mass attenuation coefficient of the radiation is expressed as:

[0064] ;

[0065] in: It is i The energy of the material is The mass attenuation coefficient of radiation, a Indicates attenuation; It is i The energy of the material is , the incident angle is The mass scattering coefficient of the radiation, s Indicates scattering; r i It is i The reference density of the material; g ( t ) is the time correction factor, which describes the material t The influence of aging and loss; It is i The material is at the position coordinate ( ) at a time-varying thickness; It is i The time-varying density correction function of the material is expressed as: = r i (t) / p i，0 ;in, r i (t) It is i Materials at time t density; r i，0 It is i The reference density of the material.

[0066] Step 2, simulate the propagation process of nuclear radiation inside the wearable nuclear radiation protection device, and identify the hot spots of radiation distribution; in the internal space of the wearable nuclear radiation protection device, divide the entire space into small grid units. Each grid unit has its own coordinate position, which is used to locate the specific position of the radiation particles inside the device. In the process of simulating radiation propagation, the radiation energy deposition in each grid unit will be counted and analyzed. According to the set dose threshold, the grid unit whose absorbed dose exceeds the threshold is marked as a radiation hot spot area. Step 2 specifically includes: according to the energy spectrum and angular distribution of the radiation source, the Monte Carlo method is used to randomly sample the initial energy E and incident angle θ of the radiation particles; the radiation particles can reach the order of millions.

[0067] For each radiation particle, the overall shielding coefficient is queried according to the current position coordinates, particle energy and incident angle; the mean free path of the particle in the current medium is calculated according to the shielding coefficient, and the actual flight distance is determined by random numbers; it is determined whether the particle undergoes elastic scattering or other interactions with the atomic nuclei in the medium. If so, the new energy, scattering angle and azimuth are randomly determined according to the corresponding cross-section and scattering phase function, and the particle state is updated.

[0068] According to the particle energy loss, the overall shielding coefficient is updated to reflect the time-varying nature of the medium. When the particle energy is lower than the set threshold, the particle process is terminated. The energy deposition in each grid unit is counted to obtain the absorbed dose of the unit. According to the set dose threshold, the grid units above the threshold are marked as radiation hotspots.

[0069] Step 3, arrange detection units in the radiation hotspot area to collect accurate time domain data of radiation dose; convert the time domain data of radiation dose into frequency domain characteristic data of radiation dose distribution; the collection of accurate time domain data of radiation dose in step 3 specifically includes: selecting a suitable detector according to the expected radiation type and energy range, arranging more detection units in the radiation hotspot area and its vicinity, improving the sampling density, using a standard radiation source to calibrate the energy and efficiency of the detection units, setting a reasonable time resolution and sampling frequency, and continuously collecting the count rate or current response of each detection unit at different time points.

[0070] The time domain radiation dose data is stored and used Indicates, where k represents the detection unit number, m Time index, n Represents a spatial index.

[0071] The selection of detectors can be based on the expected radiation type (gamma rays, neutrons, charged particles, etc.) and energy range. Commonly used detectors include: scintillation detectors (NaI (Tl), BGO, CsI, etc.); gas-filled detectors (ionization chambers, GM counters, etc.); semiconductor detectors (HPGe, Si (Li), etc.). Detector layout: more detection units are arranged in and near the hot spots of radiation distribution to increase the sampling density. For example, one detector can be arranged every 30° in the range of 10cm-1m along different azimuth angles, for a total of 12 detectors. For surface sources, the detector array can be arranged in a 10cm*10cm grid 10cm above the surface source.

[0072] Radiation calibration uses a standard radiation source (such as 137Cs, 60Co, Am-Be, etc.) to perform energy calibration and efficiency calibration on each detection unit; energy calibration: establish a calibration curve of energy and pulse amplitude; efficiency calibration: calibrate the counting efficiency at different energy points; time resolution and sampling frequency setting, for time resolution: 10ns~1μs, to capture instantaneous radiation pulse signals; sampling frequency: typically 100MHz~1GHz to meet the Nyquist sampling theorem; time domain data acquisition: for each detection unit, continuously collect the counting rate or current response at different time points to obtain a time domain data sequence.

[0073] For example, 12 NaI (Tl) detectors are used to surround the radiation source, with a time resolution of 100ns and a sampling frequency of 200MHz; a 2*2 detector array is used to scan the surface source, with a time resolution of 1μs and a sampling frequency of 500MHz; in the coupled reactor, 20 slit-type ion chambers are arranged, with a time resolution of 10ns and a sampling frequency of 1GHz; through the above process, we can obtain accurate time domain sampling data, reflecting the change of radiation dose over time at different detection points.

[0074] In step 3, converting the time domain data of the radiation dose into frequency domain characteristic data of the radiation dose distribution specifically includes:

[0075] Frequency domain feature extraction is achieved through the following expression, which is used to extract radiation dose data from the time domain Extract frequency domain features from ;

[0076] ;

[0077] in, u , v is a frequency domain variable, corresponding to the frequency in the horizontal and vertical directions, M represents the number of sampling points of the time domain data; N represents the number of sampling points of the spatial data;

[0078] Indicates the kth detection unit in the time domain m and spatial coordinates n Time domain radiation dose data collected at Indicates that the kth detection unit is in the frequency domain ( u , v ) reflects the spectral characteristics of the radiation signal.

[0079] Step 4: Establish the frequency domain characteristics of the radiation dose distribution and the renal injury prediction model based on deep learning, and integrate them into the wearable nuclear radiation protection equipment to evaluate the renal injury risk of personnel in real time and provide timely warning.

[0080] In step 4, the renal injury prediction model specifically includes:

[0081]

[0082] In the renal injury prediction model, the characteristic energy E, the incident angle θ, the position coordinates (x, y, z) and the frequency domain feature As input, calculate the predicted probability P of renal injury risk; where σ 2 represents the bandwidth parameter of the Gaussian kernel function, which is used to control the similarity between the input and reference features; represents the reference energy characteristic, represents the reference angle feature, ( ) represents the reference position coordinates; Represents the distance between the input feature and the reference feature; Represents the coefficients of the Gaussian kernel function; J represents the number of linear characteristic functions, βj Indicates j The coefficients of the characteristic functions reflect the importance of each characteristic function in the entire prediction model; in the characteristic function part, Indicates k Frequency domain characteristics of each detection unit; The skewness of is used to describe the asymmetry of the signal; Indicates k Frequency domain characteristics of a detection unit The kurtosis is used to describe the peak characteristics of the signal.

[0083] In this embodiment, the optimal model parameters can be learned through machine learning methods. , σ 2 and βj; that is, Frequency domain features are used as features of radiation dose signals and added to the probability model based on Gaussian kernel function and linear combination to jointly predict the risk of renal injury. This method of combining time domain and frequency domain features can better capture the complex characteristics of radiation signals, thereby improving the accuracy of renal injury prediction.

[0084] In step 4, real-time assessment of the risk of kidney injury of personnel and timely warning include: determining a suitable kidney injury risk warning threshold according to actual needs and previous historical data, and real-time collection of measurement data, including radiation energy E, incident angle θ, position coordinates (x, y, z) and frequency domain characteristics ; Input these features into the renal injury prediction model to calculate the current renal injury risk prediction probability P; if the prediction probability P is greater than the warning threshold, a risk warning is triggered.

[0085] In step 4, the predicted renal injury risk probability P is divided into different risk levels; the predicted renal injury risk probability P is divided into different risk levels, and when the risk level changes, the system triggers a warning signal of the corresponding level; the warning signal includes automatically recommending corresponding medical measures. For example, when the risk level changes, the system triggers a warning signal of the corresponding level.

[0086] A suggestive warning will be given at a low risk level; an early warning notice will be issued at a medium risk level; and an emergency warning will be issued immediately at a high risk level.

[0087] For example, three risk levels can be set: low risk (P<0.4), medium risk (0.4≤P<0.7), and high risk (P≥0.7). Each level corresponds to different early warning measures and emergency responses.

[0088] Early warning information can be conveyed to relevant personnel in real time through display screens, mobile phone push notifications, etc. According to different risk levels, the system will automatically recommend corresponding medical measures.

[0089] For example, when the risk is low, it can be recommended to strengthen the wearing of protective equipment; when the risk is medium, the work process can be adjusted; when the risk is high, immediate evacuation can be required. In addition, when the risk is low, the medical measures in the early warning include regular physical examinations, decontamination of radioactive materials, nutritional supplements, etc.; when the risk is medium, the medical measures in the early warning include: taking special medicines, receiving short-term hospitalization, etc., the purpose is to protect key organs, alleviate acute radiation symptoms, reduce the radiation level in the body as soon as possible, and prevent the degree of damage from worsening; when the risk is high, the medical measures in the early warning include long-term hospitalization and emergency first aid measures, the purpose is to save lives and minimize radiation damage.

[0090] These protective recommendations can help people take effective protective actions in a timely manner to reduce the risk of kidney damage. Based on the above prediction model, we can also:

[0091] Real-time risk assessment: The system monitors the radiation situation of the personnel's environment in real time and substitutes it into the prediction model to continuously assess the risk of kidney damage. Based on the risk probability P calculated in real time, the system dynamically determines the risk level of the current personnel.

[0092] The training process of the renal injury prediction model includes: collecting radiation measurement data samples; recording the actual renal injury degree corresponding to each radiation measurement data sample as a supervisory signal; calculating the high-order statistical moment features of each sample, including skewness and kurtosis; combining these statistical features with physical input features into a complete feature vector; using a gradient descent optimization algorithm to train the parameters of the prediction model; through iterative optimization, learning the optimal model parameters to minimize the loss function between the predicted probability and the actual renal injury degree; and using an independent test set to evaluate the performance of the trained prediction model on new data.

[0093] Optionally, the wearable nuclear radiation protection device in this embodiment can be a vest. As a close-fitting garment, the vest can better fit the human body and achieve radiation protection for the entire torso area.

[0094] Multiple layers of radiation shielding material can be designed inside the vest, and the shielding coefficients at different positions can be calculated based on the three-dimensional geometric model to optimize the protection effect.

[0095] Placing radiation detection units in hot spots inside the vest can more accurately collect time-domain data on the radiation dose received by the human body.

[0096] Integrate the deep learning-based kidney injury prediction model into the vest device to achieve real-time risk assessment and early warning for the wearer.

[0097] As a piece of protective equipment worn close to the body, the vest can ensure that early warning information is transmitted to the wearer in a timely manner, thereby improving the efficiency of emergency response.

[0098] This vest-based wearable nuclear radiation protection device fully utilizes the advantages of close-fitting clothing, and while achieving reliable radiation protection, it also provides intelligent safety warning functions. This is of great significance for protecting workers in high-radiation environments such as battlefields or nuclear power plants.

[0099] In general, applying the technical solution of the present invention to wearable devices such as vests can not only improve the protective performance, but also enhance the convenience of real-time monitoring and early warning, which is a very promising implementation method.

[0100] The present invention also discloses an optimized configuration device for wearable nuclear radiation protection equipment, such as Figure 2 As shown, including:

[0101] A model building module is used to establish a three-dimensional geometric model of the radiation shielding layer inside the wearable nuclear radiation protection device and calculate the shielding coefficients at different positions; the model building module specifically includes:

[0102] Establishing the overall shielding factor inside wearable nuclear radiation protection equipment The calculation model is:

[0103]

[0104] in: is the energy of the i-th material , the incident angle is The time-varying mass attenuation coefficient of the radiation is expressed as:

[0105] ;

[0106] in: It is i The energy of the material is The mass attenuation coefficient of radiation, a Indicates attenuation; It is i The energy of the material is , the incident angle is The mass scattering coefficient of the radiation, s Indicates scattering; r i It is i The reference density of the material; g ( t ) is the time correction factor, which describes the material t The influence of aging and loss; It is i The material is at the position coordinate ( ) at a time-varying thickness; It is i The time-varying density correction function of the material is expressed as: = r i (t) / p i，0 ;in, r i (t) It is i Materials at time t density; r i，0 It is i The reference density of the material.

[0107] The hot spot area identification module is used to simulate the propagation process of nuclear radiation inside the wearable nuclear radiation protection device and identify the hot spot area of ​​radiation distribution; in the internal space of the wearable nuclear radiation protection device, the entire space is divided into small grid units. Each grid unit has its own coordinate position for locating the specific position of the radiation particles inside the device. In the process of simulating radiation propagation, the radiation energy deposition in each grid unit will be counted and analyzed. According to the set dose threshold, the grid unit whose absorbed dose exceeds the threshold is marked as a radiation hot spot area. The hot spot area identification module specifically includes: according to the energy spectrum and angular distribution of the radiation source, the Monte Carlo method is used to randomly sample the initial energy and incident angle of the radiation particles; the radiation particles can reach the order of millions.

[0108] For each radiation particle, the overall shielding coefficient is queried according to the current position coordinates, particle energy and incident angle; based on the shielding coefficient, the mean free path of the particle in the current medium is calculated, and the actual flight distance is determined using a random number.

[0109] Determine whether the particle undergoes elastic scattering or other interactions with the atomic nuclei in the medium. If so, randomly determine the new energy, scattering angle, and azimuth according to the corresponding cross section and scattering phase function, and update the particle state.

[0110] According to the particle energy loss, the overall shielding coefficient is updated to reflect the time-varying nature of the medium. When the particle energy is lower than the set threshold, the particle process is terminated. The energy deposition in each grid unit is counted to obtain the absorbed dose of the unit. According to the set dose threshold, the grid units above the threshold are marked as radiation hotspots.

[0111] The radiation detection and conversion module is used to arrange detection units in the radiation hotspot area and collect accurate time domain data of radiation dose; convert the time domain data of radiation dose into frequency domain characteristic data of radiation dose distribution; the collection of accurate time domain data of radiation dose in the radiation detection and conversion module specifically includes: selecting a suitable detector according to the expected radiation type and energy range, arranging more detection units in the radiation hotspot area and its vicinity, improving the sampling density, using a standard radiation source to calibrate the energy and efficiency of the detection units, setting a reasonable time resolution and sampling frequency, and continuously collecting the count rate or current response of each detection unit at different time points.

[0112] The selection of detectors can be based on the expected radiation type (gamma rays, neutrons, charged particles, etc.) and energy range, and the appropriate detector can be selected. Commonly used detectors include: scintillation detectors (NaI (Tl), BGO, CsI, etc.); gas-filled detectors (ionization chambers, GM counters, etc.); semiconductor detectors (HPGe, Si (Li), etc.).

[0113] Detector arrangement is to arrange more detection units in and near the hot spot area of ​​radiation distribution to increase the sampling density. For example, one detector can be arranged every 30° in the range of 10cm-1m along different azimuths, for a total of 12. For surface sources, the detector array can be arranged in a 10cm*10cm grid at 10cm above the surface source.

[0114] Radiation calibration uses a standard radiation source (such as 137Cs, 60Co, Am-Be, etc.) to perform energy calibration and efficiency calibration on each detection unit; energy calibration: establish a calibration curve of energy and pulse amplitude; efficiency calibration: calibrate the counting efficiency at different energy points; time resolution and sampling frequency setting, for time resolution: 10ns~1μs, to capture instantaneous radiation pulse signals; sampling frequency: typically 100MHz~1GHz to meet the Nyquist sampling theorem; time domain data acquisition: for each detection unit, continuously collect the counting rate or current response at different time points to obtain a time domain data sequence.

[0115] For example, 12 NaI (Tl) detectors are used to surround the radiation source, with a time resolution of 100ns and a sampling frequency of 200MHz; a 2*2 detector array is used to scan the surface source, with a time resolution of 1μs and a sampling frequency of 500MHz; in the coupled reactor, 20 slit-type ion chambers are arranged, with a time resolution of 10ns and a sampling frequency of 1GHz; through the above process, we can obtain accurate time domain sampling data, reflecting the change of radiation dose over time at different detection points.

[0116] The time domain radiation dose data is stored and used Indicates, where k represents the detection unit number, m represents the time index, and n represents the spatial index.

[0117] In the radiation detection and conversion module, converting the time domain data of the radiation dose into frequency domain characteristic data of the radiation dose distribution specifically includes:

[0118] Frequency domain feature extraction is achieved through the following expression, which is used to extract radiation dose data from the time domain Extract frequency domain features from ;

[0119] ;

[0120] in, u , v is a frequency domain variable, corresponding to the frequency in the horizontal and vertical directions, M represents the number of sampling points of the time domain data; N represents the number of sampling points of the spatial data;

[0121] Indicates the kth detection unit in the time domainm and spatial coordinates n Time domain radiation dose data collected at Indicates that the kth detection unit is in the frequency domain ( u , v ) reflects the spectral characteristics of the radiation signal.

[0122] The injury assessment and alarm module is used to establish the frequency domain characteristics of the radiation dose distribution and the renal injury prediction model based on deep learning, and integrate it into the wearable nuclear radiation protection device to evaluate the renal injury risk of personnel in real time and issue a timely warning. In the injury assessment and alarm module, the renal injury prediction model specifically includes:

[0123]

[0124] In the renal injury prediction model, the characteristic energy E, the incident angle θ, the position coordinates (x, y, z) and the frequency domain feature As input, calculate the predicted probability P of renal injury risk; where σ 2 represents the bandwidth parameter of the Gaussian kernel function, which is used to control the similarity between the input and reference features; represents the reference energy characteristic, represents the reference angle feature, ( ) represents the reference position coordinates; Represents the distance between the input feature and the reference feature; represents the coefficients of the Gaussian kernel function; J represents the number of linear characteristic functions, βj Indicates j The coefficients of the characteristic functions reflect the importance of each characteristic function in the entire prediction model; in the characteristic function part, Indicates k Frequency domain characteristics of each detection unit; The skewness of is used to describe the asymmetry of the signal; Indicates k Frequency domain characteristics of a detection unit The kurtosis is used to describe the peak characteristics of the signal.

[0125] In this embodiment, the optimal model parameters can be learned through machine learning methods. , σ 2 and βj; that is, Frequency domain features are used as features of radiation dose signals and added to the probability model based on Gaussian kernel function and linear combination to jointly predict the risk of renal injury. This method of combining time domain and frequency domain features can better capture the complex characteristics of radiation signals, thereby improving the accuracy of renal injury prediction.

[0126] In the injury assessment and alarm module, the risk of renal injury of personnel is assessed in real time and timely warning is provided, specifically including: determining a suitable threshold of renal injury risk warning according to actual needs and previous historical data, and collecting measurement data in real time, including radiation energy E, incident angle θ, position coordinates (x, y, z) and frequency domain characteristics ; Input these features into the renal injury prediction model to calculate the current renal injury risk prediction probability P; if the prediction probability P is greater than the warning threshold, a risk warning is triggered.

[0127] In the injury assessment and alarm module, the predicted renal injury risk probability P is divided into different risk levels; the predicted renal injury risk probability P is divided into different risk levels, and when the risk level changes, the system triggers a warning signal of the corresponding level; the warning signal includes automatic recommendation of corresponding medical measures. For example, when the risk level changes, the system triggers a warning signal of the corresponding level.

[0128] A suggestive warning will be given at a low risk level; an early warning notice will be issued at a medium risk level; and an emergency warning will be issued immediately at a high risk level.

[0129] For example, three risk levels can be set: low risk (P<0.4), medium risk (0.4≤P<0.7), and high risk (P≥0.7). Each level corresponds to different early warning measures and emergency responses.

[0130] Early warning information can be conveyed to relevant personnel in real time through display screens, mobile phone push notifications, etc. According to different risk levels, the system will automatically recommend corresponding medical measures.

[0131] For example, when the risk is low, it can be recommended to strengthen the wearing of protective equipment; when the risk is medium, the work process can be adjusted; when the risk is high, immediate evacuation can be required. In addition, when the risk is low, the medical measures in the early warning include regular physical examinations, decontamination of radioactive materials, nutritional supplements, etc.; when the risk is medium, the medical measures in the early warning include: taking special medicines, receiving short-term hospitalization, etc., the purpose is to protect key organs, alleviate acute radiation symptoms, reduce the radiation level in the body as soon as possible, and prevent the degree of damage from worsening; when the risk is high, the medical measures in the early warning include long-term hospitalization and emergency first aid measures, the purpose is to save lives and minimize radiation damage.

[0132] These protective recommendations can help people take effective protective actions in a timely manner to reduce the risk of kidney damage. Based on the above prediction model, we can also:

[0133] Real-time risk assessment: The system monitors the radiation situation of the personnel's environment in real time and substitutes it into the prediction model to continuously assess the risk of kidney damage. Based on the risk probability P calculated in real time, the system dynamically determines the risk level of the current personnel.

[0134] The training process of the renal injury prediction model includes: collecting radiation measurement data samples; recording the actual renal injury degree corresponding to each radiation measurement data sample as a supervisory signal; calculating the high-order statistical moment features of each sample, including skewness and kurtosis; combining these statistical features with physical input features into a complete feature vector; using a gradient descent optimization algorithm to train the parameters of the prediction model; through iterative optimization, learning the optimal model parameters to minimize the loss function between the predicted probability and the actual renal injury degree; and using an independent test set to evaluate the performance of the trained prediction model on new data.

[0135] Optionally, the wearable nuclear radiation protection device in this embodiment can be a vest. As a close-fitting garment, the vest can better fit the human body and achieve radiation protection for the entire torso area.

[0136] Multiple layers of radiation shielding material can be designed inside the vest, and the shielding coefficients at different positions can be calculated based on the three-dimensional geometric model to optimize the protection effect.

[0137] Placing radiation detection units in hot spots inside the vest can more accurately collect time-domain data on the radiation dose received by the human body.

[0138] Integrate the deep learning-based kidney injury prediction model into the vest device to achieve real-time risk assessment and early warning for the wearer.

[0139] As a piece of protective equipment worn close to the body, the vest can ensure that early warning information is transmitted to the wearer in a timely manner, thereby improving the efficiency of emergency response.

[0140] This vest-based wearable nuclear radiation protection device fully utilizes the advantages of close-fitting clothing, and while achieving reliable radiation protection, it also provides intelligent safety warning functions. This is of great significance for protecting workers in high-radiation environments such as battlefields or nuclear power plants.

[0141] In general, applying the technical solution of the present invention to wearable devices such as vests can not only improve the protective performance, but also enhance the convenience of real-time monitoring and early warning, which is a very promising implementation method.

[0142] In summary, the above are only preferred embodiments of the present invention and are not intended to limit the protection scope of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

[0143] It is obvious to those skilled in the art that the embodiments of the present invention are not limited to the details of the above exemplary embodiments, and that the embodiments of the present invention can be implemented in other specific forms without departing from the spirit or basic features of the embodiments of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-restrictive, and the scope of the embodiments of the present invention is limited by the attached claims rather than the above description, so it is intended to include all changes that fall within the meaning and scope of the equivalent elements of the claims in the embodiments of the present invention. Any figure mark in the claims should not be regarded as limiting the claims involved. In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units, modules or devices stated in the system, device or terminal claims can also be implemented by the same unit, module or device through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any particular order.

[0144] Finally, it should be noted that the above implementation modes are only used to illustrate the technical solutions of the embodiments of the present invention and are not intended to limit them. Although the embodiments of the present invention have been described in detail with reference to the above preferred implementation modes, those skilled in the art should understand that the technical solutions of the embodiments of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for optimizing the configuration of a wearable nuclear radiation protection device, characterized in that: Step 1, establishing a three-dimensional geometric model of the internal radiation shielding layer of the wearable nuclear radiation protection device, and calculating the overall shielding coefficient at different positions; Step 2, simulating the propagation process of nuclear radiation inside the wearable nuclear radiation protection device to identify the hot spots of radiation distribution; Step 3, arranging detection units in the hot spot area of ​​radiation distribution to collect accurate time domain data of radiation dose; converting the time domain data of radiation dose into frequency domain characteristic data of radiation dose distribution; Step 4: establishing a deep learning-based kidney injury prediction model based on the frequency domain characteristic data of the radiation dose distribution, and integrating the model into the wearable nuclear radiation protection device to assess the risk of kidney injury of personnel in real time, and to promptly warn and inform of the medical measures that need to be taken; In step 4, the renal injury prediction model specifically includes: In the renal injury prediction model, the characteristic energy E, the incident angle θ, the position coordinates (x, y, z) and the frequency domain feature As input, calculate the predicted probability P of renal injury risk; where σ 2 represents the bandwidth parameter of the Gaussian kernel function, which is used to control the similarity between the input and reference features; represents the reference energy characteristic, represents the reference angle feature, ( ) represents the reference position coordinates; Represents the distance between the input feature and the reference feature; Represents the coefficients of the Gaussian kernel function; J represents the number of linear characteristic functions, β j Indicates j The coefficients of the characteristic functions reflect the importance of each characteristic function in the entire prediction model; in the characteristic function part, Indicates k Frequency domain characteristics of a detection unit The skewness of is used to describe the asymmetry of the signal; Indicates k Frequency domain characteristics of a detection unit The kurtosis is used to describe the peak characteristics of the signal.

2. The optimization configuration method of the wearable nuclear radiation protection device according to claim 1, characterized in that: The step 1 specifically includes: Establishing the overall shielding factor inside wearable nuclear radiation protection equipment The calculation model is: in: The energy of the i-th material is , the incident angle is The time-varying mass attenuation coefficient of the radiation is expressed as: ； in: It is i The energy of the material is The mass attenuation coefficient of radiation, a Indicates attenuation; It is i The energy of the material is , the incident angle is The mass scattering coefficient of the radiation, s Indicates scattering; ρ i It is i The reference density of the material; g ( t ) is the time correction factor, which describes the change of material over time. t The influence of aging and loss; It is i The material is at the position coordinate ( ) at a time-varying thickness; It is i The time-varying density correction function of the material is expressed as: = ρ i (t) / ρ i，0 ;in, ρ i (t) It is i Materials at time t density; ρ i，0 It is i The reference density of the material.

3. The optimization configuration method of the wearable nuclear radiation protection device according to claim 2, characterized in that: The step 2 specifically includes: randomly sampling the initial energy and incident angle of the radiation particles using the Monte Carlo method according to the energy spectrum and angular distribution of the radiation source; For each radiation particle, query the overall shielding coefficient based on the current position coordinates, initial energy and incident angle; According to the overall shielding coefficient, the mean free path of the particle in the current medium is calculated, and the actual flight distance is determined by a random number; Determine whether the particle undergoes elastic scattering or other interactions with the atomic nuclei in the medium. If so, randomly determine the new energy, scattering angle, and azimuth according to the corresponding cross section and scattering phase function, and update the particle state. According to the particle energy loss, the overall shielding coefficient is updated to reflect the time-varying nature of the medium. When the particle energy is lower than the set threshold, the particle process is terminated. The energy deposition in each grid unit is counted to obtain the absorbed dose of the grid unit. According to the set dose threshold, the grid units above the dose threshold are marked as the hot spots of the radiation distribution.

4. The method for optimizing the configuration of the wearable nuclear radiation protection device according to claim 3, characterized in that: The step 3 of collecting accurate radiation dose time domain data specifically includes: selecting a detector according to the expected radiation type and energy range, arranging more detection units in and near the hot spot area of ​​radiation distribution, improving the sampling density, using a standard radiation source to perform energy calibration and efficiency calibration on the detection unit, setting the time resolution and sampling frequency, and continuously collecting the count rate or current response of each detection unit at different time points; The time domain radiation dose data is stored and used Indicates that k Indicates the detection unit number, m Time index, n Represents a spatial index.

5. The optimization configuration method of the wearable nuclear radiation protection device according to claim 4, characterized in that: In step 3, converting the time domain data of the radiation dose into frequency domain characteristic data of the radiation dose distribution specifically includes: Frequency domain feature extraction is achieved through the following expression, which is used to extract radiation dose data from the time domain Extract frequency domain features from ; ； in, u , v is a frequency domain variable, corresponding to the frequency in the horizontal and vertical directions, M represents the number of sampling points of the time domain data; N represents the number of sampling points of the spatial data; Indicates the kth detection unit in the time domain m and spatial coordinates n Time domain radiation dose data collected at Indicates that the kth detection unit is in the frequency domain ( u , v ) reflects the spectral characteristics of the radiation signal.

6. The optimization configuration method of the wearable nuclear radiation protection device according to claim 1, characterized in that: In step 4, real-time assessment of the risk of renal injury of personnel and timely warning specifically include: determining a renal injury risk warning threshold according to actual needs and previous historical data, collecting measurement data in real time, inputting these into the characteristic renal injury prediction model, and calculating the current renal injury risk prediction probability P; if the current renal injury risk prediction probability P is greater than the warning threshold, a risk warning is triggered.

7. The optimization configuration method of the wearable nuclear radiation protection device according to claim 6, characterized in that: In step 4, the current predicted probability P of kidney injury risk is divided into different risk levels; when the risk level changes, the system triggers a warning signal of the corresponding level; the warning signal includes automatic recommendation of corresponding medical measures.

8. The optimization configuration method of the wearable nuclear radiation protection device according to claim 7, characterized in that: The method also includes a training process for the renal injury prediction model, which includes: collecting radiation measurement data samples; recording the actual renal injury degree corresponding to each radiation measurement data sample as a supervisory signal; calculating the high-order statistical moment features of each sample, including skewness and kurtosis; combining these statistical features with physical input features into a complete feature vector; using a gradient descent optimization algorithm to train the parameters of the prediction model; through iterative optimization, learning the optimal model parameters to minimize the loss function between the predicted probability and the actual renal injury degree; and using an independent test set to evaluate the performance of the trained prediction model on new data.

9. An optimized configuration device for wearable nuclear radiation protection equipment, characterized in that: A model building module, used to build a three-dimensional geometric model of the internal radiation shielding layer of the wearable nuclear radiation protection device and calculate the overall shielding coefficient at different positions; A hot spot area identification module is used to simulate the propagation process of nuclear radiation inside the wearable nuclear radiation protection device and identify the hot spot area of ​​radiation distribution; A radiation detection and conversion module is used to arrange detection units in the hot spot area of ​​radiation distribution to collect accurate time domain data of radiation dose; and convert the time domain data of radiation dose into frequency domain characteristic data of radiation dose distribution; An injury assessment and alarm module is used to establish a deep learning-based kidney injury prediction model based on the frequency domain characteristic data of the radiation dose distribution, and integrate it into the wearable nuclear radiation protection device to assess the risk of kidney injury of personnel in real time, and promptly warn and inform the medical measures that need to be taken; The renal injury prediction model specifically includes: In the renal injury prediction model, the characteristic energy E, the incident angle θ, the position coordinates (x, y, z) and the frequency domain feature As input, calculate the predicted probability P of renal injury risk; where σ 2 represents the bandwidth parameter of the Gaussian kernel function, which is used to control the similarity between the input and reference features; represents the reference energy characteristic, represents the reference angle feature, ( ) represents the reference position coordinates; Represents the distance between the input feature and the reference feature; Represents the coefficients of the Gaussian kernel function; J represents the number of linear characteristic functions, β j Indicates j The coefficients of the characteristic functions reflect the importance of each characteristic function in the entire prediction model; in the characteristic function part, Indicates k Frequency domain characteristics of a detection unit The skewness of is used to describe the asymmetry of the signal; Indicates k Frequency domain characteristics of a detection unit The kurtosis is used to describe the peak characteristics of the signal.

Images