Radiation environment online monitoring method and system based on internet of things

Abstract

This invention provides an online radiation environment monitoring method and system based on the Internet of Things (IoT), belonging to the field of IoT monitoring technology. The method includes: taking the space where the radiation device is located as the first radiation environment space and its neighboring space as the second radiation environment space; conducting online monitoring of the two spaces through an IoT device cluster to obtain first and second radiation environment monitoring sets; performing monitoring reliability correction on the two monitoring sets according to the IoT device cluster to construct first and second environment monitoring sequences; predicting the corresponding spaces based on the first and second environment monitoring sequences to obtain first and second radiation safety accident characteristics; and performing coupled prediction of radiation safety accidents based on the two types of accident characteristics to establish a radiation safety accident map. This invention solves the technical problem of low accuracy in online radiation environment monitoring in existing technologies, which leads to poor effectiveness of radiation safety management, and achieves the technical effect of improving the accuracy of online radiation environment monitoring and improving the effectiveness and quality of radiation safety management.

Description

Technical Field

[0001] This invention belongs to the field of Internet of Things (IoT) monitoring technology, specifically relating to an online monitoring method and system for radiation environment based on IoT. Background Technology

[0002] Online radiation environment monitoring is a key technology for ensuring the safe operation of radiation equipment in fields such as medicine. Although multi-device collaborative monitoring of the radiation environment has been achieved, there are still technical shortcomings: First, existing monitoring is mostly limited to the core space where the radiation equipment is located, failing to include the surrounding space and ignoring the easy diffusion characteristics of radiation, leading to the omission of cross-space safety risks. Second, the raw monitoring data collected by IoT device clusters is easily affected by fluctuations in device sensor parameters, and the lack of an effective and reliable correction mechanism results in insufficient data reliability. Furthermore, current radiation safety accident predictions largely rely on historical data from a single space, resulting in limited prediction accuracy.

[0003] Existing technologies suffer from low accuracy in online radiation environment monitoring, leading to poor effectiveness in radiation safety management. Summary of the Invention

[0004] This invention provides an Internet of Things-based online radiation environment monitoring method and system, which solves the technical problem of low accuracy in existing online radiation environment monitoring, leading to poor effectiveness of radiation safety management. It achieves the technical effect of improving the accuracy of online radiation environment monitoring and enhancing the effectiveness and quality of radiation safety management.

[0005] In view of the above problems, the present invention adopts the following technical solution:

[0006] In a first aspect, the present invention provides an online radiation environment monitoring method based on the Internet of Things (IoT). The method includes: taking the space where the radiation device is located as a first radiation environment space and the neighborhood space of the first radiation environment space as a second radiation environment space; performing online monitoring of the first radiation environment space and the second radiation environment space through an IoT device cluster to obtain a first radiation environment monitoring set and a second radiation environment monitoring set; performing monitoring reliability correction on the first radiation environment monitoring set and the second radiation environment monitoring set according to the IoT device cluster to construct a first environmental monitoring sequence and a second environmental monitoring sequence; performing radiation safety accident prediction on the first radiation environment space according to the first environmental monitoring sequence to obtain a first radiation safety accident feature; performing radiation safety accident prediction on the second radiation environment space according to the second environmental monitoring sequence to obtain a second radiation safety accident feature; and performing coupled radiation safety accident prediction based on the first radiation safety accident feature and the second radiation safety accident feature to establish a radiation safety accident map.

[0007] On the other hand, the present invention also provides an Internet of Things (IoT)-based online radiation environment monitoring system, the system comprising: a radiation environment space construction module, used to define the space where the radiation equipment is located as a first radiation environment space and the neighborhood space of the first radiation environment space as a second radiation environment space; an IoT monitoring data acquisition module, used to perform online monitoring of the first radiation environment space and the second radiation environment space through an IoT device group to obtain a first radiation environment monitoring set and a second radiation environment monitoring set; a monitoring reliability correction module, used to perform monitoring reliability correction on the first radiation environment monitoring set and the second radiation environment monitoring set according to the IoT device group, respectively, to construct a first environmental monitoring sequence and a second environmental monitoring sequence; a first radiation safety accident prediction module, used to predict radiation safety accidents in the first radiation environment space according to the first environmental monitoring sequence to obtain a first radiation safety accident feature; a second radiation safety accident prediction module, used to predict radiation safety accidents in the second radiation environment space according to the second environmental monitoring sequence to obtain a second radiation safety accident feature; and a coupling prediction and map construction module, used to perform coupled prediction of radiation safety accidents according to the first radiation safety accident feature and the second radiation safety accident feature to establish a radiation safety accident map.

[0008] One or more technical solutions provided in this invention have at least the following technical effects or advantages:

[0009] Using the space where the radiation equipment is located as the first radiation environment space, and the neighboring space of the first radiation environment space as the second radiation environment space; the first and second radiation environment spaces are monitored online through an Internet of Things (IoT) device cluster to obtain a first radiation environment monitoring set and a second radiation environment monitoring set; the first and second radiation environment monitoring sets are respectively subjected to monitoring reliability correction based on the IoT device cluster to construct a first environmental monitoring sequence and a second environmental monitoring sequence; radiation safety accidents are predicted in the first radiation environment space based on the first environmental monitoring sequence to obtain a first radiation safety accident feature; radiation safety accidents are predicted in the second radiation environment space based on the second environmental monitoring sequence to obtain a second radiation safety accident feature; radiation safety accidents are coupled and predicted based on the first and second radiation safety accident features to establish a radiation safety accident map; this achieves the technical effect of improving the accuracy of online radiation environment monitoring and enhancing the effectiveness and quality of radiation safety management. Attached Figure Description

[0010] Figure 1 This is a flowchart illustrating the online radiation environment monitoring method based on the Internet of Things of the present invention.

[0011] Figure 2This is a schematic diagram of the structure of the Internet of Things-based online radiation environment monitoring system of the present invention;

[0012] Figure labeling: Radiation environment space construction module 11, Internet of Things monitoring data acquisition module 12, monitoring reliability correction module 13, first radiation safety accident prediction module 14, second radiation safety accident prediction module 15, coupled prediction and map construction module 16. Detailed Implementation

[0013] This invention provides an Internet of Things-based online radiation environment monitoring method and system, which solves the technical problem of low accuracy in existing online radiation environment monitoring, leading to poor effectiveness of radiation safety management. It achieves the technical effect of improving the accuracy of online radiation environment monitoring and enhancing the effectiveness and quality of radiation safety management.

[0014] like Figure 1 As shown, this invention provides an online monitoring method for radiation environment based on the Internet of Things, the method comprising:

[0015] S100: The space where the radiation device is located is the first radiation environment space, and the neighborhood space of the first radiation environment space is the second radiation environment space.

[0016] Specifically, the first radiation environment space refers to the core functional area where medical radiation equipment directly operates, such as the treatment room for a linear accelerator in radiotherapy or the scanning room for a CT / DR machine in radiology. This space must completely accommodate the equipment itself and necessary operating space, such as the treatment bed and the area around the equipment control console. It is the area where radiation is directly released, with the highest initial radiation dose, and is the primary monitoring area in medical settings to prevent direct radiation leakage and ensure radiation safety for patients during treatment. The second radiation environment space refers to medical-related areas closely adjacent to the first space where radiation can easily spread, such as the doctor's control room next to the first space, the patient waiting area, the corridor connecting various machine rooms, or the medical staff inspection passage within 1-2 meters outside the walls of the first space. Although these areas are not directly where the radiation source is located, radiation may penetrate due to insufficient wall protection, making them key monitoring areas in the medical field to prevent medical staff and waiting patients from being exposed to radiation.

[0017] S200: The first radiation environment space and the second radiation environment space are monitored online through a group of Internet of Things devices to obtain a first radiation environment monitoring set and a second radiation environment monitoring set.

[0018] Specifically, for the first radiation environment space, the IoT device cluster deploys monitoring nodes around the equipment itself, at the four corners of the machine room, and near the operating positions of medical staff. These nodes focus on collecting real-time radiation dose, equipment operating status, and other information directly related to radiation safety; including radiation measurement data such as X-ray / gamma-ray dose values. For the second radiation environment space, the IoT device cluster deploys nodes at the area entrance and locations where personnel frequently stay, focusing on collecting dose data after radiation diffusion to avoid overlooking potential risks in neighboring areas. The IoT device cluster is a collaborative network composed of various types of intelligent devices adapted to medical scenarios, including radiation dose sensors and data transmission modules. It enables real-time data acquisition and transmission, ultimately summarizing the data to form the first and second radiation environment monitoring sets. Radiation dose sensors include, but are not limited to, personal dosimeters, X-ray and gamma-ray dose rate meters, portable neutron dosimeters, and surface contamination meters used to detect radiation diffusion. The first radiation environment monitoring set is the raw data collection obtained from monitoring the first radiation environment space, such as the radiotherapy linear accelerator room or CT scan room. It covers real-time radiation dose and basic equipment operating parameters within this space, directly reflecting the initial risk status of the core area of ​​the radiation source. The second radiation environment monitoring set is a collection of raw data obtained from monitoring second radiation environment spaces, such as waiting areas adjacent to the machine room and doctors' operation control rooms. It includes diffuse radiation dose, ambient temperature and humidity, etc., reflecting the initial risk of radiation transmission from the core area to the outside.

[0019] S300: Based on the IoT device group, perform monitoring reliability correction on the first radiation environment monitoring set and the second radiation environment monitoring set respectively, and construct the first environmental monitoring sequence and the second environmental monitoring sequence.

[0020] Specifically, after obtaining the first and second radiation environment monitoring sets, abnormal data from both sets are removed to form high-quality time-series data that can support subsequent accident prediction. Monitoring reliability correction involves considering the accuracy requirements of radiation monitoring in medical scenarios, such as the need for medical radiation dose accuracy to the μSv / h level. Based on the operational status of IoT device clusters, such as sensor calibration status, data transmission stability, and whether they are affected by electromagnetic interference from medical equipment, the original monitoring data is effectively screened and corrected. For example, jump data caused by instantaneous electromagnetic interference from X-ray dosimeters in radiotherapy rooms is removed, or low readings caused by power supply fluctuations in waiting areas are corrected, ensuring that the data meets the accuracy standards for medical radiation monitoring and providing a reliable basis for subsequent analysis. The first environmental monitoring sequence is a coherent time-series data chain formed by organizing the corrected original monitoring data in chronological order for the first radiation environment space. It clearly reflects the radiation intensity variation patterns in the core operating area of ​​medical radiation equipment and serves as the core data basis for subsequent prediction of radiation safety accidents in this space. The second environmental monitoring sequence is a data chain formed by chronologically processing the corrected monitoring data for the second radiation environment space. It can record dose changes from the core area to the surrounding area in real time, such as whether there is an abnormal increase in dose due to insufficient protection of the computer room walls. Together with the first environmental monitoring sequence, it forms the data basis for dual-space radiation risk analysis and supports subsequent coupled prediction.

[0021] S400: Based on the first environmental monitoring sequence, perform radiation safety accident prediction on the first radiation environment space to obtain the characteristics of the first radiation safety accident.

[0022] Specifically, after obtaining the first environmental monitoring sequence, the potential for radiation safety accidents within the space is analyzed using this sequence to extract key risk information. Specifically, the prediction combines the operational characteristics of medical radiation equipment, such as the dose output patterns of radiotherapy machines and the scanning parameter ranges of CT scanners, to capture abnormal signals from the first environmental monitoring sequence, determine the presence of radiation leakage or dose exceeding risk trends, and thus pinpoint core information related to the accident. Abnormal signals include sudden increases in radiation dose and exceeding equipment temperature limits. The characteristics of the first radiation safety accident are a set of key information output by this prediction, reflecting the essence of the risk in the first space, including the probability of accident occurrence, risk level, possible accident location, impact range, and associated abnormal equipment parameters. This provides a risk basis for the core area in subsequent dual-space coupled predictions. The probability of accident occurrence includes, for example, the probability of dose exceeding limits; the risk level includes, for example, minor leakage / severe exceeding limits; the possible accident location includes, for example, the radiation exit of the radiotherapy machine; the impact range includes, for example, the operating area within the machine room; and the associated abnormal equipment parameters include, for example, a dose rate exceeding 100 μSv / h.

[0023] S500: Based on the second environmental monitoring sequence, predict radiation safety accidents in the second radiation environment space and obtain the characteristics of the second radiation safety accident.

[0024] Specifically, after obtaining the second environmental monitoring sequence, the analysis of this sequence reveals potential safety hazards resulting from radiation spreading from the core area to adjacent areas, extracting key risk information. Specifically, the prediction will incorporate the characteristics of personnel activity in the medical adjacent spaces, such as patients staying in the waiting area or medical staff on duty in the control room, capturing abnormal radiation diffusion signals from the second environmental monitoring sequence; for example, a sudden increase in radiation dose in the waiting area exceeding the medical environment safety threshold, or a synchronous increase in the dose in the control room as the equipment in the first space operates, thereby determining whether there is a risk trend of radiation penetrating the walls and leaking from the core area to adjacent areas. The second radiation safety accident characteristic is a set of key information output by the prediction that reflects the nature of the second space diffusion risk. It includes the probability of accident occurrence, risk level, risk concentration location, affected population range, and associated diffusion dose parameters. It is used to present the risk status of radiation propagation from the core area outward, and to provide a basis for neighborhood risk for subsequent dual-space coupling prediction. Among them, the probability of accident occurrence is such as the probability of neighborhood dose exceeding the standard, the risk level is such as slight diffusion / moderate leakage, the risk concentration location is such as the side of the waiting area near the machine room, the affected population range is such as waiting patients and on-duty medical staff, and the associated diffusion dose parameters are such as peak dose exceeding 5 μSv / h.

[0025] S600: Based on the first radiation safety accident characteristics and the second radiation safety accident characteristics, perform coupled prediction of radiation safety accidents and establish a radiation safety accident map.

[0026] Specifically, after obtaining the characteristics of the first radiation safety accident and the characteristics of the second radiation safety accident, the characteristics of the first radiation safety accident and the characteristics of the second radiation safety accident are integrated, the mutual influence between the two is analyzed, and finally an intuitive risk presentation carrier is formed.

[0027] Radiation safety accident coupling prediction combines the radiation diffusion characteristics in medical scenarios, such as X / γ-ray penetration through walls and air conduction, to analyze the correlation between first-level radiation safety accident characteristics, such as the probability of radiation leakage and peak dose in the computer room, and second-level radiation safety accident characteristics, such as the diffusion dose and risk level in the waiting area. For example, it determines whether a leak in the computer room will exacerbate the risk in the waiting area through wall penetration, or whether abnormal doses in the waiting area will conversely confirm the failure of computer room protection. This allows for the prediction of the overall accident risk of the two spaces overlapping, such as the possibility of cross-space radiation exceeding standards and the scope of affected populations. The radiation safety accident atlas is a visual carrier of radiation risk in medical scenarios generated based on the coupling prediction results. It marks the accident location in the first space, the risk concentration area in the second space, and the transmission path of risks between the two spaces. It also integrates information such as risk level and types of affected personnel, providing an intuitive and comprehensive risk reference for radiation safety management in medical scenarios.

[0028] Furthermore, the present invention also provides a method for performing monitoring credibility correction on the first radiation environment monitoring set and the second radiation environment monitoring set based on the IoT device group, including: collecting device sensing accompanying parameters of the IoT device group based on each monitoring data in the first radiation environment monitoring set to obtain multiple device sensing accompanying data; performing monitoring credibility evaluation on the multiple device sensing accompanying data to obtain multiple monitoring credibility evaluation coefficients; performing outlier detection on the multiple monitoring credibility evaluation coefficients according to monitoring credibility evaluation constraints to determine monitoring credibility evaluation outliers; and performing credibility correction on the first radiation environment monitoring set based on the monitoring credibility evaluation outliers to generate the first environmental monitoring sequence.

[0029] Specifically, this involves using a network of IoT devices to collaboratively verify the first radiation environment monitoring set, generating time-series data that meets accuracy requirements, thus laying the foundation for subsequent safety analysis. Accompanying data from device sensors is a set of sensor operating status parameters synchronized with the radiation monitoring data acquisition time, directly related to the accuracy of data in the medical scenario. For example, when an X-ray dose of 200 μSv / h is collected in the radiotherapy room, the synchronously acquired accompanying data includes sensor calibration 15 days ago, operating voltage 219V, electromagnetic interference 35dB, and equipment temperature 26℃, fully reflecting the status of the equipment collecting this dose data. Specifically, the 15-day sensor calibration is within the quarterly mandatory medical calibration validity period; the operating voltage of 219V meets the 220V±5% standard; the electromagnetic interference of 35dB is below the 50dB anti-interference threshold around the radiotherapy equipment; and the equipment temperature of 26℃ is within the normal range of 15-50℃.

[0030] After acquiring sensor data from multiple devices, a reliability evaluation is performed on this data. This involves considering the high precision requirements of medical radiation monitoring and assessing reliability by correlating raw radiation data with device status parameters. If the dose data is continuous and stable, and the accompanying data indicates normal device status, the data is considered reliable. Conversely, if the dose suddenly increases, and the accompanying data indicates that the sensor is outdated and calibration is not performed, or voltage fluctuations exceed limits, the data is considered unreliable. The evaluation result is then quantified into a monitoring reliability evaluation coefficient, ranging from 0 to 1. Values ​​closer to 1 indicate that the data better meets medical-grade standards. For example, data within calibration and without interference has a coefficient of 0.95, while data outdated and subject to strong interference has a coefficient of only 0.2. Following this, a screening process is conducted based on monitoring reliability evaluation constraints, including coefficient threshold constraints and fluctuation amplitude constraints. For example, data with a coefficient threshold greater than or equal to 0.6 are considered basically reliable; the radiotherapy department defines this as the baseline for safety assessment. For fluctuation amplitude constraints, the difference between adjacent data coefficients must be less than or equal to 0.3 to avoid sudden changes in reliability. Data that do not meet the constraints are grouped into a monitoring credibility evaluation outlier cluster. For example, if the coefficients of three sets of data in a CT scan room at a certain time period are 0.5, 0.45, and 0.55, and the difference between these and the adjacent coefficients of 0.9 and 0.92 exceeds 0.3, then these three sets of dose data and evaluation coefficients belong to the outlier cluster. Finally, the first radiation environment monitoring set is corrected according to the outlier cluster. For example, the outlier values ​​are replaced by the weighted average of the credible data before and after the correction. The corrected data is then sorted by time to generate the first environmental monitoring sequence, ensuring that the data meets the safety analysis requirements of the medical scenario.

[0031] Furthermore, the present invention also provides a method for monitoring credibility evaluation of the accompanying data of the multiple devices to obtain multiple monitoring credibility evaluation coefficients, including: identifying anomalies based on the accompanying data of the multiple devices to determine each accompanying abnormal segment; conducting monitoring credibility evaluation of multiple historical accompanying abnormal segments based on multiple data experts to obtain multiple monitoring credibility evaluation sequences; calculating a central value based on each monitoring credibility evaluation sequence to obtain each monitoring credibility central value; mapping and aligning the multiple historical accompanying abnormal segments and the monitoring credibility central values ​​to generate a monitoring credibility evaluation learning set; integrating and training the monitoring credibility evaluation learning set using multiple learners to establish a monitoring credibility evaluator; and inputting each accompanying abnormal segment into the monitoring credibility evaluator to generate the multiple monitoring credibility evaluation coefficients.

[0032] Specifically, this involves integrating expert experience and machine learning to transform equipment sensor data into quantitative reliability indicators, ensuring that radiation monitoring data evaluation meets the stringent requirements of medical safety. Monitoring reliability evaluation combines radiation protection monitoring standards with the identification of sensor anomalies, integration of expert judgment and algorithmic models, to comprehensively determine the reliability of equipment sensor data.

[0033] Specifically, the first step is to identify anomalies in the accompanying data of multiple devices' sensors. For example, from data such as the calibration time of the radiotherapy linear accelerator and the working voltage of the CT scanner, continuous segments that do not meet medical standards are screened out. For example, if the sensor calibration is overdue by one month, the voltage fluctuation is ±8%, and the electromagnetic interference intensity reaches 60dB, then each abnormal segment of the sensor is a continuous segment that does not meet medical standards.

[0034] Subsequently, 3-5 experts with over 10 years of experience in medical radiation monitoring were invited to score each of the hospital's historically accumulated equipment sensor abnormality segments, with scores ranging from 0 to 1, where 1 point represents data that fully meets medical accuracy requirements. For example, they scored each of the hospital's historically accumulated radiotherapy / CT sensor abnormality segments. Each expert's scores for the same type of historical abnormality segment were arranged sequentially to form a monitoring reliability evaluation sequence. For example, expert A's scoring sequence for segments with calibration overdue by one month was [0.3, 0.28, 0.32]. To avoid subjective bias from a single expert, a monitoring reliability central tendency value was calculated for each monitoring reliability evaluation sequence, i.e., statistical methods such as mean and median were used to aggregate scores. For example, the central tendency value for the above sequence was (0.3 + 0.28 + 0.32) / 3 = 0.3, objectively integrating expert opinions.

[0035] Next, multiple historical sensor-related abnormal segments are mapped and bound one-to-one with the corresponding monitoring reliability set values ​​to generate a monitoring reliability evaluation learning set. For example, abnormal segments of data with calibration overdue by 1 month and voltage fluctuation of 8% are associated with a set value of 0.3 as samples for machine learning. The samples need to cover common abnormality types in medical scenarios such as sensor calibration overdue, electromagnetic interference, and temperature exceeding the limit.

[0036] After obtaining the monitoring credibility evaluation learning set, integrated training is performed using multiple different types of learners, such as decision trees, neural networks, and support vector machines, to simultaneously train the monitoring credibility evaluation learning set. Weights are assigned to each model based on its medical data prediction accuracy; for example, models with smaller prediction errors have higher weights. The output results of each model are merged to avoid bias in medical data from a single model, thus establishing a monitoring credibility evaluator. The monitoring credibility evaluator is an intelligent model validated by historical medical data, capable of automatically receiving sensor-related abnormal segments and outputting credibility results. Finally, each sensor-related abnormal segment is input into the evaluator. Based on the expert experience and data patterns learned during training, the model outputs a quantitative value in the range of 0-1, i.e., the monitoring credibility evaluation coefficient. For example, an output of 0.25 indicates that the radiation monitoring data corresponding to the abnormal segment has low credibility and needs to be addressed in subsequent correction.

[0037] Furthermore, the present invention also provides a method for predicting radiation safety accidents in the first radiation environment space based on the first environmental monitoring sequence to obtain first radiation safety accident features, comprising: performing a radiation safety correlation evaluation based on the first environmental monitoring sequence to obtain a first radiation safety correlation evaluation sequence; performing attention enhancement on the first environmental monitoring sequence based on the first radiation safety correlation evaluation sequence to obtain a first environmental monitoring enhancement vector; performing a historical retrieval of radiation safety accidents in the first radiation environment space to obtain a first historical set of radiation safety accidents; performing edge-cloud collaborative learning based on the first historical set of radiation safety accidents to build a first radiation safety accident prediction channel; and inputting the first environmental monitoring enhancement vector into the first radiation safety accident prediction channel to generate the first radiation safety accident features.

[0038] Specifically, this involves a closed-loop process of data evaluation, feature enhancement, historical data linkage, and model prediction to accurately extract key accident information from the core area of ​​the first radiation environment. Specifically, firstly, in accordance with radiation protection monitoring standards, each data point in the monitoring sequence is correlated with radiation safety indicators to determine whether there are any safety hazards. For example, if dose data for a certain period approaches 200 μSv / h three times consecutively and voltage fluctuations reach ±4%, then a potential risk is identified for that period; if the data remains within the safe range, then no risk is identified. Based on this, the risk assessment results, arranged chronologically, are represented by scores of 0-1, with 1 indicating high risk and 0 indicating low risk. This constitutes the first radiation safety correlation assessment sequence, such as [0.2, 0.3, 0.8, 0.7], corresponding to risk values ​​at four time points.

[0039] Next, the weights of these key data points are increased through algorithms. Some data are more critical for accident prediction, such as periods of sudden dose increases or abnormal equipment parameters. For example, if the dose in a radiotherapy room suddenly increases from 180 μSv / h to 195 μSv / h, approaching the threshold, the weight of this data in the sequence increases from 0.5 to 0.8, while the weight of stable, low-risk data decreases to 0.3. After processing, the time-series data is transformed into a vector form highlighting key risk characteristics, serving as the first environmental monitoring enhancement vector. Its dimensions cover core parameters such as dose, voltage, and temperature, with the weight of key risk dimensions significantly higher than that of ordinary dimensions.

[0040] After obtaining the enhanced vector for the first environmental monitoring, a historical radiation safety accident retrieval is performed on the first radiation environment space. This involves retrieving a set of historical accident case data consistent with the type of the first radiation environment space from the hospital information system or radiation safety management platform. This includes cases of radiation dose exceeding the standard and monitoring anomalies caused by equipment failure in radiotherapy rooms and CT rooms within the past 5 years. Each case must include monitoring data at the time of the accident, the cause of the accident, and the accident level. The monitoring data at the time of the accident includes the dose sequence and equipment status; the cause of the accident includes overdue sensor calibration and damaged wall protection; and the accident level includes information such as minor leakage and severe exceeding the standard. In other words, the first radiation safety accident historical set is a set of historical accident case data that perfectly matches the type of the first radiation environment space, retrieved from the hospital information system or radiation safety management platform, and each case contains key information to support subsequent risk analysis. For example, data from a 2023 radiotherapy room where the radiation exit shielding door was not properly closed, resulting in a momentary dose exceeding 300 μSv / h.

[0041] Subsequently, a prediction channel is built through edge-cloud collaborative learning. The edge refers to local edge computing devices in the first radiation environment space, such as industrial control hosts in the computer room, responsible for processing enhancement vector features with high real-time requirements, such as instantaneous data of sudden dose increases. The cloud refers to the hospital's cloud-based big data platform, responsible for accessing massive amounts of historical data from the first radiation safety accident to train the model. When the two collaborate, the edge transmits real-time enhancement features to the cloud, which then combines historical data to train the prediction model, such as a hybrid model fusing decision trees and neural networks. The trained model parameters are then distributed back to the edge, forming the first radiation safety accident prediction channel. This channel can simultaneously utilize the timeliness of real-time data and the regularity of historical data, avoiding the limitations of single-end processing, such as insufficient data volume at the edge or poor real-time performance at the cloud.

[0042] After obtaining the prediction channel, the first environmental monitoring enhancement vector is input into the prediction channel. The model combines real-time key features with historical accident patterns to output a set of key information reflecting the nature of the first radiation environment spatial risk, namely the first radiation safety accident characteristics, including the probability of accident occurrence, such as the probability of dose exceeding the standard, the risk level, such as mild risk, the possible accident location, such as the radiation exit of the radiotherapy machine, and related abnormal parameters, such as the dose rate exceeding 190 μSv / h, etc., to provide risk basis for the core area for subsequent dual-space coupling prediction.

[0043] Furthermore, the present invention also provides a method for constructing a first radiation safety accident prediction channel by performing edge-cloud collaborative learning based on the first radiation safety accident history set, comprising: performing supervised learning on the first radiation safety accident history set based on multiple edge nodes to obtain multiple radiation safety accident predictors; transmitting the first radiation safety accident history set and the multiple radiation safety accident predictors to cloud nodes; randomly perturbing the first radiation safety accident history set through the cloud nodes to obtain a first perturbed radiation safety accident set; testing the multiple radiation safety accident predictors based on the first perturbed radiation safety accident set to obtain multiple accident prediction loss sets; performing attention enhancement on the first perturbed radiation safety accident set based on the multiple accident prediction loss sets to obtain multiple enhanced radiation safety accident spaces; and performing aggregate learning on the multiple radiation safety accident predictors based on the multiple enhanced radiation safety accident spaces to generate the first radiation safety accident prediction channel.

[0044] Specifically, this involves constructing a spatial accident prediction model for the first radiation environment through an edge-cloud collaborative model that combines localized learning at the edge with global optimization in the cloud. This model leverages the real-time capabilities of the edge and the scale of data in the cloud to improve prediction accuracy and robustness.

[0045] Specifically, supervised learning is first performed through edge nodes. Edge nodes refer to local computing devices within the first radiation environment space. Supervised learning is the process by which edge nodes train models using a historical set of radiation safety accidents as samples, mapping inputs to outputs. The inputs are monitoring sequences prior to the historical accidents, such as dose changes and equipment voltage; the outputs are accident types, such as radiation leakage or dose exceeding thresholds, or risk levels, such as mild or severe. Through supervised learning, each edge node generates a radiation safety accident predictor—a model that can output accident prediction results based on real-time monitoring data, such as a decision tree-based predictor or a neural network model. For example, the edge nodes in a radiotherapy room can train a predictor that can predict the probability of exceeding the limit based on real-time dose increase trends by learning from 200 historical cases of dose exceeding limits.

[0046] Subsequently, the edge nodes synchronize the first set of historical radiation safety incidents with multiple trained radiation safety incident predictors to the cloud nodes, such as the hospital radiation safety management cloud platform. Random perturbation involves making small, reasonable modifications to key parameters in the first set of historical radiation safety incidents to simulate possible fluctuations in monitoring data in medical scenarios. For example, the peak dose of an incident from 200 μSv / h is fine-tuned to 195 μSv / h or 205 μSv / h, and the equipment temperature is adjusted from 28℃ to 27℃ or 29℃, generating the first perturbation set of radiation safety incidents. Random perturbation tests the stability of the predictors when data changes slightly, avoiding excessive reliance on fixed values ​​from historical data and adapting to the inherent small fluctuations in sensor data in medical scenarios.

[0047] After obtaining the first set of disturbed radiation safety incidents, the cloud platform uses this set to test multiple radiation safety incident predictors, calculating the incident prediction loss set for each predictor. Specifically, the incident prediction loss set of a predictor is the set of errors between the predictor's prediction of the disturbed set and the actual incident label, quantified using metrics such as mean squared error and cross-entropy. For example, if a predictor has significant errors in predicting 20 out of 100 sets of disturbed data, such as misjudging risk levels, these error values, arranged in order, constitute the predictor's loss set. The loss set reflects the predictor's weaknesses, such as inaccurate predictions of dose data close to the threshold.

[0048] Subsequently, based on multiple accident prediction loss sets, the cloud assigns higher weights to perturbation cases with larger errors in the loss set, highlighting their importance in training, thereby forming an enhanced radiation safety accident space. The enhanced radiation safety accident space is a dataset that focuses on difficult-to-predict cases; for example, the weights of the above 20 high-error perturbation cases are increased from 0.01 to 0.05 to ensure that the model's subsequent learning pays more attention to these critical medical safety scenarios.

[0049] Finally, the first radiation safety accident prediction channel is generated through aggregate learning. The cloud integrates the learning results of multiple radiation safety accident predictors in the enhanced radiation safety accident space, employing methods such as weighted voting and model ensemble to combine the strengths of each predictor. For example, predictor A excels at accidents caused by equipment failure, while predictor B excels at dose surge accidents, forming a comprehensive model that balances real-time performance and generalization capabilities—this is the first radiation safety accident prediction channel. After deployment, this channel can quickly receive enhanced vectors from the first environmental monitoring system, such as real-time dose characteristics of radiotherapy rooms, and output accurate accident prediction results, adapting to the high reliability requirements of radiation safety in the medical field.

[0050] Furthermore, the present invention also provides a method for predicting radiation safety accidents in the second radiation environment space based on the second environmental monitoring sequence to obtain second radiation safety accident features, comprising: obtaining a second radiation safety correlation evaluation sequence by performing a radiation safety correlation evaluation on the second environmental monitoring sequence; performing attention enhancement on the second environmental monitoring sequence based on the second radiation safety correlation evaluation sequence to obtain a second environmental monitoring enhancement vector; performing a radiation safety accident history retrieval on the second radiation environment space to obtain a second radiation safety accident history set; performing edge-cloud collaborative learning based on the second radiation safety accident history set to build a second radiation safety accident prediction channel; and inputting the second environmental monitoring enhancement vector into the second radiation safety accident prediction channel to generate the second radiation safety accident features.

[0051] Specifically, this involves predicting radiation safety risks in the second radiation environment space. Through a closed-loop process of evaluation, enhancement, linkage, and prediction, key accident information in the adjacent space is accurately extracted to match its risk characteristics, which are mainly characterized by radiation diffusion impact.

[0052] Specifically, firstly, by combining the safety threshold of the second space in the radiation protection monitoring specifications, each data point in the second environmental monitoring sequence is compared with the safety indicators to determine whether there is a risk of diffusion; for example, if the dose is continuously close to 2.5 μSv / h or increases synchronously with the operation of the equipment in the first space, it is determined to be a potential risk.

[0053] As shown in the first environmental monitoring sequence above, these risk assessment results, arranged in chronological order, serve as the second radiation safety correlation evaluation sequence, such as [0.1, 0.4, 0.8, 0.6], which intuitively reflects the risk change trend in the surrounding space.

[0054] Next, attention enhancement was performed based on the second radiation safety correlation assessment sequence, increasing the weight of key data. For example, the dose data from monitoring points near the machine room wall in the waiting area and the neighboring dose data when the equipment in the first space was started were increased from 0.4 to 0.7, while the weight of stable, low-risk data far from the diffusion source was reduced to 0.2. After this processing, the time-series data was transformed into a vector form that highlights the diffusion risk characteristics, namely the second environmental monitoring enhancement vector, whose dimensions include diffusion dose, distance correlation parameters with the first space, etc., focusing on the core risk points in the neighborhood.

[0055] Subsequently, a second set of historical radiation safety accidents was retrieved from the hospital's radiation safety management platform. The set included historical cases such as dose exceeding the standard in the waiting area due to insufficient wall protection in the machine room, and radiation diffusion caused by leakage through door gaps in the operation control room. Each case included information such as the diffusion dose data at the time of the accident, the location of the leak, and the scope of the affected population. For example, the case data of a 2022 radiotherapy machine room waiting area where the local dose reached 3.1 μSv / h due to damage to the lead plate in the wall was included in this set.

[0056] Subsequently, edge-cloud collaborative learning is performed based on the historical data set. The edge nodes are local computing devices in the second space, responsible for processing the enhanced features of real-time diffusion data. The cloud nodes call the historical data set to train the model. After the two are optimized together, a second radiation safety accident prediction channel is built. This channel takes into account both the real-time diffusion characteristics of the neighborhood data and the historical accident patterns, avoiding the limitations of single-end processing.

[0057] Finally, the second environmental monitoring enhancement vector is input into this channel. The model combines diffusion risk characteristics and historical patterns to output a set of key information reflecting the nature of the second space risk, namely the characteristics of the second radiation safety accident, including the probability of accident occurrence, risk level, risk concentration location, and affected population, providing a basis for neighborhood risk for subsequent dual-space coupling prediction.

[0058] Furthermore, the present invention also provides a method for coupled prediction of radiation safety accidents based on the first radiation safety accident characteristics and the second radiation safety accident characteristics, and for establishing a radiation safety accident atlas, comprising: performing three-dimensional modeling of the first radiation environment space and the second radiation environment space to obtain a radiation environment space model; performing boundary smoothing enhancement based on the radiation environment space model to obtain a radiation environment enhancement model; performing radiation diffusion analysis on the first radiation safety accident characteristics based on the radiation environment enhancement model to determine a first accident radiation diffusion characteristic; performing radiation diffusion analysis on the second radiation safety accident characteristics based on the radiation environment enhancement model to obtain a second accident radiation diffusion characteristic; performing coupled prediction of radiation safety accidents based on the first accident radiation diffusion characteristic and the second accident radiation diffusion characteristic based on the radiation environment enhancement model to obtain a third radiation safety accident characteristic; and generating the radiation safety accident atlas based on the first radiation safety accident characteristics, the second radiation safety accident characteristics, and the third radiation safety accident characteristics.

[0059] Specifically, this involves integrating the risks of the first and second radiation environments to construct a panoramic view of radiation safety management and control. Through spatial modeling, diffusion analysis, and risk coupling, the visualization and collaborative assessment of risks in both environments can be achieved.

[0060] Specifically, the process begins with digitally reconstructing the physical structure, equipment layout, and protective materials of the two spaces based on hospital architectural drawings and on-site measurement data. For example, when reconstructing the radiotherapy room, the coordinates of the linear accelerator need to be accurately marked, such as the central area of ​​the room, the thickness of the lead shielding wall, and the location of the ventilation openings. When reconstructing the waiting area, the distribution of seats, the distance from the room walls, and the wall material, such as gypsum board containing 1mm of lead equivalent, need to be marked to ensure a 1:1 match between the model and the actual scene. The resulting radiation environment spatial model is a digital carrier containing the physical parameters of both spaces, providing a spatial basis for subsequent radiation diffusion calculations. These physical parameters include dimensions, material properties, and equipment coordinates.

[0061] Next, details of potential interference and diffusion analysis in the radiation environment spatial model are addressed, such as the joints in the walls between the machine room and the waiting area, door gaps, and gaps between equipment and the floor. Algorithms are used for accuracy optimization and smoothing. For example, the sharp edges of the wall joints are corrected to a smooth transition structure that conforms to actual construction precision, avoiding deviations in radiation leakage path calculations caused by modeling sharp edges. Sealing parameters are annotated for door gaps to ensure the model reflects the actual radiation penetration scenario. The resulting enhanced radiation environment model more closely resembles the physical reality of the medical space and can accurately calculate the attenuation of radiation under different materials and distances. For instance, the attenuation coefficient of X-rays through a 3mm lead wall is 0.98, meaning that 200 μSv / h of radiation decreases to 4 μSv / h after passing through the wall.

[0062] Subsequently, radiation diffusion analysis was performed based on the enhanced model. Combining the physical propagation characteristics of medical radiation, such as X-rays following the square-distance attenuation law and exhibiting exponential attenuation upon penetrating materials, diffusion paths and dose distribution analyses were conducted for the characteristics of dual-space accidents. When analyzing the characteristics of the first radiation safety accident, the dose attenuation of radiation passing through the lead wall and the diffusion range in the waiting area were calculated, forming the radiation diffusion characteristics of the first accident, including the diffusion direction, dose attenuation curve, and affected area. When analyzing the characteristics of the second radiation safety accident, possible leakage sources were inferred, such as a lead layer failure at a certain location in the wall. The propagation path of radiation from the source to the waiting area was calculated, forming the radiation diffusion characteristics of the second accident, supplementing the diffusion source tracing information. The radiation diffusion characteristics of the second accident include precise leakage source location information, medium parameters of the radiation propagation path, inferred leakage source intensity, affected sub-regions associated with the source tracing, and estimated leakage duration. This comprehensively presents the entire chain of source tracing information from leakage source location, path analysis, intensity inference, and impact association, providing a basis for locating risk correlation points in the subsequent coupled prediction. Among them, the precise location information of the leakage source is as follows: at the coordinates (3.2m, 1.5m) of the wall of the adjacent machine room in the waiting area, there is a 2cm×3cm lead layer missing, or the sealant of the door between the machine room and the waiting area has aged and failed, with a gap width of 0.3cm; the medium parameters of the radiation propagation path are as follows: from the leakage source to the monitoring point exceeding the standard in the waiting area, the propagation path is: lead layer missing → 1.2m air layer → 0.5cm thick gypsum board → waiting area floor, and the radiation attenuation coefficient of each medium is marked, with an air attenuation coefficient of 0.02 / m and a gypsum board attenuation coefficient of 0.15 / cm; the intensity reflection of the leakage source is also included. Based on the 5 μSv / h dose monitored in the waiting area and combined with the path attenuation law, the initial radiation dose at the source is estimated to be 185 μSv / h, matching the typical intensity range of radiation leakage in the machine room. The affected sub-areas related to the source are the waiting area seats 1-3, which are directly corresponding to the leakage source. This area is less than 2m from the source, with a dose exceeding the public safety threshold of 2.5 μSv / h, and an average of 15-20 patients stay there daily. The leakage duration is estimated by combining the dose time series data of the waiting area, indicating that the source leakage has lasted for 4-6 hours, which is consistent with the continuous operation period of the equipment in the machine room.

[0063] Subsequently, based on the radiation environment enhancement model, the superposition effect of the radiation diffusion characteristics of the first and second accidents was analyzed. For example, if the first characteristic shows that the radiation dose from the leak in the machine room at a certain point in the waiting area is 3 μSv / h, and the second characteristic shows that the area has an additional 2 μSv / h due to wall damage, then the coupled prediction of the actual dose at that point reaches 5 μSv / h, exceeding the safety threshold. The resulting third radiation safety accident characteristic is a comprehensive risk information after the superposition of risks in two spaces, including the superimposed dose value, the high-risk superposition area, and the range of affected people, such as patients in the five seats around that point in the waiting area.

[0064] Finally, by integrating the characteristics of the first, second, and third radiation safety accidents, a radiation safety accident atlas is generated. This atlas serves as a visual representation of risks, using a 3D enhanced model as the base map and different colors to mark risk points in both spaces. For example, the X-ray exit of the machine room is marked as a red high-risk point, the diffusion path is marked as a yellow line indicating the direction of radiation propagation, and the superimposed high-risk area is marked as an orange area indicating the area of ​​the waiting area where the dose exceeds the standard. It also includes textual descriptions such as the risk level and the types of people affected, providing medical personnel with an intuitive basis for quickly locating risks and formulating control measures, such as temporarily closing high-risk areas in the waiting area.

[0065] Furthermore, the present invention also provides online monitoring of the first radiation environment space and the second radiation environment space through an Internet of Things (IoT) device group to obtain a first radiation environment monitoring set and a second radiation environment monitoring set, comprising: reading the IoT device group to obtain a multidimensional monitoring dataset; and cleaning and classifying the multidimensional monitoring dataset to obtain the first radiation environment monitoring set and the second radiation environment monitoring set.

[0066] Specifically, the multidimensional monitoring dataset is the raw data collection obtained after reading the IoT device cluster, covering multidimensional information in the medical scenario. The device cluster includes radiation dose sensors, device status sensors, and environmental sensors deployed in two spaces. The data dimensions include core indicators such as radiation dose values, device operating parameters, and environmental parameters, as well as related information such as acquisition time and sensor location. For example, a data point might be 2025-XX-XX 11:00, radiotherapy room A, dose 180μSv / h, voltage 219V, temperature 26℃.

[0067] After acquiring the multidimensional monitoring dataset, the dataset is cleaned and classified to remove invalid data in the medical scenario, such as dose jump values ​​caused by sensor failure and incomplete data due to acquisition interruption. Based on the spatial attributes of sensor deployment, all valid data from the first radiation environment space are integrated into the first radiation environment monitoring set, and valid data from the second radiation environment space are integrated into the second radiation environment monitoring set, providing basic data for subsequent reliable calibration in the two spaces.

[0068] Furthermore, the present invention also provides a method for generating a radiation safety early warning signal based on the aforementioned radiation safety accident map and a radiation alarm device.

[0069] Specifically, this involves linking the risk information in the radiation safety accident map with radiation alarm devices, transforming abstract risk data into intuitive and responsive early warning signals to ensure the safety of medical staff and patients.

[0070] The radiation safety accident map includes the dose thresholds corresponding to high / medium / low risk levels in the first radiation environment space, the safety thresholds in the second radiation environment space, and the affected areas corresponding to each risk level. Radiation alarm devices consist of audible and visual alarms deployed in the computer room, electronic screen alarm modules in the waiting area, portable alarm terminals worn by medical staff, and a backend alarm module linked to the hospital's safety management system, enabling local audible and visual alerts and remote information push notifications.

[0071] Specifically, the system retrieves real-time radiation data from the IoT device cluster, such as the real-time dose in the radiotherapy room (195 μSv / h) and the real-time dose in the waiting area (3 μSv / h). This data is compared with the corresponding risk thresholds in the radiation safety accident map. If the dose in the radiotherapy room reaches the medium-risk range, the audible and visual alarm in the room will activate, flashing a yellow light, and the screen will display a medium-risk dose of 195 μSv / h. If the dose in the waiting area exceeds the safety threshold, a red warning pop-up will appear on the electronic screen in the waiting area, and warning information will be pushed to the terminals of the radiotherapy doctors and the hospital security center. The warning information includes the risk area, the current dose, and recommended measures, such as suspending the use of the waiting area. The final radiation safety warning signal is matched with different response forms according to the risk level. Mild risk is indicated by a blue warning, including only background records and text on the electronic screen; moderate risk is indicated by a yellow audible and visual warning, including local reminders and notifications on medical staff terminals; severe risk is indicated by a red audible and visual warning and an emergency push, including linkage with the hospital's emergency broadcast system to prompt personnel to evacuate, ensuring the accuracy and timeliness of risk response in medical scenarios.

[0072] Using the space where the radiation equipment is located as the first radiation environment space, and the neighboring space of the first radiation environment space as the second radiation environment space; the first and second radiation environment spaces are monitored online through an Internet of Things (IoT) device cluster to obtain a first radiation environment monitoring set and a second radiation environment monitoring set; the first and second radiation environment monitoring sets are respectively subjected to monitoring reliability correction based on the IoT device cluster to construct a first environmental monitoring sequence and a second environmental monitoring sequence; radiation safety accidents are predicted in the first radiation environment space based on the first environmental monitoring sequence to obtain a first radiation safety accident feature; radiation safety accidents are predicted in the second radiation environment space based on the second environmental monitoring sequence to obtain a second radiation safety accident feature; radiation safety accidents are coupled and predicted based on the first and second radiation safety accident features to establish a radiation safety accident map; this achieves the technical effect of improving the accuracy of online radiation environment monitoring and enhancing the effectiveness and quality of radiation safety management.

[0073] Based on the same inventive concept as the aforementioned Internet of Things (IoT)-based online radiation environment monitoring method, this invention also provides an IoT-based online radiation environment monitoring system, such as... Figure 2As shown, the system includes:

[0074] Radiation environment space construction module 11 is used to take the space where the radiation device is located as the first radiation environment space and the neighborhood space of the first radiation environment space as the second radiation environment space.

[0075] The Internet of Things (IoT) monitoring data acquisition module 12 is used to perform online monitoring of the first radiation environment space and the second radiation environment space through a group of IoT devices, and to obtain a first radiation environment monitoring set and a second radiation environment monitoring set.

[0076] The monitoring reliability correction module 13 is used to perform monitoring reliability correction on the first radiation environment monitoring set and the second radiation environment monitoring set according to the IoT device group, and construct the first environmental monitoring sequence and the second environmental monitoring sequence.

[0077] The first radiation safety accident prediction module 14 is used to predict radiation safety accidents in the first radiation environment space based on the first environmental monitoring sequence and obtain the characteristics of the first radiation safety accident.

[0078] The second radiation safety accident prediction module 15 is used to predict radiation safety accidents in the second radiation environment space based on the second environmental monitoring sequence and obtain the characteristics of the second radiation safety accident.

[0079] The coupling prediction and map construction module 16 is used to perform coupled prediction of radiation safety accidents based on the first radiation safety accident characteristics and the second radiation safety accident characteristics, and to establish a radiation safety accident map.

[0080] Furthermore, the present invention also provides a method for performing monitoring credibility correction on the first radiation environment monitoring set and the second radiation environment monitoring set based on the IoT device group, including: collecting device sensing accompanying parameters of the IoT device group based on each monitoring data in the first radiation environment monitoring set to obtain multiple device sensing accompanying data; performing monitoring credibility evaluation on the multiple device sensing accompanying data to obtain multiple monitoring credibility evaluation coefficients; performing outlier detection on the multiple monitoring credibility evaluation coefficients according to monitoring credibility evaluation constraints to determine monitoring credibility evaluation outliers; and performing credibility correction on the first radiation environment monitoring set based on the monitoring credibility evaluation outliers to generate the first environmental monitoring sequence.

[0081] Furthermore, the present invention also provides a method for monitoring credibility evaluation of the accompanying data of the multiple devices to obtain multiple monitoring credibility evaluation coefficients, including: identifying anomalies based on the accompanying data of the multiple devices to determine each accompanying abnormal segment; conducting monitoring credibility evaluation of multiple historical accompanying abnormal segments based on multiple data experts to obtain multiple monitoring credibility evaluation sequences; calculating a central value based on each monitoring credibility evaluation sequence to obtain each monitoring credibility central value; mapping and aligning the multiple historical accompanying abnormal segments and the monitoring credibility central values ​​to generate a monitoring credibility evaluation learning set; integrating and training the monitoring credibility evaluation learning set using multiple learners to establish a monitoring credibility evaluator; and inputting each accompanying abnormal segment into the monitoring credibility evaluator to generate the multiple monitoring credibility evaluation coefficients.

[0082] Furthermore, the present invention also provides a method for predicting radiation safety accidents in the first radiation environment space based on the first environmental monitoring sequence to obtain first radiation safety accident features, comprising: performing a radiation safety correlation evaluation based on the first environmental monitoring sequence to obtain a first radiation safety correlation evaluation sequence; performing attention enhancement on the first environmental monitoring sequence based on the first radiation safety correlation evaluation sequence to obtain a first environmental monitoring enhancement vector; performing a historical retrieval of radiation safety accidents in the first radiation environment space to obtain a first historical set of radiation safety accidents; performing edge-cloud collaborative learning based on the first historical set of radiation safety accidents to build a first radiation safety accident prediction channel; and inputting the first environmental monitoring enhancement vector into the first radiation safety accident prediction channel to generate the first radiation safety accident features.

[0083] Furthermore, the present invention also provides a method for constructing a first radiation safety accident prediction channel by performing edge-cloud collaborative learning based on the first radiation safety accident history set, comprising: performing supervised learning on the first radiation safety accident history set based on multiple edge nodes to obtain multiple radiation safety accident predictors; transmitting the first radiation safety accident history set and the multiple radiation safety accident predictors to cloud nodes; randomly perturbing the first radiation safety accident history set through the cloud nodes to obtain a first perturbed radiation safety accident set; testing the multiple radiation safety accident predictors based on the first perturbed radiation safety accident set to obtain multiple accident prediction loss sets; performing attention enhancement on the first perturbed radiation safety accident set based on the multiple accident prediction loss sets to obtain multiple enhanced radiation safety accident spaces; and performing aggregate learning on the multiple radiation safety accident predictors based on the multiple enhanced radiation safety accident spaces to generate the first radiation safety accident prediction channel.

[0084] Furthermore, the present invention also provides a method for predicting radiation safety accidents in the second radiation environment space based on the second environmental monitoring sequence to obtain second radiation safety accident features, comprising: obtaining a second radiation safety correlation evaluation sequence by performing a radiation safety correlation evaluation on the second environmental monitoring sequence; performing attention enhancement on the second environmental monitoring sequence based on the second radiation safety correlation evaluation sequence to obtain a second environmental monitoring enhancement vector; performing a radiation safety accident history retrieval on the second radiation environment space to obtain a second radiation safety accident history set; performing edge-cloud collaborative learning based on the second radiation safety accident history set to build a second radiation safety accident prediction channel; and inputting the second environmental monitoring enhancement vector into the second radiation safety accident prediction channel to generate the second radiation safety accident features.

[0085] Furthermore, the present invention also provides a method for coupled prediction of radiation safety accidents based on the first radiation safety accident characteristics and the second radiation safety accident characteristics, and for establishing a radiation safety accident atlas, comprising: performing three-dimensional modeling of the first radiation environment space and the second radiation environment space to obtain a radiation environment space model; performing boundary smoothing enhancement based on the radiation environment space model to obtain a radiation environment enhancement model; performing radiation diffusion analysis on the first radiation safety accident characteristics based on the radiation environment enhancement model to determine a first accident radiation diffusion characteristic; performing radiation diffusion analysis on the second radiation safety accident characteristics based on the radiation environment enhancement model to obtain a second accident radiation diffusion characteristic; performing coupled prediction of radiation safety accidents based on the first accident radiation diffusion characteristic and the second accident radiation diffusion characteristic based on the radiation environment enhancement model to obtain a third radiation safety accident characteristic; and generating the radiation safety accident atlas based on the first radiation safety accident characteristics, the second radiation safety accident characteristics, and the third radiation safety accident characteristics.

[0086] Furthermore, the present invention also provides online monitoring of the first radiation environment space and the second radiation environment space through an Internet of Things (IoT) device group to obtain a first radiation environment monitoring set and a second radiation environment monitoring set, comprising: reading the IoT device group to obtain a multidimensional monitoring dataset; and cleaning and classifying the multidimensional monitoring dataset to obtain the first radiation environment monitoring set and the second radiation environment monitoring set.

[0087] Furthermore, the present invention also provides a method for generating a radiation safety early warning signal based on the aforementioned radiation safety accident map and a radiation alarm device.

[0088] This manual uses a progressive approach, focusing on the differences from other parts. The foregoing... Figure 1The IoT-based online radiation environment monitoring method and its specific content also apply to the IoT-based online radiation environment monitoring system described in this section. Through the foregoing detailed description of the IoT-based online radiation environment monitoring method, those skilled in the art can clearly understand the IoT-based online radiation environment monitoring system; therefore, for the sake of brevity, it will not be described in detail here. For the publicly disclosed system, since it corresponds to the publicly disclosed method, the description is relatively simple; relevant details can be found in the method section.

[0089] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0090] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for online monitoring of radiation environment based on the Internet of Things, characterized in that, The method includes: The space where the radiation device is located is defined as the first radiation environment space, and the neighborhood space of the first radiation environment space is defined as the second radiation environment space. The first radiation environment space and the second radiation environment space are monitored online by a group of Internet of Things devices to obtain a first radiation environment monitoring set and a second radiation environment monitoring set. Based on the IoT device group, the first radiation environment monitoring set and the second radiation environment monitoring set are respectively subjected to monitoring reliability correction to construct a first environmental monitoring sequence and a second environmental monitoring sequence. Based on the first environmental monitoring sequence, radiation safety accidents are predicted for the first radiation environment space to obtain the characteristics of the first radiation safety accident. Based on the second environmental monitoring sequence, radiation safety accidents are predicted for the second radiation environment space to obtain the characteristics of the second radiation safety accident. Radiation safety accident coupled prediction is performed based on the first radiation safety accident characteristics and the second radiation safety accident characteristics to establish a radiation safety accident map.

2. The method for online monitoring of radiation environment based on the Internet of Things as described in claim 1, characterized in that, Based on the IoT device group, monitoring reliability calibration is performed on the first radiation environment monitoring set and the second radiation environment monitoring set, including: Based on each monitoring data in the first radiation environment monitoring set, the IoT device group is subjected to device sensing accompanying parameter collection to obtain multiple device sensing accompanying data. The monitoring reliability evaluation is performed on the accompanying data of the multiple devices to obtain multiple monitoring reliability evaluation coefficients; Outlier detection is performed on the multiple monitoring credibility evaluation coefficients based on the monitoring credibility evaluation constraints to determine the monitoring credibility evaluation outlier clusters. The first radiation environment monitoring set is reliably corrected based on the monitoring reliability evaluation outcry, and the first environmental monitoring sequence is generated.

3. The method for online monitoring of radiation environment based on the Internet of Things as described in claim 2, characterized in that, The monitoring reliability evaluation is performed on the accompanying data from the multiple devices to obtain multiple monitoring reliability evaluation coefficients, including: Anomalies are identified based on the accompanying data from the multiple devices, and each accompanying abnormal segment of the sensor is determined. Based on multiple data experts' monitoring and credibility evaluation of multiple historical sensing-related abnormal segments, multiple monitoring credibility evaluation sequences were obtained. The central value is calculated based on each monitoring confidence evaluation sequence to obtain the confidence central value of each monitoring; The multiple historical sensing-related abnormal segments and the values ​​in each monitoring credibility set are mapped and aligned to generate a monitoring credibility evaluation learning set. A monitoring credibility evaluator is established by integrating and training the monitoring credibility evaluation learning set using multiple learners. Each sensor-associated abnormal segment is input into the monitoring reliability evaluator to generate the multiple monitoring reliability evaluation coefficients.

4. The method for online monitoring of radiation environment based on the Internet of Things as described in claim 1, characterized in that, Based on the first environmental monitoring sequence, a radiation safety accident prediction is performed on the first radiation environment space to obtain the characteristics of the first radiation safety accident, including: A radiation safety correlation assessment is performed based on the first environmental monitoring sequence to obtain the first radiation safety correlation assessment sequence. The first environmental monitoring sequence is enhanced with attention based on the first radiation safety correlation evaluation sequence to obtain the first environmental monitoring enhancement vector. Based on the first radiation environment space, a historical retrieval of radiation safety accidents is performed to obtain the first historical set of radiation safety accidents. Based on the first set of historical radiation safety accidents, edge-cloud collaborative learning is performed to build a first radiation safety accident prediction channel. The first environmental monitoring enhancement vector is input into the first radiation safety accident prediction channel to generate the first radiation safety accident feature.

5. The method for online monitoring of radiation environment based on the Internet of Things as described in claim 4, characterized in that, Based on the first set of historical radiation safety accidents, edge-cloud collaborative learning is performed to build a first radiation safety accident prediction channel, including: Multiple radiation safety accident predictors are obtained by performing supervised learning on the first set of historical radiation safety accidents using multiple edge nodes. Transmit the first set of historical radiation safety incidents and the plurality of radiation safety incident predictors to the cloud node; The first set of historical radiation safety accidents is randomly perturbed by the cloud node to obtain the first perturbed set of radiation safety accidents. Test the multiple radiation safety accident predictors according to the first set of disturbed radiation safety accidents to obtain multiple accident prediction loss sets; Based on the multiple accident prediction loss sets, attention enhancement is performed on the first disturbance radiation safety accident set to obtain multiple enhanced radiation safety accident spaces. The multiple radiation safety accident predictors are aggregated and learned based on the multiple enhanced radiation safety accident spaces to generate the first radiation safety accident prediction channel.

6. The method for online monitoring of radiation environment based on the Internet of Things as described in claim 1, characterized in that, Based on the second environmental monitoring sequence, radiation safety accidents are predicted for the second radiation environment space to obtain the characteristics of the second radiation safety accident, including: A second radiation safety correlation evaluation sequence is obtained by performing a radiation safety correlation evaluation on the second environmental monitoring sequence; Based on the second radiation safety correlation evaluation sequence, the second environmental monitoring sequence is enhanced with attention to obtain the second environmental monitoring enhancement vector. A historical retrieval of radiation safety accidents was performed on the second radiation environment space to obtain a second set of historical radiation safety accidents. Based on the second set of historical radiation safety accidents, edge-cloud collaborative learning is used to build a second radiation safety accident prediction channel. The second environmental monitoring enhancement vector is input into the second radiation safety accident prediction channel to generate the second radiation safety accident feature.

7. The method for online monitoring of radiation environment based on the Internet of Things as described in claim 1, characterized in that, Based on the first radiation safety accident characteristics and the second radiation safety accident characteristics, a coupled prediction of radiation safety accidents is performed to establish a radiation safety accident map, including: Three-dimensional modeling is performed on the first radiation environment space and the second radiation environment space to obtain a radiation environment space model; Based on the aforementioned radiation environment spatial model, boundary smoothing enhancement is performed to obtain a radiation environment enhancement model; Based on the radiation environment enhancement model, the radiation diffusion characteristics of the first radiation safety accident are analyzed to determine the radiation diffusion characteristics of the first accident. The radiation diffusion characteristics of the second radiation safety accident are analyzed based on the radiation environment enhancement model to obtain the radiation diffusion characteristics of the second accident. Based on the radiation environment enhancement model, radiation safety accident coupling prediction is performed on the radiation diffusion characteristics of the first accident and the radiation diffusion characteristics of the second accident to obtain the third radiation safety accident characteristics. The radiation safety accident map is generated based on the first radiation safety accident characteristics, the second radiation safety accident characteristics, and the third radiation safety accident characteristics.

8. The method for online monitoring of radiation environment based on the Internet of Things as described in claim 1, characterized in that, Online monitoring of the first and second radiation environment spaces is conducted using a cluster of IoT devices to obtain a first radiation environment monitoring set and a second radiation environment monitoring set, including: Read the IoT device group to obtain a multi-dimensional monitoring dataset; The multidimensional monitoring dataset is cleaned and classified to obtain the first radiation environment monitoring set and the second radiation environment monitoring set.

9. The method for online monitoring of radiation environment based on the Internet of Things as described in claim 1, characterized in that, Based on the radiation safety accident map, a radiation safety early warning signal is generated according to the radiation alarm device.

10. An online radiation environment monitoring system based on the Internet of Things, characterized in that, The system is used to perform the Internet of Things-based online radiation environment monitoring method according to any one of claims 1 to 9, the system comprising: The radiation environment space construction module is used to take the space where the radiation device is located as the first radiation environment space and the neighborhood space of the first radiation environment space as the second radiation environment space. The Internet of Things (IoT) monitoring data acquisition module is used to perform online monitoring of the first radiation environment space and the second radiation environment space through a group of IoT devices, and to obtain a first radiation environment monitoring set and a second radiation environment monitoring set. The monitoring reliability correction module is used to perform monitoring reliability correction on the first radiation environment monitoring set and the second radiation environment monitoring set according to the IoT device group, and construct the first environmental monitoring sequence and the second environmental monitoring sequence. The first radiation safety accident prediction module is used to predict radiation safety accidents in the first radiation environment space based on the first environmental monitoring sequence and obtain the characteristics of the first radiation safety accident. The second radiation safety accident prediction module is used to predict radiation safety accidents in the second radiation environment space based on the second environmental monitoring sequence and obtain the characteristics of the second radiation safety accident. The coupled prediction and map construction module is used to perform coupled prediction of radiation safety accidents based on the first radiation safety accident characteristics and the second radiation safety accident characteristics, and to establish a radiation safety accident map.

Images