A method and system for multi-mode joint detection of hazardous chemical trace radiation

Abstract

The application provides a kind of irradiation dangerous chemical trace multi-mode joint detection method and system, it is related to the intelligent monitoring technical field of dangerous chemical trace detection, including obtaining the three-way original response signal of electrochemical impedance, surface enhanced raman and plasma assisted ionization mass spectrometry arranged in irradiation environment, obtain multi-modal feature matrix;Core tensor and factor matrix are obtained by Tucker decomposition, after vectorization after explainability screening, form low-dimensional fusion feature vector;With low-dimensional fusion feature vector as mutual information to construct attribute graph, embedded through three-layer graph attention network, output trace dangerous chemical category label and concentration estimate;The category label and concentration estimate are substituted into ERPG parameter, to generate digital twin risk atlas;Through federal learning consensus risk map, drive ventilation, isolation, spray executor to complete closed-loop response.The application provides an efficient and reliable technical solution for dangerous chemical trace detection and safety control under irradiation environment.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring technology for hazardous chemicals, and more specifically, to a method and system for multi-mode joint detection of trace amounts of irradiated hazardous chemicals. Background Technology

[0002] In irradiated environments such as nuclear facilities and radiation processing, trace detection of hazardous chemicals (such as volatile organic compounds and toxic gases) is crucial for ensuring personnel safety, equipment stability, and environmental protection. Current detection technologies are gradually developing towards multi-parameter fusion and real-time response, but still face many technical bottlenecks. On the one hand, traditional single-modal detection technologies (such as independent electrochemical sensor detection and offline gas chromatography-mass spectrometry) have significant limitations: electrochemical sensors are easily affected by radiation dose, exhibiting severe signal drift and amplitude errors often exceeding 10% in high-dose environments (≥10 kGy / h); while offline chromatography-mass spectrometry offers high detection accuracy, its complex sample pretreatment and long detection cycles (ranging from minutes to hours) fail to meet real-time early warning requirements; Raman spectroscopy is susceptible to baseline drift and fluorescence interference, with characteristic peak extraction accuracy for trace substances (concentrations below 1 ppm) less than 60%. On the other hand, existing multimodal fusion solutions mostly adopt simple data splicing or weighted summation methods, without considering the time asynchronicity and modal specificity of different modal signals, and lack dynamic correction mechanisms for interference factors such as temperature, humidity and dose rate under irradiation environment. This results in weak anti-interference ability of detection results and low category recognition accuracy (usually below 85%). At the same time, it is impossible to achieve real-time visualization and diffusion prediction of risk fields, making it difficult to meet the closed-loop control requirements in industrial scenarios. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for multi-mode joint detection of trace amounts of irradiated hazardous chemicals, thereby improving the aforementioned problems. To achieve the above objective, the technical solution adopted by this invention is as follows:

[0004] Firstly, this application provides a method for multi-mode detection of trace amounts of irradiated hazardous chemicals, including:

[0005] The original response signals of electrochemical impedance spectroscopy, surface-enhanced Raman spectroscopy, and plasma-assisted ionization mass spectrometry, which are arranged in an irradiated environment, are obtained. After passing through Wiener filter, airPLS baseline correction and mode-specific characteristic peak fitting, the output is time-synchronized and denoised and normalized to obtain the multimodal feature matrix.

[0006] The multimodal feature matrix is ​​constructed as a three-dimensional tensor, and the core tensor and factor matrix are obtained by Tucker decomposition. After interpretability screening, the vectorization is used to form a low-dimensional fusion feature vector.

[0007] An attribute graph is constructed using low-dimensional fused feature vectors as mutual information. After being embedded through a three-layer graph attention network, the trace hazardous chemical category label and concentration estimate are output.

[0008] Substitute the category labels and concentration estimates into the ERPG parameters, combine them with real-time temperature, humidity, airflow and dose rate, calculate the spatial risk field according to the physical response surface model and generate a digital twin risk map.

[0009] By using a federated learning consensus risk map and a convolutional long short-term memory network to predict risk diffusion in the next 30 seconds, the system outputs standardized linkage control signals that are sent through the industrial interoperability standard and protocol after comparing thresholds. This signals then drive ventilation, isolation, and spray actuators to complete a closed-loop response.

[0010] Preferably, the coefficients of the Wiener filter are dynamically updated in the following manner: the unloaded signal without a sample is pre-acquired in the same irradiation environment, and the noise power spectrum is estimated online.

[0011] Preferably, the modality-specific feature extraction includes:

[0012] The electrochemical impedance spectroscopy module obtains the charge transfer resistance and double-layer capacitance by fitting the equivalent circuit Randall equivalent circuit model, which is denoted as the first feature.

[0013] The Raman spectroscopy submodule employs adaptive peak fitting, outputting peak position, full width at half maximum (FWHM), and relative intensity, denoted as the second feature.

[0014] The mass spectrometry submodule uses integrated ion current and mass spectrometry entropy as features, denoted as the third feature;

[0015] The first, second, and third features are combined to form a modality-specific feature.

[0016] Preferably, the core tensor and factor matrix are obtained through Tucker decomposition, wherein the Tucker decomposition adopts an online randomized Tucker algorithm, and the core tensor dimension is adaptively selected by the Bayesian Information Criterion (BIC) to meet the real-time requirements of the embedded edge computing platform.

[0017] Preferably, the mutual information in the attribute graph constructed using low-dimensional fused feature vectors is calculated using the Kraskov estimation method.

[0018] Preferably, the graph neural network embedded in the three-layer graph attention network is a deep learning model composed of three stacked graph attention mechanisms, with 4 attention heads per layer, a random inactivation rate of 20%, an activation function of 0.2, and a skip knowledge mechanism in the last layer to fuse information from nodes at different depths.

[0019] Preferably, the step of substituting category labels and concentration estimates into the ERPG parameters, combining real-time temperature, humidity, airflow, and dose rate, and calculating the spatial risk field according to the physical response surface model to generate a digital twin risk map includes:

[0020] Read the hazardous chemical category and concentration value one by one, and match them with the corresponding emergency response plan guidelines concentration and risk coefficient to form a toxicity parameter vector;

[0021] The toxicity parameter vector, along with the real-time measured temperature, humidity, airflow velocity, and dose rate, is fed into the quadratic response surface model to calculate the environmental correction factor.

[0022] Spatial risk values ​​are calculated point-by-point using environmental correction factors and dose rates according to a calculation formula.

[0023] The calculated spatial risk values ​​are incorporated into a three-dimensional array using 0.5m cubic grids. Isosurface extraction and region segmentation are then performed to construct a digital twin risk map.

[0024] Secondly, this application also provides a multimodal graphite component internal damage assessment system, including:

[0025] Acquisition and processing module: It is used to synchronously acquire five-modal data, and after spatiotemporal alignment and standardization, generate a standardized multimodal data stream, in which the five-modal data include ultrasound, acoustic emission, electrical impedance, temperature and neutron flux;

[0026] The building module is used to extract acoustic and electro-thermal physical features from multimodal data streams and perform adaptive weighted fusion based on the principle of maximum relevant information to construct a multidimensional damage-sensitive feature vector.

[0027] Inversion Extraction Module: This module is used to input multidimensional damage-sensitive feature vectors into a deep learning network that has been corrected by a physical constraint layer, in order to invert and reconstruct a high-resolution three-dimensional crack probability field and extract crack quantization parameters from it.

[0028] The calculation module is used to input crack quantification parameters into the digital twin-driven irradiation-creep-fatigue coupled damage model, predict crack propagation trajectory in real time, and then calculate the remaining life and failure probability of the overall structure.

[0029] Generation module: Used to determine the warning level based on the remaining life and failure probability, and combined with the spatial coordinate information of the three-dimensional crack probability field, to generate and optimize preventive maintenance instructions and schedules for high-risk modular components.

[0030] Thirdly, this application also provides a multi-mode detection device for trace amounts of irradiated hazardous chemicals, comprising:

[0031] Memory, used to store computer programs;

[0032] A processor is used to implement the steps of the multi-mode detection method for trace amounts of irradiated hazardous chemicals when executing the computer program.

[0033] Fourthly, this application also provides a readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method for multi-mode detection of trace amounts of irradiated hazardous chemicals.

[0034] The beneficial effects of this invention are as follows:

[0035] This invention deeply integrates three sensing modes: electrochemical impedance spectroscopy, surface-enhanced Raman scattering, and plasma-assisted ionization mass spectrometry. Through cutting-edge signal processing and artificial intelligence algorithms, it achieves intelligent linkage from data acquisition to closed-loop control. Specifically, this invention first uses adaptive filters and the airPLS baseline correction algorithm to specifically suppress irradiation noise and signal drift. Then, it utilizes tensor decomposition and graph neural network technology to achieve efficient fusion and accurate analysis of multi-source heterogeneous information, ultimately outputting the category and concentration of hazardous chemicals. Building upon this, this invention innovatively introduces a physical mechanism-based dynamic risk field calculation model and digital twin map mapping technology, and leverages a federated learning framework to achieve distributed decision consensus, thereby driving actuators to complete automatic and rapid emergency response. This method effectively overcomes the shortcomings of existing technologies, such as weak anti-interference capability, low identification efficiency, and system isolation, achieving highly reliable, real-time, and intelligent joint detection and disposal of various trace hazardous chemicals under extreme irradiation environments. Other features and advantages of this invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the embodiments of the invention. The objects and other advantages of the present invention can be realized and obtained by means of the structures particularly pointed out in the written description, claims and drawings. Attached Figure Description

[0036] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 This is a schematic diagram of the multi-mode joint detection method for trace amounts of irradiated hazardous chemicals described in this embodiment of the invention;

[0038] Figure 2 This is a schematic diagram of the multi-mode joint detection system for trace amounts of irradiated hazardous chemicals described in this embodiment of the invention;

[0039] Figure 3This is a schematic diagram of the structure of the multi-mode joint inspection equipment for trace amounts of irradiated hazardous chemicals described in an embodiment of the present invention. In the figure: 701, acquisition module; 702, decomposition module; 703, output module; 704, calculation module; 705, drive module; 800, multi-mode joint inspection equipment for trace amounts of irradiated hazardous chemicals; 801, processor; 802, memory; 803, multimedia component; 804, I / O interface; 805, communication component. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0041] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0042] Example 1:

[0043] This embodiment provides a multi-mode detection method for trace amounts of irradiated hazardous chemicals, which can be used for trace detection of hazardous chemicals.

[0044] S100: Acquire the three raw response signals of electrochemical impedance spectroscopy, surface-enhanced Raman spectroscopy, and plasma-assisted ionization mass spectrometry arranged in the irradiation environment. After passing through Wiener filter, airPLS baseline correction, and mode-specific characteristic peak fitting, the output is time-synchronized and denoised to obtain the multimodal feature matrix.

[0045] The core tensor and factor matrix are obtained through Tucker decomposition. The Tucker decomposition adopts the online random Tucker algorithm, and the core tensor dimension is adaptively selected by the Bayesian Information Criterion (BIC) to meet the real-time requirements of the embedded edge computing platform.

[0046] The mutual information in the attribute graph constructed using low-dimensional fused feature vectors is calculated using the Kraskov estimation method. The formula for calculating the adjacency threshold is as follows:

[0047]

[0048] In the formula, θ The adjacency threshold when constructing the mutual information graph. The average of the mutual information of all feature pairs. Let be the standard deviation of the mutual information of all feature pairs; where and The mean and standard deviation of the mutual information of all feature pairs are used to construct a sparse graph and reduce the computational load of GNN;

[0049] The process involves substituting category labels and concentration estimates into the ERPG parameters, combining real-time temperature, humidity, airflow, and dose rate, and calculating the spatial risk field using a physical response surface model to generate a digital twin risk map. This includes:

[0050] Read the hazardous chemical category and concentration value one by one, and match them with the corresponding emergency response plan guidelines concentration and risk coefficient to form a toxicity parameter vector;

[0051] The toxicity parameter vector, along with the real-time measured temperature, humidity, airflow velocity, and dose rate, is fed into the quadratic response surface model to calculate the environmental correction factor.

[0052] Spatial risk values ​​are calculated point-by-point using environmental correction factors and dose rates according to the following formula:

[0053]

[0054] In the formula, For risk value field, For concentration, As an environmental correction factor, The concentration limit for the emergency response plan guidelines for the i-th substance is... Let be the inherent toxicity / risk coefficient of the i-th substance. Where D is the radiation enhancement factor and D is the real-time dose rate at the site;

[0055] The calculated spatial risk values ​​are incorporated into a three-dimensional array using 0.5m cubic grids. Isosurface extraction and region segmentation are then performed to construct a digital twin risk map.

[0056] S200. The multimodal feature matrix is ​​constructed into a three-dimensional tensor. The core tensor and factor matrix are obtained by Tucker decomposition. After interpretability screening, the vectorization is used to form a low-dimensional fused feature vector.

[0057] S300 constructs an attribute graph using low-dimensional fused feature vectors as mutual information, embeds it through a three-layer graph attention network, and outputs trace hazardous chemical category labels and concentration estimates.

[0058] S400: Substitute the category label and concentration estimate into the ERPG parameters, combine real-time temperature, humidity, airflow and dose rate, calculate the spatial risk field according to the physical response surface model and generate a digital twin risk map.

[0059] The S500 uses a federated learning consensus risk map and a convolutional long short-term memory network to predict the risk diffusion in the next 30 seconds. After comparing the threshold, it outputs a standardized linkage control signal that meets industrial interoperability standards and is sent through the protocol, thereby driving ventilation, isolation, and spray actuators to complete a closed-loop response.

[0060] The coefficients of the Wiener filter are dynamically updated in the following manner:

[0061] The no-sample signal was pre-acquired in the same irradiation environment, and the noise power spectrum was estimated online.

[0062] The modality-specific feature extraction includes:

[0063] The electrochemical impedance spectroscopy module obtains the charge transfer resistance and double-layer capacitance by fitting the equivalent circuit Randall equivalent circuit model, which is denoted as the first feature.

[0064] The Raman spectroscopy submodule employs adaptive peak fitting, outputting peak position, full width at half maximum (FWHM), and relative intensity, denoted as the second feature.

[0065] The mass spectrometry submodule uses integrated ion current and mass spectrometry entropy as features, denoted as the third feature;

[0066] The first, second, and third features are combined to form a modality-specific feature.

[0067] The core tensor and factor matrix are obtained through Tucker decomposition, where the Tucker decomposition adopts an online randomized Tucker algorithm, and the core tensor dimension is adaptively selected by the Bayesian Information Criterion (BIC) to meet the real-time requirements of the embedded edge computing platform.

[0068] The mutual information in the attribute graph constructed using low-dimensional fused feature vectors as mutual information is calculated using the Kraskov estimation method.

[0069] The graph neural network embedded in the three-layer graph attention network is a deep learning model composed of three stacked graph attention mechanisms. Each layer has 4 attention heads, a random inactivation rate of 20%, an activation function of 0.2, and a skip knowledge mechanism is used in the last layer to fuse information from nodes at different depths.

[0070] The process involves substituting category labels and concentration estimates into the ERPG parameters, combining real-time temperature, humidity, airflow, and dose rate, and calculating the spatial risk field using a physical response surface model to generate a digital twin risk map. This includes:

[0071] Read the hazardous chemical category and concentration value one by one, and match them with the corresponding emergency response plan guidelines concentration and risk coefficient to form a toxicity parameter vector;

[0072] The toxicity parameter vector, along with the real-time measured temperature, humidity, airflow velocity, and dose rate, is fed into the quadratic response surface model to calculate the environmental correction factor.

[0073] Spatial risk values ​​are calculated point by point using environmental correction factors and dose rates according to the calculation formula. The calculated spatial risk values ​​are then included in a three-dimensional array using 0.5m cubic grids. Isosurface extraction and region segmentation are performed to construct a digital twin risk map.

[0074] Example 2:

[0075] like Figure 2 As shown, this embodiment provides a multi-mode joint detection system for trace amounts of irradiated hazardous chemicals. See [link to documentation]. Figure 2 The system includes:

[0076] Acquisition module 701: used to acquire the three raw response signals of electrochemical impedance, surface-enhanced Raman and plasma-assisted ionization mass spectrometry arranged in the irradiation environment. After Wiener filter, airPLS baseline correction and mode-specific characteristic peak fitting, the output is time-synchronized and denoised to obtain the multimodal feature matrix.

[0077] Decomposition module 702: Used to construct a three-dimensional tensor from the multimodal feature matrix, obtain the core tensor and factor matrix through Tucker decomposition, and vectorize it after interpretability screening to form a low-dimensional fused feature vector;

[0078] Output module 703: Used to construct an attribute graph with low-dimensional fused feature vectors as mutual information, and embedded through a three-layer graph attention network to output trace hazardous chemical category labels and concentration estimates;

[0079] Calculation module 704: Used to input category labels and concentration estimates into ERPG parameters, combine real-time temperature, humidity, airflow and dose rate, calculate the spatial risk field according to the physical response surface model and generate a digital twin risk map;

[0080] Drive module 705: It is used to predict the risk diffusion in the next 30 seconds by learning the consensus risk map through a convolutional long short-term memory network, compare the threshold, output the standardized linkage control signal sent through the industrial interoperability standard and protocol, and then drive the ventilation, isolation and spray actuators to complete the closed-loop response.

[0081] It should be noted that the specific methods by which each module performs operations in the system described in the above embodiments have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0082] Example 3:

[0083] Corresponding to the above method embodiments, this embodiment also provides a multi-mode joint inspection device for trace amounts of irradiated hazardous chemicals. The multi-mode joint inspection device for trace amounts of irradiated hazardous chemicals described below and the multi-mode joint inspection method for trace amounts of irradiated hazardous chemicals described above can be referred to in correspondence.

[0084] Figure 3 This is a block diagram illustrating a multi-mode detection device 800 for trace amounts of hazardous irradiated chemicals, according to an exemplary embodiment. Figure 3 As shown, the multi-mode inspection device 800 for trace amounts of irradiated hazardous chemicals includes a processor 801 and a memory 802. The device 800 also includes one or more of a multimedia component 803, an I / O interface 804, and a communication component 805.

[0085] The processor 801 controls the overall operation of the irradiated trace hazardous chemical multi-mode inspection device 800 to complete all or part of the steps in the aforementioned irradiated trace hazardous chemical multi-mode inspection method. The memory 802 stores various types of data to support the operation of the irradiated trace hazardous chemical multi-mode inspection device 800. This data may include, for example, instructions for any application or method operating on the irradiated trace hazardous chemical multi-mode inspection device 800, as well as application-related data such as contact data, sent and received messages, images, audio, video, etc. The memory 802 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touchscreen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in the memory 802 or transmitted via the communication component 805. The audio component also includes at least one speaker for outputting audio signals. I / O interface 804 provides an interface between processor 801 and other interface modules, such as keyboards, mice, or buttons. These buttons can be virtual or physical. Communication component 805 is used for wired or wireless communication between the irradiated hazardous chemical trace multi-mode inspection equipment 800 and other devices. Wireless communication includes, for example, Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination thereof. Therefore, the corresponding communication component 805 may include a Wi-Fi module, a Bluetooth module, or an NFC module.

[0086] In an exemplary embodiment, the irradiated hazardous chemical trace multi-mode inspection device 800 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above-described irradiated hazardous chemical trace multi-mode inspection method.

[0087] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, which, when executed by a processor, implement the steps of the above-described method for multi-mode detection of trace amounts of irradiated hazardous chemicals. For example, the computer-readable storage medium may be the memory 802 including the program instructions, which may be executed by the processor 801 of the multi-mode detection device 800 for irradiated hazardous chemicals to complete the above-described method for multi-mode detection of trace amounts of irradiated hazardous chemicals.

[0088] Example 4:

[0089] Corresponding to the above method embodiments, this embodiment also provides a readable storage medium. The readable storage medium described below can be referred to in conjunction with the above-described method for multi-mode detection of trace amounts of irradiated hazardous chemicals.

[0090] A computer program is stored on a readable storage medium, and when the computer program is executed by a processor, it implements the steps of the multi-mode detection method for trace amounts of irradiated hazardous chemicals described in the above method embodiments.

[0091] Specifically, the readable storage medium can be a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, or any other readable storage medium capable of storing program code.

[0092] This invention overcomes existing technological bottlenecks through an integrated design of multimodal signal collaborative processing, intelligent feature fusion, and dynamic risk assessment. First, it integrates three signals: electrochemical impedance spectroscopy, surface-enhanced Raman spectroscopy, and plasma-assisted ionization mass spectrometry. It employs Wiener filters for dynamic denoising, airPLS baseline correction, and modality-specific feature peak fitting to achieve time synchronization and denoising normalization. Next, it reduces the dimensionality of the multimodal feature matrix using Tucker decomposition and adaptively selects the core tensor dimension using Bayesian information criteria, balancing real-time performance and feature integrity. Then, it calculates feature mutual information using the Kraskov estimation method to construct an attribute map and utilizes a three-layer graph attention network to achieve high-precision category recognition and concentration estimation. Finally, it combines ERPG parameters and a physical response surface model, fusing temperature, humidity, airflow, and dose rate to construct a digital twin risk map. Through federated learning and convolutional long short-term memory networks, it predicts risk diffusion and outputs standardized linkage control signals to drive the actuator's closed-loop response.

[0093] The method employed in this invention can maintain an amplitude error of ≤3% under high-dose environments of ≥10 kGy / h, improve the category identification accuracy to over 95%, and simultaneously achieve risk diffusion prediction within 30 seconds, providing an efficient and reliable technical solution for the detection and safety management of trace hazardous chemicals under irradiation environments.

[0094] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0095] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for multi-mode joint detection of trace amounts of irradiated hazardous chemicals, characterized in that, include: The original response signals of electrochemical impedance spectroscopy, surface-enhanced Raman spectroscopy, and plasma-assisted ionization mass spectrometry, which are arranged in an irradiated environment, are obtained. After passing through Wiener filter, airPLS baseline correction and mode-specific characteristic peak fitting, the output is time-synchronized and denoised and normalized to obtain the multimodal feature matrix. The multimodal feature matrix is ​​constructed as a three-dimensional tensor, and the core tensor and factor matrix are obtained by Tucker decomposition. After interpretability screening, the vectorization is used to form a low-dimensional fusion feature vector. An attribute graph is constructed using low-dimensional fused feature vectors as mutual information. After being embedded through a three-layer graph attention network, the trace hazardous chemical category label and concentration estimate are output. Substitute the category labels and concentration estimates into the ERPG parameters, combine them with real-time temperature, humidity, airflow and dose rate, calculate the spatial risk field according to the physical response surface model and generate a digital twin risk map. By using a federated learning consensus risk map and a convolutional long short-term memory network to predict the risk diffusion in the next 30 seconds, the risk is compared with the threshold and then outputs a standardized linkage control signal that is sent through the industrial interoperability standard and protocol, thereby driving the ventilation, isolation and spray actuators to complete the closed-loop response. The core tensor and factor matrix are obtained through Tucker decomposition. The Tucker decomposition adopts the online random Tucker algorithm, and the core tensor dimension is adaptively selected by the Bayesian Information Criterion (BIC) to meet the real-time requirements of the embedded edge computing platform. The mutual information in the attribute graph constructed using low-dimensional fused feature vectors is calculated using the Kraskov estimation method. The formula for calculating the adjacency threshold is as follows: In the formula, θ The adjacency threshold when constructing the mutual information graph. The average of the mutual information of all feature pairs. Let be the standard deviation of the mutual information of all feature pairs; where and The mean and standard deviation of the mutual information of all feature pairs are used to construct a sparse graph and reduce the computational load of GNN; The process involves substituting category labels and concentration estimates into the ERPG parameters, combining real-time temperature, humidity, airflow, and dose rate, and calculating the spatial risk field using a physical response surface model to generate a digital twin risk map. This includes: Read the hazardous chemical category and concentration value one by one, and match them with the corresponding emergency response plan guidelines concentration and risk coefficient to form a toxicity parameter vector; The toxicity parameter vector, along with the real-time measured temperature, humidity, airflow velocity, and dose rate, is fed into the quadratic response surface model to calculate the environmental correction factor. Spatial risk values ​​are calculated point-by-point using environmental correction factors and dose rates according to the following formula: In the formula, For risk value field, For concentration, As an environmental correction factor, The concentration limit for the emergency response plan guidelines for the i-th substance is... Let be the inherent toxicity / risk coefficient of the i-th substance. Where D is the radiation enhancement factor and D is the real-time dose rate at the site; The calculated spatial risk values ​​are incorporated into a three-dimensional array using 0.5m cubic grids. Isosurface extraction and region segmentation are then performed to construct a digital twin risk map.

2. The method for multi-mode joint detection of trace amounts of irradiated hazardous chemicals according to claim 1, characterized in that, The coefficients of the Wiener filter are dynamically updated in the following manner: The empty signal without a sample was pre-acquired in the same irradiation environment, and the noise power spectrum was estimated online. The calculation formula is as follows: In the formula, For noise power spectral density, To estimate the noise power spectrum at ω at the previous time t-1, This represents the background radiation signal measured at ω at the current time.

3. The method for multi-mode joint detection of trace amounts of irradiated hazardous chemicals according to claim 1, characterized in that, The modality-specific feature extraction includes: The electrochemical impedance spectroscopy module obtains the charge transfer resistance and double-layer capacitance by fitting the equivalent circuit Randall equivalent circuit model, which is denoted as the first feature. The Raman spectroscopy submodule employs adaptive peak fitting, outputting peak position, full width at half maximum (FWHM), and relative intensity, denoted as the second feature. The mass spectrometry submodule uses integrated ion current and mass spectrometry entropy as features, denoted as the third feature; The first, second, and third features are combined to form a modality-specific feature.

4. The method for multi-mode joint detection of trace amounts of irradiated hazardous chemicals according to claim 1, characterized in that, The graph neural network embedded in the three-layer graph attention network is a deep learning model composed of three stacked graph attention mechanisms. Each layer has 4 attention heads, a random inactivation rate of 20%, an activation function of 0.2, and a skip knowledge mechanism is used in the last layer to fuse information from nodes at different depths.

5. A multi-mode joint detection system for trace amounts of irradiated hazardous chemicals, based on the multi-mode joint detection method for trace amounts of irradiated hazardous chemicals as described in claim 1, characterized in that, include: Acquisition module: used to acquire the raw response signals of electrochemical impedance, surface-enhanced Raman and plasma-assisted ionization mass spectrometry arranged in the irradiation environment. After Wiener filter, airPLS baseline correction and mode-specific characteristic peak fitting, the output is time-synchronized and denoised to obtain the multimodal feature matrix. Decomposition module: Used to construct a three-dimensional tensor from the multimodal feature matrix. After Tucker decomposition, the core tensor and factor matrix are obtained. After interpretability screening, the vectorized form a low-dimensional fused feature vector. Output module: Used to construct an attribute graph with low-dimensional fused feature vectors as mutual information, embedded through a three-layer graph attention network, and output trace hazardous chemical category labels and concentration estimates; The calculation module is used to input the category labels and concentration estimates into the ERPG parameters, combine them with real-time temperature, humidity, airflow and dose rate, calculate the spatial risk field according to the physical response surface model and generate a digital twin risk map. The drive module is used to predict the risk spread in the next 30 seconds by using a federated learning consensus risk map and a convolutional long short-term memory network. After comparing the threshold, it outputs a standardized linkage control signal that meets the industrial interoperability standard and is sent through the protocol, thereby driving the ventilation, isolation, and spray actuators to complete the closed-loop response. In the decomposition module, the core tensor and factor matrix are obtained through Tucker decomposition. The Tucker decomposition adopts the online random Tucker algorithm, and the core tensor dimension is adaptively selected by the Bayesian Information Criterion (BIC) to meet the real-time requirements of the embedded edge computing platform. The output module includes: constructing mutual information in the attribute graph using low-dimensional fused feature vectors as mutual information, calculating the mutual information using the Kraskov estimation method, and the formula for calculating the adjacency threshold is as follows: In the formula, θ The adjacency threshold when constructing the mutual information graph. The average of the mutual information of all feature pairs. Let be the standard deviation of the mutual information of all feature pairs; where and The mean and standard deviation of the mutual information of all feature pairs are used to construct a sparse graph and reduce the computational load of GNN; The calculation module includes: reading the hazardous chemical category and concentration value one by one, and matching them with the corresponding emergency response plan guidelines concentration and risk coefficient to form a toxicity parameter vector; The toxicity parameter vector, along with the real-time measured temperature, humidity, airflow velocity, and dose rate, is fed into the quadratic response surface model to calculate the environmental correction factor. Spatial risk values ​​are calculated point-by-point using environmental correction factors and dose rates according to the following formula: In the formula, For risk value field, For concentration, As an environmental correction factor, The concentration limit for the emergency response plan guidelines for the i-th substance is... Let be the inherent toxicity / risk coefficient of the i-th substance. Where D is the radiation enhancement factor and D is the real-time dose rate at the site; The calculated spatial risk values ​​are incorporated into a three-dimensional array using 0.5m cubic grids. Isosurface extraction and region segmentation are then performed to construct a digital twin risk map.

6. A multi-mode joint detection device for trace amounts of irradiated hazardous chemicals, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the multi-mode detection method for trace amounts of irradiated hazardous chemicals as described in any one of claims 1 to 4 when executing the computer program.

7. A readable storage medium, characterized in that: The readable storage medium stores a computer program that, when executed by a processor, implements the multi-mode detection method for trace amounts of irradiated hazardous chemicals as described in any one of claims 1 to 4.

Images