A method and system for assessing the state of a nuclear power plant main control room operator

Abstract

The present application relates to a kind of nuclear power plant main control room operator's state evaluation method and system, comprising: obtaining the physiological data of evaluation object;The physiological data is preprocessed, and preprocessing data is obtained;According to the type identification of preprocessing data, and according to the different types of preprocessing data, difference characteristic extraction is carried out, and multi-modal fusion feature data is obtained;On the basis of multi-modal fusion feature data, on-site personnel state evaluation algorithm model is constructed;According to on-site personnel state evaluation algorithm model, the state of current on-site personnel is evaluated, and personnel state evaluation result is output.The present application can accurately evaluate the stress, activity, emotional fluctuation level and other states of on-site personnel by constructing on-site personnel state evaluation algorithm model based on multi-modal fusion feature data, so as to identify the fatigue of nuclear power plant main control room operator in advance, reduce human error and improve the safety of unit operation.

Description

Technical Field

[0001] This invention relates to the technical field of personnel status assessment in nuclear power plants, and more specifically, to a method and system for assessing the status of operators in the main control room of a nuclear power plant. Background Technology

[0002] The safe operation of a nuclear power plant requires timely and accurate operation by personnel, making them crucial to the entire system. Since control room operators are required to remain constantly vigilant in front of the control panels, special attention must be paid to their mental state. In the context of digitalization, strengthening the monitoring of operator fatigue is particularly important, as timely detection of fatigue can reduce human error.

[0003] Current multimodal physiological signal acquisition can collect users' physiological signals in real time and send the data to a host computer via TCP protocol within a local area network. This allows the host computer to perform data analysis and processing, enabling customized wristbands to collect and transmit multiple physiological signals in real time, ensuring the system features data synchronization, low latency, high-precision acquisition, and stable transmission. It is possible to consider using customized wristbands to collect physiological data, perform data processing, and monitor employee fatigue. However, current host computers lack further preprocessing and feature extraction of the collected physiological data, making it impossible to build an on-duty employee status assessment algorithm model based on physiological data. This hinders the accurate assessment of employees' current stress, activity levels, and emotional fluctuations. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method and system for assessing the status of operators in the main control room of a nuclear power plant, addressing the problems existing in the prior art.

[0005] The technical solution adopted by this invention to solve its technical problem is: constructing a state assessment method for operators in the main control room of a nuclear power plant, comprising: Step S10: Obtain physiological data of the evaluation subject; Step S20: Preprocess the physiological data to obtain preprocessed data; Step S30: Perform type identification based on the preprocessed data, and extract differentiated features based on different types of preprocessed data to obtain multimodal fusion feature data; Step S40: Construct an on-duty personnel status assessment algorithm model based on the multimodal fusion feature data; Step S50: Evaluate the status of the current on-duty personnel according to the on-duty personnel status evaluation algorithm model, and output the personnel status evaluation results.

[0006] In the method for assessing the condition of operators in the main control room of a nuclear power plant according to the present invention, step S20 involves preprocessing the physiological data to obtain preprocessed data, including: The physiological data are subjected to bandpass filtering and median filtering to obtain the preprocessed data.

[0007] In the state assessment method for nuclear power plant main control room operators described in this invention, the different types of preprocessed data include: pulse data, skin conductance data, acceleration data, angular velocity data, and heart rate data; In step S30, type identification is performed based on the preprocessed data, and differential feature extraction is performed based on different types of preprocessed data to obtain multimodal fusion feature data, including: If the data is pulse data, then feature extraction is performed on the pulse data to obtain pulse feature data; If the data is skin conductance data, then feature extraction is performed on the skin conductance data to obtain skin conductance feature data; If the data is acceleration data, then feature extraction is performed on the acceleration data to obtain acceleration feature data; If the data is angular velocity, then feature extraction is performed on the angular velocity data to obtain angular velocity feature data; If the data is heart rate data, then feature extraction is performed on the heart rate data to obtain heart rate feature data; The pulse feature data, skin conductance feature data, acceleration feature data, angular velocity feature data, and heart rate feature data are fused to obtain the multimodal fused feature data.

[0008] In the nuclear power plant main control room operator status assessment method of the present invention, the step of extracting features from the pulse data to obtain pulse feature data includes: The pulse data is reconstructed using the maximum non-overlapping wavelet decomposition method to obtain post-processed data; Feature extraction is performed on the post-processed data to obtain the pulse feature data; The pulse characteristic data includes: mean RR interval, standard deviation of interval, average difference between adjacent RR intervals, percentage of consecutive RR intervals exceeding 20ms, percentage of consecutive RR intervals exceeding 50ms, absolute power in the very low frequency band, low frequency power, and high frequency power.

[0009] In the nuclear power plant main control room operator status assessment method of the present invention, step S40, which involves constructing an on-duty personnel status assessment algorithm model based on the multimodal fusion feature data, includes: Construct the structure of the on-the-job personnel status assessment algorithm model; Based on the multimodal fusion feature data, the feature inputs of the on-duty personnel status assessment algorithm model are constructed; The on-duty personnel status assessment algorithm model is trained; The trained model is validated until it meets the accuracy requirements, thus obtaining the on-duty personnel status assessment algorithm model.

[0010] In the nuclear power plant main control room operator status assessment method of the present invention, the feature inputs for constructing the on-duty personnel status assessment algorithm model based on the multimodal fusion feature data include: The multimodal fusion feature data is input in layers according to different types; Add location information to the features of different layers.

[0011] In the nuclear power plant main control room operator status assessment method of the present invention, the on-duty personnel status assessment algorithm model is a deep learning model based on the attention mechanism; The attention-based deep learning model includes an encoder and a decoder; The encoder is used to transform the input multimodal fusion features into a set of representation vectors; The decoder is used to transform the representation vector into the status level of the currently on-duty personnel and output the personnel status assessment result.

[0012] In the nuclear power plant main control room operator status assessment method of the present invention, step S50, which assesses the status of the currently on-duty personnel according to the on-duty personnel status assessment algorithm model and outputs the personnel status assessment results, includes: Obtain the real-time physiological data of the currently on-duty personnel; The real-time physiological data is preprocessed to obtain real-time preprocessed data; Based on the real-time preprocessed data, differential features are extracted to obtain real-time fused features; The on-duty personnel status assessment algorithm model is used to encode and identify the real-time fused features to obtain a representation vector. The representation vector is mapped to output the personnel status assessment result of the currently on-duty personnel.

[0013] In the nuclear power plant main control room operator status assessment method of the present invention, the step of extracting differential features based on the real-time preprocessed data to obtain real-time fused features includes: Extract the first single-modal feature that is strongly correlated with the state of the operator in the main control room of the nuclear power plant from the real-time preprocessed data; Extract the second single-modal features that are strongly correlated with the state of the operator in the main control room of the nuclear power plant under the target operating conditions; The first unimodal feature and the second unimodal feature are fused with multimodal physiological data to obtain the real-time fused feature.

[0014] This invention also provides a status assessment system for operators in the main control room of a nuclear power plant, comprising a host computer, the host computer including: The experimental data acquisition unit is used to acquire physiological data of the evaluation subjects; A data preprocessing unit is used to preprocess the physiological data to obtain preprocessed data; The feature extraction unit is used to perform type identification based on the preprocessed data and to perform differentiated feature extraction based on different types of preprocessed data to obtain multimodal fusion feature data. The evaluation model construction unit is used to construct an on-the-job personnel status evaluation algorithm model based on the multimodal fusion feature data; The personnel status assessment unit is used to assess the status of the current on-duty personnel according to the on-duty personnel status assessment algorithm model and output the personnel status assessment results.

[0015] The nuclear power plant main control room operator status assessment system described in this invention also includes: a customized wristband that communicates with the host computer; The customized wristband is used to collect real-time physiological data of on-duty personnel.

[0016] In the nuclear power plant main control room operator status assessment system described in this invention, the customized wristband includes: a sensor module, a data processing module, and a communication module; The sensor module is used to collect the real-time physiological data of the on-duty personnel. The data processing module is used to perform preliminary processing and formatting of the real-time physiological data; The communication module is used to transmit the real-time generated data to the host computer.

[0017] In the nuclear power plant main control room operator status assessment system described in this invention, the sensor module includes: a pulse sensor, a skin temperature sensor, an acceleration sensor, a gyroscope, a geomagnetic sensor, an EDA skin conductance sensor, an EMG electromyography sensor, an ambient temperature and humidity sensor, a light intensity sensor, and an altitude and air pressure sensor.

[0018] In the nuclear power plant main control room operator status assessment system described in this invention, the data transmission between the customized wristband and the host computer adopts TCP protocol or IP protocol; The data packets uploaded by the customized wristband to the host computer include: timestamp, sensor ID, data value, and verification code.

[0019] In the nuclear power plant main control room operator status assessment system described in this invention, the customized wristband uploads the real-time physiological data to the host computer according to the following steps: The customized wristband packages the real-time physiological data and uploads it to the host computer. After uploading the real-time physiological data to the host computer, it enters a waiting state. If no confirmation message is received from the host computer within the set waiting time, the real-time physiological data will be resent.

[0020] In the nuclear power plant main control room operator status assessment system described in this invention, the customized wristband is also used to perform the following operations: When the network connection is interrupted, the customized bracelet enters reconnection mode and caches data; Once the network connection is restored, the customized wristband automatically continues to upload data to the host computer.

[0021] The method and system for assessing the status of nuclear power plant control room operators according to the present invention have the following beneficial effects: It includes: acquiring physiological data of the assessment subject; preprocessing the physiological data to obtain preprocessed data; identifying the type of the preprocessed data and extracting differentiated features based on different types of preprocessed data to obtain multimodal fusion feature data; constructing an on-duty personnel status assessment algorithm model based on the multimodal fusion feature data; assessing the status of the currently on-duty personnel based on the on-duty personnel status assessment algorithm model, and outputting the personnel status assessment results. This invention, by constructing an on-duty personnel status assessment algorithm model based on multimodal fusion feature data, can accurately assess the current stress, activity level, and emotional fluctuation levels of on-duty personnel, thereby identifying fatigue conditions of nuclear power plant control room operators in advance, reducing human error, and improving unit operation safety. Attached Figure Description

[0022] The present invention will be further described below with reference to the accompanying drawings and embodiments. In the accompanying drawings: Figure 1 This is a flowchart illustrating the status assessment method for operators in the main control room of a nuclear power plant provided by the present invention. Figure 2 This is a schematic diagram of the structure of the on-the-job personnel status assessment algorithm model provided by the present invention; Figure 3 This is a schematic diagram illustrating the multi-head attention mechanism introduced into the on-the-job personnel status assessment algorithm model provided by this invention; Figure 4 This is a system architecture diagram for assessing the status of operators in the main control room of a nuclear power plant, provided by the present invention. Detailed Implementation

[0023] 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 embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] This invention proposes a method and system for assessing the status of nuclear power plant main control room operators based on physiological data collected by a customized wristband. The system collects physiological data through the customized wristband and transmits it to a host computer. The host computer preprocesses and extracts features from the series of physiological data transmitted by the customized wristband, and constructs a personnel status assessment algorithm model based on physiological data to accurately assess the current stress, activity level, and emotional fluctuation level of on-duty personnel.

[0025] Specifically, in a preferred embodiment, such as Figure 1 As shown, the status assessment method for operators in the main control room of this nuclear power plant includes: Step S10: Obtain physiological data of the evaluation subject.

[0026] Step S20: Preprocess the physiological data to obtain preprocessed data.

[0027] Step S30: Perform type identification based on the preprocessed data, and extract differentiated features based on different types of preprocessed data to obtain multimodal fusion feature data.

[0028] Step S40: Construct an on-the-job personnel status assessment algorithm model based on multimodal fusion feature data.

[0029] Step S50: Evaluate the status of the current on-duty personnel based on the on-duty personnel status evaluation algorithm model, and output the personnel status evaluation results.

[0030] In some embodiments, the physiological data acquired in step S10 is experimental data. Specifically, before model development, a simulated testing environment is set up and nuclear power plant shift workers are recruited to conduct experiments, simulating various task scenarios in their work, and physiological data under their task status is collected through customized wristbands. The collected physiological data includes, but is not limited to: pulse signal data, skin conductance signal data, acceleration signal data, angular velocity signal data, heart rate data, skin temperature data, ambient temperature and humidity data, light intensity data, altitude and air pressure data, etc.

[0031] In some embodiments, step S20 involves preprocessing the physiological data to obtain preprocessed data, including bandpass filtering and median filtering. Preprocessing the physiological data enhances its stability and reliability, thereby improving the accuracy of subsequent data identification.

[0032] Preferably, different types of preprocessed data include, but are not limited to: pulse data, skin conductance data, acceleration data, angular velocity data, and heart rate data.

[0033] In some embodiments, step S30, which involves type identification based on preprocessed data and differential feature extraction based on different types of preprocessed data to obtain multimodal fusion feature data, includes: If the data is pulse, feature extraction is performed to obtain pulse feature data; if the data is ductal skin, feature extraction is performed to obtain ductal skin feature data; if the data is acceleration, feature extraction is performed to obtain acceleration feature data; if the data is angular velocity, feature extraction is performed to obtain angular velocity feature data; if the data is heart rate, feature extraction is performed to obtain heart rate feature data; and the pulse feature data, ductal skin feature data, acceleration feature data, angular velocity feature data, and heart rate feature data are fused to obtain multimodal fused feature data.

[0034] In practical applications, feature extraction of pulse data to obtain pulse feature data includes: reconstructing pulse data using the maximum non-overlapping wavelet decomposition method to obtain post-processed data; and extracting features from the post-processed data to obtain pulse feature data. The pulse feature data includes: mean RR interval, standard deviation of interval, average difference between adjacent RR intervals, percentage of consecutive RR intervals exceeding 20 ms, percentage of consecutive RR intervals exceeding 50 ms, absolute power of the very low frequency band, low frequency power, and high frequency power. The RR interval is the time interval between the R waves of two adjacent QRS complexes on an electrocardiogram, representing the length of a complete cycle. The mean RR interval is the average of all RR intervals over a period of time, typically measured in milliseconds.

[0035] Specifically, for pulse data, since it is easily affected by motion noise, the maximum non-overlapping wavelet decomposition method can be used to reconstruct the signal and remove motion artifacts. Then, based on the post-processed data, features in the time and frequency domains such as heart rate variability can be extracted, such as mean RR interval (IBI), RR interval standard deviation (SDNN), average difference between adjacent RR intervals (RMSSD), percentage of consecutive RR intervals exceeding 20ms (PNN20), percentage of consecutive RR intervals exceeding 50ms (PNN50), absolute power in the very low frequency band (VLF), low frequency power (LF), and high frequency power (HF).

[0036] For skin conductance data, it is necessary to extract features such as mean skin conductance level (SCL-Ave), standard deviation of skin conductance level (SCL-Std), maximum skin conductance level (SCL-Max), minimum skin conductance level (SCL-Min), skin conductance response area (SCR-Area), mean skin conductance response (SCR-Ave), standard deviation of skin conductance response (SCR-Std), maximum skin conductance response (SCR-Max), and minimum skin conductance response (SCR-Min).

[0037] For acceleration data, features such as average acceleration (Acc-Ave), maximum acceleration (Acc-Max), minimum acceleration (Acc-Min), and standard deviation of acceleration (Acc-Std) are extracted.

[0038] For angular velocity data, features such as average angular velocity (Ang-Ave), maximum angular velocity (Ang-Max), minimum angular velocity (Ang-Min), and standard deviation of angular velocity (Ang-Std) are extracted.

[0039] For heart rate data, it is necessary to extract features such as the average heart rate (Ang-Ave) and the standard deviation of heart rate (Ang-Std) over a period of time.

[0040] The details of each feature are shown in Table 1 below.

[0041] Table 1. Extraction characteristics of physiological data signals collected by the customized wristband Specifically, in step S30, data that can be used to assess the current state of on-duty personnel, such as stress, activity level, and mood fluctuation, are first identified from physiological data to obtain different types of preprocessed data. Then, based on the differential features of different types of preprocessed data, multimodal fusion feature data is obtained.

[0042] Specifically as follows: Step 1: Preprocess physiological data such as pulse signal data, skin conductance signal data, acceleration signal data, angular velocity signal data, heart rate data, skin temperature data, ambient temperature and humidity data, light intensity data, altitude and air pressure data, and identify different types of preprocessed data such as pulse data, skin conductance data, acceleration data, angular velocity data and heart rate data. Step 2: First, extract the first single-modal features (such as pulse feature data, skin conductance feature data, acceleration feature data, etc.) that are strongly correlated with the state of the operators in the main control room of the nuclear power plant. Step 3: Extract the second single-modal features (such as angular velocity feature data, heart rate data, etc.) that are strongly correlated with the state of the nuclear power plant main control room operator in the target operating condition; the target operating condition includes two conditions: normal operation condition and accident operation condition. Step 4: Perform multimodal physiological data fusion on the first and second unimodal features to obtain multimodal fused feature data.

[0043] In some embodiments, step S40, constructing an on-duty personnel status assessment algorithm model based on multimodal fusion feature data, includes: constructing the structure of the on-duty personnel status assessment algorithm model; constructing the feature inputs of the on-duty personnel status assessment algorithm model based on multimodal fusion feature data; training the on-duty personnel status assessment algorithm model; and validating the trained model until it meets the accuracy requirements, thereby obtaining the on-duty personnel status assessment algorithm model. Specifically, the feature inputs for constructing the on-duty personnel status assessment algorithm model based on multimodal fusion feature data include: layered input according to different types of multimodal fusion feature data; and adding positional information to the features of different layers.

[0044] Optionally, in this embodiment of the invention, the on-duty personnel status assessment algorithm model is a deep learning model based on an attention mechanism. The deep learning model based on an attention mechanism includes an encoder and a decoder; the encoder is used to transform the input multimodal fusion features into a set of representation vectors; the decoder is used to transform the representation vectors into the current on-duty personnel's status level and output the personnel status assessment result. Specifically, as shown... Figure 2 As shown, the on-duty personnel status assessment algorithm model constructed in this invention consists of an encoder and a decoder. The encoder transforms the fusion features of peripheral physiological data, facial expressions, eye movement indicators, and speech information of the output model into a set of representation vectors through a self-attention mechanism, a feedforward neural network, and a layer normalization mechanism. Then, the decoder uses this representation vector to transform the current on-duty personnel's status level for output.

[0045] Furthermore, in this embodiment of the invention, different levels of the multimodal fusion features represent indicators from different sources, such as pulse features, skin conductance features (i.e., skin electrical conductivity features), acceleration features, and angular velocity features. Since each layer has different feature modalities, dimensions, and weights, positional information is added to the features of different layers when the fusion features are input into the model. The calculation method is as follows: ; In the formula, The input to the model after adding location information represents the features of the current layer, which in turn represents the location encoding of the current layer. The location encoding dimension is consistent with the feature dimension. For input; The position encoding function is expressed as follows: ; in, Let d represent the angular frequency of the k-th sine / cosine function, and d be the feature dimension of the current layer.

[0046] In some embodiments, step S50, which assesses the status of currently on-duty personnel based on the on-duty personnel status assessment algorithm model and outputs the personnel status assessment results, includes: acquiring real-time physiological data of currently on-duty personnel; preprocessing the real-time physiological data to obtain real-time preprocessed data; extracting differentiated features based on the real-time preprocessed data to obtain real-time fused features; performing feature encoding and feature recognition on the real-time fused features through the on-duty personnel status assessment algorithm model to obtain a representation vector; and mapping the representation vector to output the personnel status assessment results of currently on-duty personnel. The personnel status assessment results include three categories of assessment indicators: stress level, activity level, and mood fluctuation level, with each category including high, medium, and low levels.

[0047] Among them, the real-time fusion features obtained by extracting differentiated features based on real-time preprocessed data include: extracting the first single-modal features that are strongly correlated with the state of the nuclear power plant main control room operator in the real-time preprocessed data; extracting the second single-modal features that are strongly correlated with the state of the nuclear power plant main control room operator in the target operating condition; and fusing multimodal physiological data on the first single-modal features and the second single-modal features to obtain the real-time fusion features.

[0048] It should be noted that the data preprocessing and feature recognition in this step can be performed in the same way as in the aforementioned steps S20 and S30.

[0049] The core feature recognition calculation method of the on-the-job personnel status assessment algorithm model is the self-attention mechanism and the multi-head attention mechanism. During the calculation process, the input features are processed through matrix operations to obtain q (Query), k (Key), and v (Value). Then, based on q, k, and v, the self-attention mechanism is used for calculation, as shown below. The core process is as follows: ; in, This represents the result of the self-attention mechanism operation, where k represents the vector dimension, and Softmax() represents the normalization function. For any q, its dot product with k is calculated, and then the Softmax() activation function is used to obtain the current attention weights, which are then multiplied by v to obtain the output of the current position. This represents the dimension of the q / k vector.

[0050] Meanwhile, the introduction of a parallel multi-head attention mechanism into the model introduces richer features, the computation process of which is as follows: Figure 3 As shown.

[0051] Furthermore, to avoid the gradient vanishing problem in the on-the-job personnel status assessment algorithm model, residual connections and layer normalization are added to the model. The core processes involved are as follows: ; ; in, This represents the embedding vector of the input sequence. This represents the output after passing through the attention layer; The output of multi-head attention is obtained by concatenating and linearly transforming multiple parallel scaled dot product attentions. Representation layer normalization normalizes the feature dimensions of each sample, stabilizing the training process and accelerating convergence.

[0052] This invention directly uses the original input Adding it to the attention output can prevent gradient vanishing and ensure that the model can be trained to a deeper level; at the same time, it can also allow the model to retain the original input information while learning new features, thus improving training stability.

[0053] refer to Figure 4 The present invention also provides a status assessment system for operators in the main control room of a nuclear power plant. For example... Figure 4 As shown, the status assessment system for operators in the main control room of the nuclear power plant includes: a host computer and multiple customized wristbands (as shown in the figure, customized wristband #1, customized wristband #2, ..., customized wristband #N).

[0054] In this embodiment of the invention, the host computer includes: The experimental data acquisition unit is used to acquire physiological data of the evaluation subjects; The data preprocessing unit is used to preprocess physiological data to obtain preprocessed data; The feature extraction unit is used to identify the type of preprocessed data and extract differentiated features based on different types of preprocessed data to obtain multimodal fusion feature data. The evaluation model building unit is used to construct an on-the-job personnel status evaluation algorithm model based on multimodal fusion feature data; The personnel status assessment unit is used to assess the status of currently on-duty personnel based on the on-duty personnel status assessment algorithm model and output the personnel status assessment results.

[0055] In this embodiment of the invention, each customized wristband is used to collect real-time physiological data of on-duty personnel.

[0056] Preferably, the customized wristband includes a sensor module, a data processing module, and a communication module. The sensor module collects real-time physiological data from on-duty personnel; the data processing module performs preliminary processing and formatting of the real-time physiological data; and the communication module transmits the generated real-time data to a host computer. Preliminary processing and formatting of the real-time physiological data through the data processing module reduces noise and errors.

[0057] Optionally, the sensor module includes 10 types of sensors, such as a pulse sensor, a skin temperature sensor, an accelerometer, a gyroscope, a geomagnetic sensor, an EDA skin conductance sensor, an EMG electromyography sensor, an ambient temperature and humidity sensor, a light intensity sensor, and an altitude and barometric pressure sensor.

[0058] Specifically, in order to address the issue that existing wristbands can only collect physiological data and transmit it to mobile phones, lacking a clear data transmission communication protocol with the host computer, this invention achieves effective data transmission between the customized wristband and the host computer by customizing a data transmission mechanism between them.

[0059] In terms of data acquisition, this customized wristband supports the collection of various physiological data such as EDA, heart rate, and blood oxygen, with sampling frequencies meeting the monitoring needs of different physiological indicators. For data transmission, it uses either TCP or IP protocols; for example, it transmits collected data to a host computer in real time within a local area network via TCP. Regarding data caching, if the network connection is interrupted, the system can cache data and resume transmission after the connection is restored. That is, when the network connection is interrupted, the customized wristband enters reconnection mode and caches data; when the network connection is detected to be restored, the customized wristband automatically resumes uploading data to the host computer.

[0060] Customized wristband interface design: Sensor interface: Supports connection of EDA, heart rate and blood oxygen sensors.

[0061] Communication interface: TCP connection is implemented through Wi-Fi module to enable communication with host computer in local area network.

[0062] Customized data packet structure for wristband transmission: Each data packet includes: timestamp, sensor ID, data value, and checksum.

[0063] For example, data packet format: { "timestamp": "2024-11-04T14:32:00Z", "EDA": 0.56, "heart_rate": 75, "blood_oxygen": 98, "checksum": "ABC123"} Furthermore, in this embodiment of the invention, the customized wristband uploads real-time physiological data to the host computer according to the following steps: The customized wristband packages real-time physiological data and uploads it to the host computer; After uploading real-time physiological data to the host computer, it enters a waiting state. If no acknowledgment (ACK) is received from the host computer within the set waiting time, the real-time physiological data will be resent.

[0064] This invention utilizes a customized wristband to collect various data, including pulse signals, skin conductance signals, acceleration signals, angular velocity signals, heart rate, skin temperature, ambient temperature and humidity, light intensity, altitude, and air pressure. A host computer then constructs an algorithm model for assessing the on-duty status of nuclear power plant control room operators. The model's feature inputs are designed, and self-attention and multi-head attention mechanisms are used for feature recognition. This allows for accurate assessment of stress levels, activity levels, and mood fluctuations of nuclear power plant control room operators, enabling timely detection of fatigue and improving unit safety.

[0065] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0066] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0067] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0068] The above embodiments are only for illustrating the technical concept and features of the present invention, and are intended to enable those skilled in the art to understand the content of the present invention and implement it accordingly. They do not limit the scope of protection of the present invention. All equivalent changes and modifications made within the scope of the claims of the present invention should fall within the scope of the claims of the present invention.

Claims

1. A method for assessing the condition of operators in the main control room of a nuclear power plant, characterized in that, include: Step S10: Obtain physiological data of the evaluation subject; Step S20: Preprocess the physiological data to obtain preprocessed data; Step S30: Perform type identification based on the preprocessed data, and extract differentiated features based on different types of preprocessed data to obtain multimodal fusion feature data; Step S40: Construct an on-duty personnel status assessment algorithm model based on the multimodal fusion feature data; Step S50: Evaluate the status of the current on-duty personnel according to the on-duty personnel status evaluation algorithm model, and output the personnel status evaluation results.

2. The method for assessing the status of operators in the main control room of a nuclear power plant according to claim 1, characterized in that, In step S20, the physiological data is preprocessed to obtain preprocessed data, including: The physiological data are subjected to bandpass filtering and median filtering to obtain the preprocessed data.

3. The method for assessing the status of operators in the main control room of a nuclear power plant according to claim 1, characterized in that, The different types of preprocessed data include: pulse data, skin conductance data, acceleration data, angular velocity data, and heart rate data; In step S30, type identification is performed based on the preprocessed data, and differential feature extraction is performed based on different types of preprocessed data to obtain multimodal fusion feature data, including: If the data is pulse data, then feature extraction is performed on the pulse data to obtain pulse feature data; If the data is skin conductance data, then feature extraction is performed on the skin conductance data to obtain skin conductance feature data; If the data is acceleration data, then feature extraction is performed on the acceleration data to obtain acceleration feature data; If the data is angular velocity, then feature extraction is performed on the angular velocity data to obtain angular velocity feature data; If the data is heart rate data, then feature extraction is performed on the heart rate data to obtain heart rate feature data; The pulse feature data, skin conductance feature data, acceleration feature data, angular velocity feature data, and heart rate feature data are fused to obtain the multimodal fused feature data.

4. The method for assessing the status of operators in the main control room of a nuclear power plant according to claim 3, characterized in that, The step of extracting features from the pulse data to obtain pulse feature data includes: The pulse data is reconstructed using the maximum non-overlapping wavelet decomposition method to obtain post-processed data; Feature extraction is performed on the post-processed data to obtain the pulse feature data; The pulse characteristic data includes: mean RR interval, standard deviation of interval, average difference between adjacent RR intervals, percentage of consecutive RR intervals exceeding 20ms, percentage of consecutive RR intervals exceeding 50ms, absolute power in the very low frequency band, low frequency power, and high frequency power.

5. The method for assessing the status of operators in the main control room of a nuclear power plant according to claim 1, characterized in that, In step S40, constructing the on-duty personnel status assessment algorithm model based on the multimodal fusion feature data includes: Construct the structure of the on-the-job personnel status assessment algorithm model; Based on the multimodal fusion feature data, the feature inputs of the on-duty personnel status assessment algorithm model are constructed; The on-duty personnel status assessment algorithm model is trained; The trained model is validated until it meets the accuracy requirements, thus obtaining the on-duty personnel status assessment algorithm model.

6. The method for assessing the status of operators in the main control room of a nuclear power plant according to claim 5, characterized in that, The feature inputs for constructing the on-the-job personnel status assessment algorithm model based on the multimodal fusion feature data include: The multimodal fusion feature data is input in layers according to different types; Add location information to the features of different layers.

7. The method for assessing the condition of a nuclear power plant main control room operator according to any one of claims 1-6, characterized in that, The on-duty personnel status assessment algorithm model is a deep learning model based on the attention mechanism; The attention-based deep learning model includes an encoder and a decoder; The encoder is used to transform the input multimodal fusion features into a set of representation vectors; The decoder is used to transform the representation vector into the status level of the currently on-duty personnel and output the personnel status assessment result.

8. The method for assessing the status of operators in the main control room of a nuclear power plant according to claim 1, characterized in that, In step S50, the status of the current on-duty personnel is assessed according to the on-duty personnel status assessment algorithm model, and the personnel status assessment results are output, including: Obtain the real-time physiological data of the currently on-duty personnel; The real-time physiological data is preprocessed to obtain real-time preprocessed data; Based on the real-time preprocessed data, differential features are extracted to obtain real-time fused features; The on-duty personnel status assessment algorithm model is used to perform feature encoding and feature recognition on the real-time fused features to obtain a representation vector; The representation vector is mapped to output the personnel status assessment result of the currently on-duty personnel.

9. The method for assessing the status of operators in the main control room of a nuclear power plant according to claim 8, characterized in that, The step of extracting differential features based on the real-time preprocessed data to obtain real-time fused features includes: Extract the first single-modal feature that is strongly correlated with the state of the operator in the main control room of the nuclear power plant from the real-time preprocessed data; Extract the second single-modal features that are strongly correlated with the state of the operator in the main control room of the nuclear power plant under the target operating conditions; The first unimodal feature and the second unimodal feature are fused with multimodal physiological data to obtain the real-time fused feature.

10. A status assessment system for operators in the main control room of a nuclear power plant, characterized in that, The host computer includes: The experimental data acquisition unit is used to acquire physiological data of the evaluation subjects; A data preprocessing unit is used to preprocess the physiological data to obtain preprocessed data; The feature extraction unit is used to perform type identification based on the preprocessed data and to perform differentiated feature extraction based on different types of preprocessed data to obtain multimodal fusion feature data. The evaluation model construction unit is used to construct an on-the-job personnel status evaluation algorithm model based on the multimodal fusion feature data; The personnel status assessment unit is used to assess the status of the current on-duty personnel according to the on-duty personnel status assessment algorithm model and output the personnel status assessment results.

11. The status assessment system for nuclear power plant main control room operators according to claim 10, characterized in that, Also includes: A customized wristband that communicates with the host computer; The customized wristband is used to collect real-time physiological data of on-duty personnel.

12. The status assessment system for nuclear power plant main control room operators according to claim 11, characterized in that, The customized bracelet includes: a sensor module, a data processing module, and a communication module; The sensor module is used to collect the real-time physiological data of the on-duty personnel. The data processing module is used to perform preliminary processing and formatting of the real-time physiological data; The communication module is used to transmit the real-time generated data to the host computer.

13. The status assessment system for nuclear power plant main control room operators according to claim 12, characterized in that, The sensor module includes: a pulse sensor, a skin temperature sensor, an accelerometer, a gyroscope, a geomagnetic sensor, an EDA skin conductance sensor, an EMG electromyography sensor, an ambient temperature and humidity sensor, a light intensity sensor, and an altitude and barometric pressure sensor.

14. The status assessment system for nuclear power plant main control room operators according to any one of claims 11-13, characterized in that, The data transmission between the customized wristband and the host computer uses either the TCP protocol or the IP protocol. The data packets uploaded by the customized wristband to the host computer include: timestamp, sensor ID, data value, and verification code.

15. The status assessment system for nuclear power plant main control room operators according to claim 14, characterized in that, The customized wristband uploads the real-time physiological data to the host computer according to the following steps: The customized wristband packages the real-time physiological data and uploads it to the host computer. After uploading the real-time physiological data to the host computer, it enters a waiting state. If no confirmation message is received from the host computer within the set waiting time, the real-time physiological data will be resent.

16. The status assessment system for nuclear power plant main control room operators according to claim 14, characterized in that, The customized bracelet is also used to perform the following operations: When the network connection is interrupted, the customized bracelet enters reconnection mode and caches data; Once the network connection is restored, the customized wristband automatically continues to upload data to the host computer.

Images