Intelligent protection and monitoring system for railway construction

Abstract

The invention discloses an intelligent protection and monitoring system for railway construction, and the system comprises a data collection module which is used for obtaining the physiological baseline parameters, behavior modes and historical operation data of constructors, constructing a digital twinborn body model through combining with the environment and mechanical operation data collected by a multi-mode sensor disposed at a construction site, and transmitting the digital twinborn body model to a monitoring module; outputting a personalized safety threshold value; the risk prediction module is used for constructing an individualized risk prediction model based on the individualized safety threshold value and the multi-modal sensor data, and the individualized risk prediction model is used for carrying out joint modeling on behavior characteristics of construction personnel and environment changes so as to form behavior-environment coupling state vectors capable of being used for risk trajectory prediction; and the track embedding module is used for inputting the behavior-environment coupling state vector into the individualized risk prediction model. According to the method, prediction is more suitable for individual differences, the false alarm rate is remarkably reduced, a closed loop is formed by self-adaptive updating and risk evolution records, and continuous optimization of the system is ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of railway construction protection and monitoring technology, and in particular to a railway construction intelligent protection and monitoring system. BACKGROUND

[0002] In recent years, with the rapid development of railway construction, railway construction safety monitoring systems have become an important technical means to ensure construction quality and personnel safety. Existing railway construction safety monitoring technology mainly relies on multi-modal sensors, video recognition and Internet of Things, etc. to realize real-time monitoring and management of the construction site. For example, by deploying high-definition cameras and environmental sensors to collect on-site images, temperature, humidity and mechanical vibration data, combined with image processing algorithms for foreign object intrusion detection and construction progress tracking; at the same time, using wireless communication and big data analysis platform, the collected data is fused and processed to form a safety protection system, such as using space simulation and machine learning technology to generate a site model to assist in risk warning and equipment control during operating line construction. These technologies have been applied in railway engineering supervision, line inspection and disaster rescue, etc., improving construction efficiency and basic safety protection.

[0003] However, the existing railway construction safety monitoring technology has significant limitations in practical application, mainly relying on static thresholds and single modal data, which is difficult to effectively capture the interactive influence of individual physiological behavior differences of construction personnel and dynamic environmental changes, resulting in a lack of personalized adaptability of risk prediction models, and unable to realize precise embedding and adaptive updating of future risk trajectories, thus limiting the timely generation of active protection instructions and closed-loop responsibility tracing, which is prone to cause lag response of safety hazards and difficulty in tracing accident responsibility. SUMMARY

[0004] This section aims to summarize some aspects of the embodiments of the present application and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the abstract and title of the specification to avoid obscuring the purpose of this section, abstract and title, and such simplifications or omissions cannot be used to limit the scope of the present application.

[0005] In view of the above problems existing in the prior art, the present application is proposed.

[0006] Therefore, the purpose of the present application is to provide a railway construction intelligent protection and monitoring system, which is suitable for solving the problem that the prior art lacks personalized adaptability and cannot realize precise embedding and adaptive updating of future risk trajectories, which is prone to cause lag response of safety hazards.

[0007] To solve the above technical problems, the present application provides the following technical solutions: In a first aspect, the embodiments of the present application provide a railway construction intelligent protection and monitoring system, comprising: A data acquisition module is configured to acquire physiological baseline parameters, behavior patterns and historical operation data of construction personnel, and combine environmental and mechanical operation data collected by multi-modal sensors arranged at the construction site to construct a digital twin model and output individualized safety thresholds; A risk prediction module is configured to construct an individualized risk prediction model based on the individualized safety thresholds and the multi-modal sensor data, wherein the individualized risk prediction model is configured to jointly model behavior characteristics of the construction personnel and environmental changes to form a behavior-environment coupled state vector that can be used for risk trajectory prediction; A trajectory embedding module is configured to input the behavior-environment coupled state vector into the individualized risk prediction model to obtain a risk trajectory embedding in a future time window; A protection control module is configured to determine a risk level of the construction personnel according to the risk trajectory embedding and the individualized safety thresholds output by the digital twin model, and generate corresponding active protection instructions to prompt the construction personnel or control the construction machinery; A model updating module is configured to adaptively update the digital twin model based on feedback of execution of the active protection instructions, and record a risk evolution process to realize closed-loop responsibility tracing.

[0008] As a preferred scheme of the railway construction intelligent protection and monitoring system, the physiological baseline parameters are heart rate, blood pressure and body temperature data of the construction personnel acquired by wearable devices or sensors; The behavior patterns are action trajectories, posture changes and operation frequency data of the construction personnel captured by multi-modal sensors; The historical operation data are operation records, risk event records and operation duration data of the construction personnel in past construction tasks; The environmental and mechanical operation data are environmental temperature, humidity, wind speed, and operation state, vibration amplitude and power output data of the construction machinery collected by multi-modal sensors arranged at the construction site.

[0009] As a preferred scheme of the railway construction intelligent protection and monitoring system, the construction of the digital twin model and the output of the individualized safety thresholds comprise: The composite effect of the physiological baseline parameters and the behavior patterns of the construction personnel is processed by introducing a logarithmic function to obtain individualized physiological-behavior characteristic values, thereby enhancing the response capability of the digital twin model to fatigue or abnormal behavior scenarios, and the specific formula is as follows: ; wherein, is a personalized physiological-behavioral characteristic value, representing the nonlinear interaction of the physiological baseline parameters and the behavior patterns of the construction personnel, HR is the heart rate, representing the number of heartbeats per minute of the construction personnel, BP is the blood pressure, representing the mean arterial pressure of the construction personnel, is a motion trajectory feature, dimensionless, 、 and are weight coefficients for adjusting the contribution of heart rate, blood pressure and motion trajectory to the characteristic value; the compound influence of environmental temperature, humidity and wind speed is quantitatively processed by introducing a fractional structure to obtain an environmental adaptation value, thereby reducing the interference of dynamic environmental changes on the construction site on the accuracy of the digital twin model, and the specific formula is as follows: ; wherein, is an environmental adaptation value, representing the adaptive adjustment of environmental parameters on the digital twin model, T is the average environmental temperature, representing the average temperature of the construction site, H is the average environmental humidity, representing the average relative humidity of the construction site, is the average wind speed, representing the average wind speed of the construction site, is an environmental adjustment coefficient for controlling the influence of environmental disturbance on the model; the interactive influence of historical operation data and mechanical operation data is processed by introducing an exponential function to obtain an operation-mechanical state value, thereby improving the dynamic response capability of the digital twin model in complex construction scenarios, and the specific formula is as follows: ; wherein, is an operation-mechanical state value, representing the exponential comprehensive influence of historical operation data and mechanical operation state, is an operation record characteristic value, representing the standardized features based on risk events and operation duration in historical operation, is the mechanical vibration amplitude, representing the vibration intensity of the construction machinery, is an operation adjustment coefficient for controlling the weight of operation records on the state value, is a mechanical adjustment coefficient for controlling the weight of mechanical vibration on the state value; According to the obtained personalized physiological-behavioral characteristic value, environmental adaptation value and operation-mechanical state value, a digital twin model is constructed to comprehensively evaluate the influence of physiological behavior characteristics, environmental parameters and historical operation and mechanical operation data on the personalized safety threshold of the construction personnel.

[0010] As a preferred scheme of the railway construction intelligent protection and monitoring system, the specific formula of the digital twin model is as follows: ; wherein, The personalized safety threshold represents a threshold value calculated based on physiological behavior characteristics of the construction personnel, environmental parameters, and historical operation and mechanical operation data.

[0011] As a preferred scheme of the railway construction intelligent protection and monitoring system, the method for constructing an individualized risk prediction model based on the personalized safety threshold and the multi-modal sensor data comprises the following steps: By introducing a nonlinear combination method to process the dynamic interaction of the behavior characteristics of the construction personnel, a behavior risk correction value is obtained, thereby enhancing the response capability of the individualized risk prediction model to abnormal behavior of the construction personnel, and the specific formula is as follows: ; wherein, is the behavior risk correction value, representing the dynamic interaction of the behavior characteristics of the construction personnel, is a behavior characteristic value, representing the standardized characteristics of the action trajectory and the operation frequency of the construction personnel, is a behavior influence coefficient, used to adjust the weight of the behavior characteristics on the risk correction; by introducing a fractional structure to quantitatively process the combined influence of the environmental changes and the mechanical operation state in the multi-modal sensor data, an environment-mechanical inhibition value is obtained, thereby reducing the interference of the dynamic environment and the mechanical disturbance of the construction site on the risk prediction accuracy, and the specific formula is as follows: ; wherein, is the environment-mechanical inhibition value, representing the inhibitory effect of the environmental changes and the mechanical operation state on the risk prediction, is an environmental data characteristic value, representing the standardized characteristics of the environmental temperature, humidity, and wind speed, is a mechanical operation data characteristic value, representing the standardized characteristics of the mechanical vibration amplitude and power output, is the rate of change of the environmental temperature, and are adjustment coefficients, used to control the weight of the environmental and mechanical disturbance on the inhibition value; Based on the personalized safety threshold and in combination with the behavior risk correction value and the environment-mechanical inhibition value obtained by optimization, an individualized risk prediction model is constructed to comprehensively evaluate the dynamic influence of the behavior characteristics of the construction personnel, the environmental changes, and the mechanical operation state on the risk trajectory prediction, and the specific formula of the individualized risk prediction model is as follows: ; wherein, is the risk prediction value, representing the risk evaluation result optimized based on the behavior characteristics, the environmental changes, and the mechanical operation state, and is used to generate a behavior-environment coupling state vector to predict the risk trajectory in a future time window.

[0012] As a preferred scheme of the railway construction intelligent protection and monitoring system, the behavior-environment coupling state vector is input into the individualized risk prediction model to obtain a risk trajectory embedding in a future time window, including: According to the behavior-environment coupling state vector and the risk prediction value output by the individualized risk prediction model, the state vector is subjected to time sequence embedding processing to obtain a preliminary risk trajectory representation; The values are taken from the preliminary risk trajectory representation in descending order, and if the risk trajectory representation with the largest value is greater than a first preset risk threshold, the risk trajectory representation is determined as the risk trajectory embedding in the future time window; The risk level of the construction personnel is determined according to the risk trajectory embedding and the individualized safety threshold output by the digital twin model, including: The risk trajectory embedding is compared with the individualized safety threshold to obtain a risk level.

[0013] As a preferred scheme of the railway construction intelligent protection and monitoring system, the risk level of the construction personnel is determined according to the risk trajectory embedding and the individualized safety threshold output by the digital twin model, including: If the risk trajectory embedding with the largest value is less than or equal to a first preset risk threshold, and the risk prediction value of the behavior-environment coupling state vector corresponding to the risk trajectory embedding is greater than a second preset risk threshold, risk warning information corresponding to the risk trajectory embedding is generated through a risk assessment process; Wherein, the first preset risk threshold is greater than the second preset risk threshold; According to the environmental change of the risk trajectory embedding and the risk warning information, the risk level of the construction personnel is determined, and the corresponding active protection instruction is generated.

[0014] As a preferred scheme of the railway construction intelligent protection and monitoring system, the risk level of the construction personnel is determined according to the risk trajectory embedding and the individualized safety threshold output by the digital twin model, including: If the risk prediction value of the behavior-environment coupling state vector of the risk trajectory embedding is less than or equal to the second preset risk threshold, the associated environmental data associated with the risk trajectory embedding is determined; Based on the associated environmental data, the risk level of the construction personnel is determined, and the corresponding active protection instruction is generated.

[0015] In a second aspect, an embodiment of the present application provides a computer device comprising a memory and a processor, the memory storing a computer program, wherein the computer program, when executed by the processor, implements any step of the railway construction intelligent protection and monitoring system according to the first aspect of the present application.

[0016] In a third aspect, an embodiment of the present application provides a computer readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements any step of the railway construction intelligent protection and monitoring system according to the first aspect of the present application.

[0017] The present application has the following beneficial effects: through multi-source data fusion of the data acquisition module and construction of the digital twin model, the system outputs personalized safety thresholds, avoiding the deviation of the universal thresholds in the prior art; in combination with the individualized model of the risk prediction module and the behavior-environment coupling state vector, a future risk trajectory embedding is formed, which makes the prediction more suitable for individual differences and significantly reduces the false alarm rate; the hierarchical threshold judgment compares the risk trajectory embedding with the threshold to generate targeted instructions, compared with the lag response of the prior art, the system can intervene within a forward-looking time window to reduce the probability of accidents; at the same time, the adaptive update of the model update module and the risk evolution record form a closed loop to ensure continuous optimization of the system, through recording the risk evolution process, the system supports closed-loop responsibility tracing, facilitating post-analysis and legal accountability, and changes from passive monitoring to active intelligent protection, and is applicable to railway infrastructure enterprises, construction units and safety management agencies. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor. Among them: Fig. 1 A schematic diagram of the overall process of the railway construction intelligent protection and monitoring system according to the present application; Fig. 2 A schematic diagram of risk level judgment of the railway construction intelligent protection and monitoring system according to the present application. DETAILED DESCRIPTION

[0019] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings of the specification.

[0020] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, it will be apparent to one skilled in the art that the present application can be practiced without the specific details set forth in this description. In other instances, well-known methods, procedures, components, and circuits have not been described in detail so as not to unnecessarily obscure aspects of the present application.

[0021] Secondly, the "one embodiment" or "an embodiment" referred to herein means a specific feature, structure, or characteristic under discussion. Each of the various aspects presented in this description can be implemented in many different embodiments and each of the described aspects can be implemented alone or in combination with other aspects. It will thus be appreciated that the embodiments described herein are not limited to only those embodiments that can possibly be explicitly present in the above description.

[0022] Thirdly, the present application is described in detail in conjunction with the schematic diagram. In the detailed description of the embodiments of the present application, the cross-sectional view of the device structure is locally enlarged without the general proportion for the convenience of description, and the schematic diagram is only an example which should not limit the scope of protection of the present application herein. In addition, the three-dimensional spatial dimensions of length, width and depth should be included in actual production.

[0023] Embodiment one Reference Figs. 1-2 For one embodiment of the present application, a railway construction intelligent protection and monitoring system is provided.

[0024] The existing railway construction safety monitoring technology has significant limitations in actual application, mainly manifested as relying on static threshold and single modal data, which is difficult to effectively capture the interactive influence of individual physiological behavior differences of construction personnel and dynamic environmental changes, resulting in lack of personalized adaptability of risk prediction model, inability to realize precise embedding and adaptive updating of future risk trajectory, and further limiting the timely generation of active protection instructions and closed-loop responsibility tracing, which is easy to cause lag response of safety hazards and difficulty in tracing accident responsibility.

[0025] The present application provides a railway construction intelligent protection and monitoring system which can effectively solve the above-mentioned problems. Next, how to realize the system will be described in detail in conjunction with multiple embodiments.

[0026] Fig. 1 A whole flowchart of a railway construction intelligent protection and monitoring system is shown, wherein the system includes the following functional modules: a data acquisition module, a risk prediction module, a trajectory embedding module, a protection control module, and a model updating module. The functions of the above modules are introduced one by one as follows: Data acquisition module: used for acquiring physiological baseline parameters, behavior patterns and historical operation data of construction personnel, and combining with environmental and mechanical operation data collected by multi-modal sensors arranged at the construction site; Further, the physiological baseline parameters refer to the heart rate, blood pressure, and body temperature data of the construction workers obtained through wearable devices or sensors; the behavior patterns refer to the action trajectory, posture change, and work frequency data of the construction workers captured through multi-modal sensors; the historical work data refer to the operation records, risk event records, and work duration data of the construction workers in past construction tasks; and the environmental and mechanical operation data refer to the environmental temperature, humidity, and wind speed as well as the running state, vibration amplitude, and power output data of the construction machinery collected through multi-modal sensors deployed on the construction site.

[0027] It should be noted that the railway construction site involves a complex dynamic environment and various behaviors of construction workers, and in order to ensure the accuracy of safety monitoring, the present application obtains various types of data through the data acquisition module, including the physiological baseline parameters (such as heart rate, blood pressure, and body temperature) of the construction workers, the behavior patterns (such as action trajectory, posture change, and work frequency) of the construction workers, the historical work data (such as operation records, risk event records, and work duration) of the construction workers, and the environmental and mechanical operation data (such as environmental temperature, humidity, wind speed, mechanical vibration amplitude, and power output) of the construction site. These data are obtained through wearable devices, multi-modal sensors, and historical databases, and are transmitted to the central processing unit for subsequent modeling and analysis. The diversity of data acquisition is due to the complexity of construction safety risks, for example, physiological parameters reflect the fatigue state of construction workers, behavior patterns reveal potential operational error risks, environmental data capture the influence of weather or site disturbances, and mechanical operation data reflect the influence of equipment state on safety.

[0028] In a specific implementation, the data acquisition module adopts a multi-modal sensor network, including but not limited to the following devices: (1) wearable sensors (such as smart bracelets or chest straps) for real-time monitoring of the heart rate, blood pressure, and body temperature of construction workers; (2) visual sensors (such as high-definition cameras or laser radars) for capturing the action trajectory, posture change, and work frequency of construction workers through image processing or point cloud analysis; (3) environmental sensors (such as thermohygrographs and anemometers) for collecting the temperature, humidity, and wind speed of the construction site; and (4) mechanical sensors (such as vibration sensors and power meters) for monitoring the vibration amplitude and power output of construction machinery. The historical work data are extracted from the construction management system database, including operation records (such as construction task logs), risk event records, and work duration.

[0029] To ensure the accuracy and real-time of data collection, the present application adopts data preprocessing and feature extraction technology, classifies and standardizes the collected data by analyzing the metadata (such as timestamp, device ID) or content features (such as physiological data fluctuation amplitude, behavior trajectory spatial distribution) of sensor data. In specific implementation, the data can be processed through preset feature analysis algorithm (such as Fourier transform based on time series signal or machine learning classifier) to output standardized feature vector. For example, heart rate data extracts abnormal fluctuation features through sliding window analysis, action trajectory identifies abnormal behavior patterns through spatial clustering algorithm, environmental data eliminates noise interference through statistical filtering, and mechanical data extracts vibration features through frequency domain analysis. The output of the data collection module is a multi-dimensional feature vector, including physiological feature values, behavior feature values, environmental feature values and mechanical operation feature values, which provides accurate basis for the construction of subsequent digital twin model.

[0030] The advantage of this operation is that through multi-modal data fusion, the dynamic risk factors of the construction site can be fully captured, ensuring that the digital twin model can generate personalized safety thresholds for individual construction personnel and specific environments. For example, construction personnel data identified as high heart rate and abnormal action trajectory can be used to model fatigue risk first, and environmental data under adverse weather conditions can be used for mechanical stability assessment. In contrast, alternative technologies include single-modal data collection (such as relying only on video monitoring) or threshold detection based on fixed rules, but these methods are difficult to adapt to individual differences in construction personnel and dynamic changes in the environment, resulting in low risk prediction accuracy or response lag. The present application selects a multi-modal feature analysis method, which can automatically process diversified data and adapt to real-time and personalized requirements in complex railway construction scenarios, significantly improving the efficiency and accuracy of model construction.

[0031] It should be further explained that in the above data preprocessing process, the multi-modal feature analysis method adopted by the present application specifically includes the following steps: first, through a modal fusion algorithm (such as a neural network based on attention mechanism), cross-modal alignment is performed on different modal data (such as time sequence of physiological signals, spatial data of behavior trajectory, and numerical vector of environmental parameters), to ensure synchronization of data in time and space dimensions; second, a feature extraction model (such as convolutional neural network CNN for image / trajectory data, and recurrent neural network RNN for time sequence physiological data) is used to extract high-dimensional feature vectors from each modality, for example, RNN is applied to physiological parameters to capture heart rate fluctuation patterns, and principal component analysis (PCA) is applied to environmental data to reduce dimension and extract key disturbance factors; then, a multi-modal attention fusion layer (such as multi-head attention of Transformer) is used to calculate the interaction weight between modalities, for example, higher weight is given to the interaction between behavior pattern and environmental data to capture the combined effect of "wind speed affecting posture stability"; finally, the fused multi-dimensional feature vector is output for input of the downstream digital twin model.

[0032] Preferably, the digital twin model is constructed, and the personalized safety threshold is output, including: By introducing a logarithmic function to process the combined effect of the physiological baseline parameters and behavior patterns of the construction personnel, a personalized physiological-behavioral characteristic value is obtained, thereby enhancing the response capability of the digital twin model to fatigue or abnormal behavior scenarios, and the specific formula is as follows: ; wherein, is the personalized physiological-behavioral characteristic value, representing the nonlinear interaction between the physiological baseline parameters and behavior patterns of the construction personnel, HR is the heart rate, representing the number of heartbeats per minute of the construction personnel, BP is the blood pressure, representing the mean arterial pressure of the construction personnel, is the action trajectory feature (unit is dimensionless, and is processed by, for example, division by a reference speed value of 1 m / s), 、 and are weight coefficients for adjusting the contribution of heart rate, blood pressure, and action trajectory to the characteristic value; By introducing a fractional structure to quantitatively process the combined effect of environmental temperature, humidity, and wind speed, an environmental adaptation value is obtained, thereby reducing the interference of dynamic environmental changes on the accuracy of the digital twin model on the construction site, and the specific formula is as follows: ; wherein, The environmental adaptation value represents the adaptive adjustment effect of environmental parameters on the digital twin model. T is the mean ambient temperature (normalized, dimensionless), representing the average temperature at the construction site. H is the mean ambient humidity (normalized, dimensionless), representing the average relative humidity at the construction site. The wind speed is the mean (normalized, dimensionless), representing the average wind speed at the construction site. This is the environmental adjustment coefficient, used to control the impact of environmental disturbances on the model; By introducing an exponential function to process the interaction between historical operation data and machinery operation data, an operation-machinery state value is obtained, thereby improving the dynamic response capability of the digital twin model in complex construction scenarios. The specific formula is as follows: ; in, This is the work-machinery status value, representing the exponential combined effect of historical work data and machine operating status. The operation record feature value represents the standardized features based on risk events and operation duration in historical operations (the unit is dimensionless and normalized, such as by dividing by the reference risk event frequency value). Mechanical vibration amplitude (unit: (), indicating the vibration intensity of the construction machinery. This is the job adjustment coefficient, used to control the weighting of status values ​​in the operation log. Mechanical adjustment coefficient (unit: ), used to control the weighting of mechanical vibration on state values; Based on the obtained personalized physiological-behavioral characteristic values, environmental adaptation values, and work-machinery status values, a digital twin model is constructed to comprehensively evaluate the impact of construction workers' physiological and behavioral characteristics, environmental parameters, and historical work and machinery operation data on personalized safety thresholds. The specific formula for the digital twin model is as follows: ; in, Personalized safety thresholds represent thresholds calculated based on the physiological and behavioral characteristics of construction workers, environmental parameters, and historical work and machinery operation data. This represents the interactive multiplicative effect of personalized physiological-behavioral traits and environmental adaptation, emphasizing that environmental factors amplify physiological and behavioral risks. Nonlinear multiplication captures the coupling effect, and then... Representing independent historical operations and machine condition contributions, this formula reflects the cumulative effect of these factors as baseline risk, balancing interaction and accumulation in the overall formula.

[0033] In this embodiment, the weight coefficients of the contributions of heart rate, blood pressure and motion trajectory to the personalized physiological-behavioral characteristic values 、 and , and the environmental adjustment coefficient , the work adjustment coefficient , and the mechanical adjustment coefficient are derived as follows: the weight coefficients 、 and can be automatically adjusted by using gradient descent method to minimize the prediction error by using historical physiological and behavioral data, ensuring that the weighted contributions of heart rate, blood pressure and motion trajectory sum to 1, thereby balancing the influence of each physiological and behavioral characteristic; the environmental adjustment coefficient can be obtained by analyzing the correlation between environmental data (such as temperature, humidity, wind speed) and risk events, combined with machine learning methods (such as random forest); the work adjustment coefficient and the mechanical adjustment coefficient can be determined by time series analysis of historical work data and mechanical operation data using least squares method, or can be fine-tuned combined with expert experience method, which is not specifically limited in this embodiment.

[0034] It should be noted that the traditional railway construction safety monitoring system usually uses fixed threshold method, which is difficult to effectively handle the combined effects of individual physiological behavior differences of construction personnel, dynamic changes of environment and mechanical operation state, resulting in lack of personalization in threshold setting, which is prone to risk prediction deviation. This embodiment builds a digital twin model, introduces a logarithmic function to handle the combined effects of physiological baseline parameters and behavior patterns, quantifies the dynamic effects of environmental temperature, humidity and wind speed with a fractional structure, and processes the interaction of historical work and mechanical operation data with an exponential function, generates personalized physiological-behavioral characteristic values, environmental adaptation values and work-mechanical state values, and realizes multi-dimensional nonlinear modeling. This model can accurately adapt to individual differences of construction personnel and complex construction scenes, significantly improve the accuracy of personalized safety threshold, overcome the limitations of traditional fixed threshold method, reduce the false positive or false negative risk caused by threshold deviation, and provide reliable technical support for subsequent risk prediction and protection control.

[0035] Risk prediction module: for constructing an individualized risk prediction model based on the personalized safety threshold and multi-modal sensor data, the individualized risk prediction model is used to jointly model the behavior characteristics of the construction personnel and the environmental changes to form a behavior-environment coupling state vector that can be used for risk trajectory prediction.

[0036] For constructing an individualized risk prediction model based on the personalized safety threshold and multi-modal sensor data, including the following steps: The dynamic interaction of the behavior characteristics of the construction personnel is processed by introducing a nonlinear combination method to obtain a behavior risk correction value, thereby enhancing the response capability of the individualized risk prediction model to the abnormal behavior of the construction personnel, and the specific formula is as follows: ; Among them, is the behavior risk correction value, indicating the dynamic interaction of the behavior characteristics of the construction personnel, is the behavior characteristic value, dimensionless, indicating the standardized characteristics of the action trajectory and the operation frequency of the construction personnel, is the behavior influence coefficient, used to adjust the weight of the behavior characteristics on the risk correction; The compound influence of the environmental change and the mechanical operation state in the multi-modal sensor data is quantitatively processed by introducing a fractional structure to obtain an environment-mechanical inhibition value, thereby reducing the interference of the dynamic environment and the mechanical disturbance of the construction site on the risk prediction accuracy, and the specific formula is as follows: ; Among them, is the environment-mechanical inhibition value, indicating the inhibitory effect of the environmental change and the mechanical operation state on the risk prediction, is the environmental data characteristic value, dimensionless, indicating the standardized characteristics of the environmental temperature, humidity and wind speed, is the mechanical operation data characteristic value, dimensionless, indicating the standardized characteristics of the mechanical vibration amplitude and power output, is the environmental temperature change rate, with the unit of , and are adjustment coefficients, used to control the weight of the environmental and mechanical disturbance on the inhibition value, with the unit of , with the unit of ; Based on the individualized safety threshold and in combination with the behavior risk correction value and the environment-mechanical inhibition value obtained by optimization, an individualized risk prediction model is constructed to comprehensively evaluate the dynamic influence of the behavior characteristics of the construction personnel, the environmental change and the mechanical operation state on the risk trajectory prediction, and the specific formula of the individualized risk prediction model is as follows: ; Among them, is the risk prediction value, indicating the risk evaluation result optimized based on the behavior characteristics, the environmental change and the mechanical operation state, used to generate a behavior-environment coupling state vector to predict the risk trajectory in the future time window.

[0037] In this embodiment, the behavior influence coefficient and the adjustment coefficient for controlling the weight of the environmental and mechanical disturbance on the environment-mechanical inhibition value and are derived as follows: behavior influence coefficient By analyzing the correlation between the behavior characteristics (such as motion trajectory, operation frequency) of the construction personnel and the historical risk events, a support vector machine or similar machine learning method is used for optimization to ensure the accurate contribution of the behavior characteristics to the risk correction; adjustment coefficient and By fitting the historical interaction between the environmental data (such as temperature, humidity, wind speed) and the mechanical operation data (such as vibration amplitude, power output), combined with the least squares method or Bayesian optimization algorithm to determine, to minimize the risk prediction error, it can also be fine-tuned through construction site measured data and expert experience method, this embodiment does not make specific limitation.

[0038] It should be noted that the existing railway construction safety monitoring system usually relies on single modal data or static model, and it is difficult to effectively capture the combined influence of construction personnel behavior and dynamic environment-mechanical interaction, resulting in lack of individualization in risk prediction and inability to prospectively identify potential accident trends. The embodiment of the present application builds an individualized risk prediction model, introduces a nonlinear combination method to process the dynamic interaction of the behavior characteristics of the construction personnel, quantifies the combined influence of environmental changes and mechanical operation states in a fractional structure, generates behavior risk correction values and environment-mechanical inhibition values, and realizes multi-dimensional dynamic modeling. This model can accurately identify the superimposed effect of abnormal behavior of construction personnel (such as fatigue operation) and environmental-mechanical disturbance (such as sudden weather or equipment failure) on risk, significantly improve the accuracy of risk trajectory prediction, overcome the limitations of traditional static models, fill the gap of individualized dynamic risk prediction, and provide prospective support for subsequent trajectory embedding and protection control, reducing the risk of safety accidents caused by prediction lag.

[0039] Trajectory embedding module: used for inputting the behavior-environment coupling state vector into the individualized risk prediction model to obtain the risk trajectory embedding in the future time window.

[0040] It should be noted that the safety risk of railway construction site has high dynamicity and complexity, involving the interaction of construction personnel behavior, environmental changes and mechanical operation state. In order to realize prospective risk prediction, the present application inputs the behavior-environment coupling state vector into the individualized risk prediction model through the trajectory embedding module, generates the risk trajectory embedding in the future time window to represent the potential risk trend of the construction personnel in a specific time period. The behavior-environment coupling state vector is generated by the risk prediction module and includes the physiological behavior characteristics (such as heart rate anomaly, motion trajectory deviation) of the construction personnel, environmental characteristics (such as temperature, wind speed change) and mechanical operation characteristics (such as vibration amplitude), providing multi-dimensional input for risk trajectory prediction.

[0041] Further, as Fig. 2As shown, in a specific implementation, according to the risk prediction value output by the behavior-environment coupling state vector and the individualized risk prediction model, the state vector is subjected to time sequence embedding processing to obtain a preliminary risk trajectory representation; the time sequence embedding processing adopts a time sequence analysis method (such as a long short-term memory network LSTM or a sliding window convolution) to analyze the time sequence characteristics (such as the time evolution of the behavior characteristics and the fluctuation period of the environmental data) of the state vector and the risk prediction value output by the individualized risk prediction model, and construct a multi-dimensional time sequence vector to represent the dynamic evolution trend of the risk. The preliminary risk trajectory representation is a set of multi-dimensional vectors, including the risk scores at each time step; According to the values in the preliminary risk trajectory representation in descending order, if the risk trajectory representation with the largest value is greater than a first preset risk threshold, the risk trajectory representation is determined as a risk trajectory embedding in a future time window, which is used as an input for subsequent risk level assessment; According to the risk trajectory embedding and the individualized safety threshold output by the digital twin model, the risk level of the construction personnel is determined, including: The risk trajectory embedding is compared with the individualized safety threshold to obtain the risk level.

[0042] The operation has the advantage that through time sequence embedding and threshold screening, high-risk trends in the future time window can be accurately captured, and reliable input is provided for the protection control module. For example, the high-risk embedding of the abnormal action trajectory of the construction personnel in the strong wind environment can trigger the protection instruction preferentially. In contrast, alternative technologies include risk assessment based on a static threshold or single-time-point prediction, but these methods are difficult to capture the dynamic evolution of behavior-environment interaction, resulting in prediction lag or misjudgment. The present application selects the time sequence embedding and dynamic threshold screening method because it can automatically process multi-dimensional time sequence data, adapt to the real-time and predictive requirements in complex railway construction scenarios, and significantly improve the accuracy and forward-looking nature of the risk trajectory embedding.

[0043] The protection control module is used to determine the risk level of the construction personnel according to the risk trajectory embedding and the individualized safety threshold output by the digital twin model, and generate corresponding active protection instructions to prompt the construction personnel or control the construction machinery.

[0044] Further, the safety management of the railway construction site needs to take timely and targeted protective measures based on the dynamic risk assessment results to reduce the probability of accidents and ensure the safety of construction personnel and equipment. The present application realizes accurate determination of risk level and generation of active protection instructions through the protection control module, compares the risk trajectory embedding with the individualized safety threshold based on risk trajectory embedding and individualized safety threshold, and considers the interactive influence of construction personnel behavior, environmental changes and mechanical operation state to generate protective instructions suitable for different risk scenarios. The risk trajectory embedding is generated by the trajectory embedding module, representing the potential risk trend in the future time window; the individualized safety threshold is output by the digital twin model, reflecting the dynamic safety boundary of the individual physiological behavior characteristics of the construction personnel and the environmental conditions on site.

[0045] In specific implementation, first, compare the risk trajectory embedding with the individualized safety threshold output by the digital twin model to determine the risk level of the construction personnel; the comparison process uses a threshold judgment algorithm (such as a logical judgment or classifier based on comparison operations), determines the risk level (high risk, medium risk and low risk) by calculating the relative relationship between the maximum value of the risk trajectory embedding and the individualized safety threshold: Further, to realize fine-grained risk assessment, a multi-level threshold judgment mechanism is introduced. If the numerically largest risk trajectory embedding is less than or equal to the first preset risk threshold, and the risk prediction value of the behavior-environment coupling state vector corresponding to the risk trajectory embedding is greater than the second preset risk threshold, then generate medium risk warning information corresponding to the risk trajectory embedding through the risk assessment process; Wherein, the first preset risk threshold is greater than the second preset risk threshold; According to the environmental changes of the risk trajectory embedding and the risk warning information, determine the risk level of the construction personnel and generate the corresponding active protection instructions.

[0046] In specific implementation, the risk assessment process can use a preset analysis algorithm (such as a decision tree based on rules or a machine learning model) to generate warning information (such as "be careful of fatigue" or "be careful of strong wind") according to the features (such as abnormal motion frequency, environmental wind speed sudden change) in the behavior-environment coupling state vector. Subsequently, combined with the environmental change data (such as temperature, humidity fluctuation) in the risk trajectory embedding and the warning information, determine the medium risk level and generate the corresponding active protection instructions, for example, send voice prompts (such as "please rest for 5 minutes") to the construction personnel through wearable devices or control the construction machinery to slow down through the wireless communication module.

[0047] Further, according to the risk trajectory embedding and the individualized safety threshold output by the digital twin model, determine that the construction personnel is in a low risk level, If the risk prediction value of the behavior-environment coupled state vector embedded in the risk trajectory is less than or equal to the second preset risk threshold, the associated environmental data (such as the historical trend of environmental temperature, humidity or mechanical vibration amplitude) associated with the risk trajectory embedding is determined; based on the associated environmental data, the risk level of the construction personnel is determined, and the corresponding active protection instruction (for example, sending a low-level prompt (such as "keep normal operation") to the construction personnel or adjusting the mechanical operation parameter (such as maintaining the current power output)) is generated. The processing of the low-risk scene ensures that the system does not intervene excessively in the safe state, and optimizes the use of resources.

[0048] The advantage of this operation is that by multi-level threshold judgment and environment association analysis, high, medium and low risk scenarios can be accurately distinguished, and hierarchical response and dynamic protection are realized. In contrast, alternative technologies include alarm systems based on a single threshold or fixed rule control strategies, but these methods are difficult to adapt to dynamic changes and individual differences in the construction site, resulting in delayed early warning or frequent false alarms. The present application selects a multi-level threshold comparison and environment association evaluation method, which can automatically handle complex risk scenarios and consider both immediate risks and environmental backgrounds, significantly improving the relevance and timeliness of protection instructions, effectively reducing the accident rate, and optimizing construction efficiency and equipment management.

[0049] For example, assume that a construction worker is working at night at a high-speed rail construction site, and the system detects that his heart rate is continuously rising, the motion trajectory is abnormally deviated (for example, frequently deviating from the safe operation area), and the environmental data indicates that the wind speed has suddenly increased. Based on the comparison of the risk trajectory embedding and the personalized safety threshold, the protection control module determines that the risk prediction value is in the medium risk interval (greater than the second preset risk threshold but less than the first preset risk threshold). The system generates a warning message (such as "pay attention to fatigue and strong wind") through the risk assessment process and sends a vibration prompt to the construction worker, suggesting that he pause work and rest for 5 minutes, while reducing the operating power of the nearby excavator by 10% through the wireless communication module to avoid potential collision risks. After receiving the medium-level early warning, the maintenance personnel analyze the environmental change data and confirm that the strong wind is the main risk factor, and adjust the construction plan in time to avoid safety accidents caused by fatigue and environmental disturbance, ensuring the safety of the construction personnel and the stable operation of the equipment.

[0050] In the embodiments of the present application, for each threshold value in the protection control module, the first preset risk threshold value is calculated by statistical analysis of historical construction data (including personnel behavior, environment and mechanical operation state), distribution characteristics and confidence interval of risk trajectory embedding are calculated, combined with railway construction safety specifications and equipment operation parameter setting; the second preset risk threshold value is determined based on the risk prediction value distribution of the behavior-environment coupling state vector, combined with the characteristic analysis of typical risk scenarios (such as fatigue, severe weather) in the construction site, aiming to distinguish between medium and low risk scenarios; the evaluation threshold of environment-related data is set by analyzing the correlation between environmental parameters (such as temperature, humidity, wind speed) and historical risk events, using regression model and quantile method, to ensure accurate identification of low-risk scenarios. The initial values of the above threshold values can be preliminarily determined through small-scale test data, and then optimized through continuous monitoring and feedback, or a machine learning method (such as neural network training based on historical data) is introduced to obtain a more optimal threshold combination, which is not specifically limited in the present embodiment.

[0051] The model updating module is used to adaptively update the digital twin model based on the execution feedback of the active protection instruction, and record the risk evolution process to realize closed-loop responsibility tracing.

[0052] Preferably, the dynamic and complex nature of the railway construction site requires the safety monitoring system to continuously optimize the digital twin model according to the actual protection effect, in order to improve the long-term prediction accuracy and support clear tracing of accident responsibility. The present application realizes the adaptive updating of the digital twin model through the model updating module, adjusts the parameters of the digital twin model using the execution feedback of the active protection instruction (such as whether the instruction is effective, the response of the construction personnel, the adjustment result of the machine), and records the risk evolution process to form structured log data for closed-loop responsibility tracing. The execution feedback includes the execution state of the protection instruction (such as whether the prompt is accepted, whether the machine adjusts according to the instruction), environmental response data (such as the stability after the wind speed changes) and personnel behavior changes (such as the heart rate returning to normal), providing multi-dimensional basis for model optimization and tracing.

[0053] In a specific implementation, the model updating module first collects the execution feedback data of the active protection instruction, obtains real-time feedback through a sensor network and a communication interface, such as response confirmation signals returned by the construction personnel's wearable device, state updates of the machine control system (such as execution logs of power reduction) and subsequent data of the environmental sensor (such as measured values after the wind speed stabilizes). These feedback data are standardized by preprocessing algorithms (such as time series filtering or anomaly detection), and feedback feature vectors are extracted, including instruction execution success rate, response time and environmental change trend.

[0054] Subsequently, the module adjusts the parameters of the digital twin model using an adaptive update algorithm (such as online gradient descent or incremental learning); to optimize the digital twin model, the update process calculates the actual risk state based on the execution feedback data, compares it with the risk prediction value output by the individualized risk prediction model, generates a prediction error, and uses it to adjust the parameters of the digital twin model. The actual risk state is represented by the risk level in the feedback data, reflecting the actual effect of the protective instructions. For example, if the feedback shows that the heart rate of a construction worker does not return to normal after a high-risk instruction, the module increases the weight of the physiological parameter in the logarithmic function to enhance the sensitivity of the individualized physiological-behavioral feature value to the fatigue scenario; if the environmental feedback shows that the mechanical stability improves after strong winds, the module optimizes the environmental parameter weight in the fractional structure, adjusts the environmental adaptation value, and reduces false positives. The update process is achieved by minimizing the prediction error, as follows: ; wherein, is the loss function value at time t, representing the deviation between the individualized risk prediction model and the actual effect, is the actual risk state calculated based on the execution feedback data, standardized by the risk level and feedback features (such as instruction execution success rate), and dimensionless, is the logarithmic function weight of the individualized physiological-behavioral feature value, used to adjust the influence of physiological and behavioral parameters, is the fractional structure weight of the environmental adaptation value, used to adjust the interactive influence of historical work and mechanical data, is the exponential function weight of the work-mechanical state value, used to adjust the interactive influence of historical work and mechanical data, is the regularization coefficient, used to control the stability of weight update, and T is the time window length.

[0055] Specifically, the adaptive update process is as follows by minimizing the prediction error formula: ①: At each time step t, calculate the difference between the output of the individualized risk prediction model and the actual risk state , which constitutes the first term of the loss function , for example, if the feedback shows that the heart rate of a construction worker returns to normal after a high-risk instruction ( ), the single prediction value is high risk ( ), then the error is large, indicating that the model needs to be adjusted; ②: Based on the loss function, use the online gradient descent algorithm to adjust the parameters of the digital twin model (i.e., the weights corresponding to , and , , and ) is optimized, with the following steps: First, the partial derivatives of the loss function with respect to each weight are calculated: ; Similarly, the partial derivatives of and can be calculated, where = , = , = 1 (according to ); Second, the weights are adjusted according to the partial derivatives, for example: ; where is the updated log function weight of the personalized physiological-behavioral feature value, is the log function weight of the personalized physiological-behavioral feature value before updating, is the learning rate, and similarly, the weights of and can be updated; ③: Adjust (the weight of the physiological-behavioral parameter in the log function), (the weight of the environmental parameter in the fractional structure), and (the weight of the work-machine data in the exponential function) using the updated weights , , and , thereby optimizing the personalized safety threshold of the digital twin model ; ④: Iterative optimization: repeat the above steps within each time window (e.g., 10 minutes) to continuously minimize the loss function, ensuring that adapts to dynamic construction scenarios.

[0056] In the embodiments of the present application, the log function weight of the personalized physiological-behavioral feature value , the fractional structure weight of the environmental adaptation value , the exponential function weight of the work-machine state value , and the regularization coefficient The optimal values can be found by fitting historical construction data to minimize the loss function, which is most suitable for the current application dataset. Specifically, the initial values of these parameters are based on statistical analysis of historical multi-modal data of railway construction sites (such as physiological parameter fluctuation records, environmental change logs, and mechanical operation events), and the deviation between the output of the digital twin model formula and the actual safety threshold (annotated from the risk event database) is fitted by least squares method or gradient descent algorithm; for example, the logarithmic function weight (adjusting the interaction intensity of physiological baseline parameters and behavior patterns) is calculated by analyzing the correlation between heart rate and motion trajectory in fatigue scenarios, the fractional structure weight (adjusting the inhibitory effect of environmental temperature, humidity and wind speed) is fitted based on the distribution of environmental disturbances in adverse weather, the exponential function weight (adjusting the nonlinear interaction between historical operation data and mechanical vibration) is optimized by Monte Carlo method simulation of complex construction scenarios, and the regularization coefficient (controlling the overall update stability) selects the value that minimizes the risk of overfitting by cross-validation; then, these parameters are iterated through online learning to ensure the robustness of the model in dynamic scenarios, which is not limited in the embodiment.

[0057] The existing railway construction safety monitoring system generally has the defects of static threshold dependence and lack of closed-loop optimization, which cannot effectively identify and handle the development process of individual differences of construction personnel gradually accumulating into serious risks, resulting in failure to early warn potential accidents; the embodiment introduces an adaptive updating mechanism and designs an optimization strategy based on execution feedback and loss function, which realizes the scientific adjustment and iteration of the parameters of the digital twin model; the log record of the risk evolution process ensures the complete tracking of the historical trajectory of the system, effectively reducing the interference of environmental disturbances or behavior abnormalities on the model; this adaptive updating and closed-loop tracing method fills the gap in feedback optimization in traditional monitoring technology, enabling the system to mine potential risk patterns and development trends from the historical trajectory of construction operation, so as to optimize the protection response before the accident fully appears; this method effectively solves the technical problems of difficult to capture dynamic risks and insufficient individual adaptability in railway construction, provides a scientific basis for predictive safety management, greatly improves the predictability and reliability of the system, reduces the construction interruption and accident loss caused by model lag, and ensures the continuity of project progress and personnel safety.

[0058] In summary, through the multi-source data fusion of the data acquisition module and the construction of the digital twin model, the system outputs personalized safety thresholds, avoiding the deviation of the universal thresholds in the prior art; combined with the individualized model of the risk prediction module and the behavior-environment coupling state vector, the future risk trajectory embedding is formed, which makes the prediction more suitable for individual differences and significantly reduces the false positive rate; the hierarchical threshold judgment compares the risk trajectory embedding with the threshold to generate targeted instructions, compared with the lag response of the prior art, the system can intervene within the forward-looking time window to reduce the probability of accidents; at the same time, the adaptive update of the model update module and the risk evolution record form a closed loop to ensure continuous optimization of the system, and through the record of the risk evolution process, the system supports closed-loop responsibility tracing, facilitating post-analysis and legal accountability, and shifting from passive monitoring to active intelligent protection.

[0059] Embodiment two, as an embodiment of the present application, is different from the previous embodiment in that: If the functions are realized in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application or the parts that essentially contribute to the prior art or parts of the technical solutions can be embodied in the form of a software product, which is stored in a storage medium and includes instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk and various program code storage media.

[0060] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered a list of executable instructions for implementing logic functions, and can be specifically embodied in any computer-readable medium for use by an instruction execution system, device or apparatus, such as a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, device or apparatus, or in conjunction with these instructions execution system, device or apparatus. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transport programs for use by an instruction execution system, device or apparatus, or in conjunction with these instruction execution system, device or apparatus.

[0061] More specific examples (a non-exhaustive list) of the computer-readable medium include the following: an electrical connection (electronic) having one or more wires, a portable computer diskette (magnetic), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can also be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, for example, via an optical scanner, then compiled, interpreted, or otherwise processed, as necessary, and stored in a computer memory.

[0062] It should be understood that aspects of the application can be implemented in hardware, software, firmware or combinations thereof. In the embodiments described above, various steps or methods can be implemented, for example, by software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, and in another embodiment, any of the following technologies, known in the art, or combinations thereof, can be used: discrete logic circuitry having logic gates for implementing logic functions upon an application of data signals, application specific integrated circuits having appropriate combinational logic gates, programmable gate arrays (PGA), field programmable gate arrays (FPGA), and the like.

[0063] It should be noted that the above-mentioned embodiments are merely used to illustrate the technical solutions of the present application but not to limit the present application, and although the present application is described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalent replaced without departing from the spirit and scope of the present application, and all of them should be covered in the scope of the claims of the present application.

Claims

1. A railway construction intelligent protection and monitoring system, characterized in that, The method comprises the following steps: a data acquisition module: used for acquiring physiological baseline parameters, behavior patterns and historical operation data of construction personnel, and combining environmental and mechanical operation data collected by multi-modal sensors arranged on the construction site to construct a digital twin model and output individualized safety thresholds; a risk prediction module: used for constructing an individualized risk prediction model based on the individualized safety thresholds and the multi-modal sensor data, the individualized risk prediction model being used for jointly modeling behavior characteristics of construction personnel and environmental changes to form a behavior-environment coupling state vector that can be used for risk trajectory prediction; a trajectory embedding module: used for inputting the behavior-environment coupling state vector into the individualized risk prediction model to obtain a risk trajectory embedding in a future time window; a protection control module: used for determining a risk level of construction personnel according to the risk trajectory embedding and the individualized safety thresholds output by the digital twin model, and generating corresponding active protection instructions to prompt the construction personnel or control the construction machinery; a model updating module: used for adaptively updating the digital twin model based on feedback of execution of the active protection instructions, and recording a risk evolution process to realize closed-loop responsibility tracing.

2. The intelligent protection and monitoring system for railway construction according to claim 1, characterized in that: The physiological baseline parameters refer to heart rate, blood pressure and body temperature data of construction personnel acquired by wearable devices or sensors; The behavior patterns refer to action trajectories, posture changes and operation frequency data of construction personnel captured by multi-modal sensors; The historical operation data refer to operation records, risk event records and operation duration data of construction personnel in past construction tasks; The environmental and mechanical operation data refer to environmental temperature, humidity, wind speed, and operation state, vibration amplitude and power output data of construction machinery collected by multi-modal sensors arranged on the construction site.

3. The intelligent protection and monitoring system for railway construction according to claim 2, characterized in that: The construction of the digital twin model and the output of the individualized safety thresholds comprise the following steps: The compound effect of physiological baseline parameters and behavior patterns of construction personnel is processed by introducing a logarithmic function to obtain individualized physiological-behavior characteristic values, thereby enhancing the response capability of the digital twin model to fatigue or abnormal behavior scenarios; The compound influence of environmental temperature, humidity and wind speed is quantitatively processed by introducing a fractional structure to obtain an environmental adaptation value, thereby reducing the interference of dynamic environmental changes on the precision of the digital twin model on the construction site; The interaction influence of historical operation data and mechanical operation data is processed by introducing an exponential function to obtain an operation-mechanical state value, thereby improving the dynamic response capability of the digital twin model in complex construction scenarios; Based on the obtained individualized physiological-behavior characteristic values, environmental adaptation values and operation-mechanical state values, the digital twin model is constructed to comprehensively evaluate the influence of physiological behavior characteristics, environmental parameters and historical operation and mechanical operation data of construction personnel on the individualized safety thresholds.

4. The intelligent protection and monitoring system for railway construction according to claim 3, characterized in that: The specific formula of the digital twin model is as follows: ； wherein is a personalized safety threshold, is a personalized physiological-behavioral characteristic value, is an environmental adaptation value, is a work-machine status value.

5. The intelligent protection and monitoring system for railway construction according to claim 4, characterized in that: The construction of the individualized risk prediction model based on the individualized safety thresholds and the multi-modal sensor data comprises the following steps: The dynamic interaction of the behavior characteristics of the construction personnel is processed by introducing a nonlinear combination method to obtain a behavior risk correction value, thereby enhancing the response capability of the individualized risk prediction model to abnormal behavior of the construction personnel; The compound influence of environmental changes and mechanical operating states in the multi-modal sensor data is quantitatively processed by introducing a fractional structure to obtain an environment-mechanical inhibition value, thereby reducing the interference of the dynamic environment and mechanical disturbance of the construction site on the risk prediction accuracy; Based on the individualized safety threshold and in combination with the behavior risk correction value and the environment-mechanical inhibition value obtained by optimization, an individualized risk prediction model is constructed to comprehensively evaluate the dynamic influence of the behavior characteristics of the construction personnel, the environmental changes and the mechanical operating states on the risk trajectory prediction, and the specific formula of the individualized risk prediction model is as follows: ； wherein, is a risk prediction value, is a behavioral risk modifier value, is an environmental-mechanical inhibition value.

6. The intelligent protection and monitoring system for railway construction according to claim 5, characterized in that: The behavior-environment coupling state vector is input into the individualized risk prediction model to obtain a risk trajectory embedding in a future time window, including: According to the behavior-environment coupling state vector and the risk prediction value output by the individualized risk prediction model, the state vector is processed by time series embedding to obtain a preliminary risk trajectory representation; The values are taken from the preliminary risk trajectory representation in descending order, and if the risk trajectory representation with the largest value is greater than a first preset risk threshold, the risk trajectory representation is determined as the risk trajectory embedding in the future time window; The risk level of the construction personnel is determined according to the risk trajectory embedding and the individualized safety threshold output by the digital twin model, including: The risk trajectory embedding is compared with the individualized safety threshold to obtain a risk level.

7. The intelligent protection and monitoring system for railway construction according to claim 6, characterized in that: The risk level of the construction personnel is determined according to the risk trajectory embedding and the individualized safety threshold output by the digital twin model, including: If the risk trajectory embedding with the largest value is less than or equal to a first preset risk threshold, and the risk prediction value of the behavior-environment coupling state vector corresponding to the risk trajectory embedding is greater than a second preset risk threshold, risk warning information corresponding to the risk trajectory embedding is generated through a risk assessment process; Wherein, the first preset risk threshold is greater than the second preset risk threshold; According to the environmental changes of the risk trajectory embedding and the risk warning information, the risk level of the construction personnel is determined, and the corresponding active protection instruction is generated.

8. The intelligent protection and monitoring system for railway construction according to claim 7, characterized in that: The risk level of the construction personnel is determined according to the risk trajectory embedding and the individualized safety threshold output by the digital twin model, including: If the risk prediction value of the behavior-environment coupling state vector corresponding to the risk trajectory embedding is less than or equal to the second preset risk threshold, the associated environmental data associated with the risk trajectory embedding is determined; Based on the associated environmental data, it is determined that the construction personnel is in a low risk level, and the corresponding active protection instruction is generated. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is characterized in that: The processor executes the computer program to realize the steps of the railway construction intelligent protection and monitoring system according to any one of claims 1-8.

10. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to realize the steps of the railway construction intelligent protection and monitoring system according to any one of claims 1-8.

Images