Highway construction safety risk intelligent identification method and system

Abstract

The invention is suitable for the technical field of safety monitoring, and provides a highway construction safety risk intelligent identification method and system, and the method comprises the following steps: constructing a physical constraint dynamic Bayesian network model used for predicting the collapse risk in front of tunneling; performing real-time risk probabilistic reasoning according to the real-time monitoring data based on a physical constraint dynamic Bayesian network model to obtain a risk posterior probability value; according to the multi-source dynamic data related to the construction environment, a safety probability early warning threshold value is adjusted in a self-adaptive mode, and a dynamic safety probability early warning threshold value is obtained; and on the basis of the dynamic safety probability early warning threshold, according to the risk posterior probability value, the risk of collapse occurring in front of tunneling is identified. According to the method, on the basis of the risk posterior probability value output by the physical constraint dynamic Bayesian network model, a self-adaptive safety threshold adjustment mechanism based on environment multi-source dynamic data is fused, and the accuracy and the intelligent level of tunnel front collapse risk identification can be remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of safety monitoring technology, and in particular relates to an intelligent identification method and system for safety risks during highway construction. Background Technology

[0002] In the construction of underground engineering projects such as highway tunnels, the risk of surrounding rock collapse is one of the main safety threats. Current technologies generally rely on various sensors (such as stress gauges, displacement gauges, and ground-penetrating radar) to monitor the state of the surrounding rock and issue warnings by comparing the monitoring data with preset fixed safety thresholds. While this method is direct, it fails to fully consider the dynamic physical mechanisms and complex environmental coupling effects during tunnel excavation, resulting in insufficient accuracy in early warnings and frequent false alarms and missed alarms.

[0003] In recent years, some improved technologies have attempted to introduce data fusion or simple machine learning models to integrate and analyze multi-source monitoring data in order to improve the accuracy of risk identification. However, the decision-making logic of these methods lacks clear physical interpretation. More importantly, their safety judgment criteria are usually static or require human experience to set, and cannot be adaptively adjusted according to dynamic factors such as environmental disturbances, thus limiting their level of intelligence.

[0004] Therefore, there is an urgent need to develop a risk prediction scheme for highway tunnel construction that can achieve more accurate and reliable forward-looking safety early warning. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent identification method for safety risks during highway construction, aiming to solve the aforementioned technical problems.

[0006] This invention is implemented as follows: a method for intelligent identification of safety risks during highway construction, comprising the following steps:

[0007] Based on the geological conditions and construction parameters of tunnel engineering, a physically constrained dynamic Bayesian network model is constructed to predict the risk of collapse ahead of tunnel excavation.

[0008] Based on a physically constrained dynamic Bayesian network model, real-time risk probability inference is performed based on real-time monitoring data to obtain the posterior risk probability value.

[0009] Based on multi-source dynamic data related to the construction environment, the safety probability early warning threshold is adaptively adjusted to obtain the dynamic safety probability early warning threshold.

[0010] Based on the dynamic safety probability early warning threshold, the risk of collapse ahead of tunnel excavation is identified according to the risk posterior probability value.

[0011] Furthermore, based on the tunnel engineering geological conditions and construction parameters, the steps for constructing a physically constrained dynamic Bayesian network model to predict the risk of collapse ahead of tunnel excavation specifically include:

[0012] Based on the geological conditions and construction parameters of the tunnel project, the key physical state variables and monitoring variables affecting the collapse were determined;

[0013] Based on the fundamental principles of geotechnical mechanics and engineering experience, a causal relationship network among various key physical state variables is established, which serves as the node and structural topology of a dynamic Bayesian network.

[0014] The physical laws in the causal relationship network are transformed into constraints on the node state transition probabilities and conditional probability distributions, thus completing the construction of a physically constrained dynamic Bayesian network model.

[0015] Furthermore, the key physical state variables include the surrounding rock stress state, the range of the plastic zone, the equivalent stiffness of the support structure, and the groundwater seepage pressure; the monitoring variables include the tunnel perimeter convergence displacement, the arch crown settlement displacement, the anchor bolt axial force, the initial support concrete stress, the advanced geological prediction results, and the tunneling parameters.

[0016] Furthermore, based on a physically constrained dynamic Bayesian network model, the step of performing real-time risk probability inference based on real-time monitoring data to obtain the posterior probability value of the risk specifically includes:

[0017] Real-time monitoring data is collected and used as evidence input for the corresponding monitoring variables in the physical constraint dynamic Bayesian network model.

[0018] Based on the evidence input, the conditional probability distribution of the state of each node in the physical constraint dynamic Bayesian network model is updated using the Bayesian inference algorithm, and the posterior probability value of the risk of collapse in front of the tunnel at the current moment is output.

[0019] Furthermore, the step of adaptively adjusting the safety probability early warning threshold based on multi-source dynamic data related to the construction environment to obtain the dynamic safety probability early warning threshold specifically includes:

[0020] Simultaneously acquire multi-source dynamic environmental data related to the construction environment at the current moment;

[0021] Establish a mapping relationship model between multi-source dynamic environmental data and early warning sensitivity adjustment coefficients;

[0022] Based on the current dynamic environmental data from multiple sources, the early warning sensitivity adjustment coefficient for the current moment is calculated using a mapping relationship model.

[0023] Based on the current warning sensitivity adjustment coefficient, the preset baseline safety probability warning threshold is dynamically adjusted to generate a dynamic safety probability warning threshold suitable for the current construction environment.

[0024] Furthermore, the steps for establishing a mapping model between multi-source dynamic environmental data and early warning sensitivity adjustment coefficients specifically include:

[0025] Comprehensive environmental factors are determined based on multi-source dynamic environmental data, including advanced geological forecasting results, blasting vibration monitoring data, tunnel temperature and humidity data, and rainfall data. The comprehensive environmental factors are determined by a weighted linear fusion of terrain deterioration trend factors, blasting vibration intensity factors, temperature influence factors, humidity influence factors, and rainfall influence factors. The terrain deterioration trend factor is determined through geological condition ratings in the advanced geological forecasting results. The blasting vibration intensity factor is determined through blasting vibration monitoring data. The temperature and humidity influence factors are determined through tunnel temperature and humidity data. The rainfall influence factor is determined through rainfall data.

[0026] Based on the improved Sigmoid function, a mapping relationship model for the adjustment coefficient of early warning sensitivity is determined according to comprehensive environmental factors.

[0027] Furthermore, the expression for the mapping relationship model is as follows:

[0028] ;

[0029] In the formula, Z is the early warning sensitivity adjustment coefficient; Z is the comprehensive environmental factor, whose value is not less than zero; k is the steepness coefficient greater than zero.

[0030] Furthermore, based on the dynamic safety probability early warning threshold, and according to the risk posterior probability value, the steps for identifying the risk of collapse ahead of tunnel excavation specifically include:

[0031] The posterior probability value of the risk is compared with the dynamic safety probability warning threshold to obtain the comparison result;

[0032] Based on the comparison results, the risk level of collapse ahead of the tunnel excavation is determined, and corresponding early warning signals are triggered.

[0033] Another objective of this invention is to provide an intelligent identification system for highway construction safety risks, used to implement the aforementioned intelligent identification method for highway construction safety risks, comprising:

[0034] The model building module is used to construct a physically constrained dynamic Bayesian network model based on the geological conditions and construction parameters of the tunnel project to predict the risk of collapse ahead of tunnel excavation.

[0035] The risk reasoning module is used to perform real-time risk probability reasoning based on physical constraint dynamic Bayesian network model and real-time monitoring data to obtain the posterior probability value of risk.

[0036] The threshold adaptive adjustment module is used to adaptively adjust the safety probability early warning threshold based on multi-source dynamic data related to the construction environment, so as to obtain the dynamic safety probability early warning threshold.

[0037] The risk identification module is used to identify the risk of collapse ahead of tunnel excavation based on the dynamic safety probability warning threshold and the risk posterior probability value.

[0038] Furthermore, the model building module specifically includes:

[0039] The variable determination unit is used to determine the key physical state variables and monitoring variables that affect the collapse based on the tunnel engineering geological conditions and construction parameters.

[0040] The causal relationship establishment unit is used to establish a causal relationship network between various key physical state variables based on the basic principles of geotechnical mechanics and engineering experience, which serves as the node and structural topology of the dynamic Bayesian network.

[0041] The model generation unit is used to transform the physical laws in the causal relationship network into constraints on the node state transition probabilities and conditional probability distributions, thereby completing the construction of a physically constrained dynamic Bayesian network model.

[0042] The risk reasoning module specifically includes:

[0043] The monitoring data acquisition unit is used to collect monitoring data in real time, obtain real-time monitoring data, and use the real-time monitoring data as evidence input for the corresponding monitoring variables in the physical constraint dynamic Bayesian network model.

[0044] The Bayesian inference unit is used to update the conditional probability distribution of the state of each node in the physical constraint dynamic Bayesian network model based on evidence input and using the Bayesian inference algorithm, and output the posterior probability value of the risk of collapse in front of the tunnel at the current moment.

[0045] The threshold adaptive adjustment module specifically includes:

[0046] The environmental data acquisition unit is used to synchronously acquire multi-source dynamic environmental data related to the construction environment at the current moment.

[0047] The mapping relationship establishment unit is used to establish a mapping relationship model between multi-source dynamic environmental data and early warning sensitivity adjustment coefficients.

[0048] The adjustment coefficient calculation unit is used to calculate the early warning sensitivity adjustment coefficient for the current moment based on the current multi-source dynamic environmental data and through a mapping relationship model.

[0049] The threshold dynamic adjustment unit is used to dynamically adjust the preset benchmark safety probability warning threshold based on the warning sensitivity adjustment coefficient at the current moment, and generate a dynamic safety probability warning threshold suitable for the current construction environment.

[0050] The risk identification module specifically includes:

[0051] The risk probability comparison unit is used to compare the posterior risk probability value with the dynamic safety probability warning threshold to obtain the comparison result.

[0052] The risk level determination unit is used to determine the risk level of a collapse ahead of the tunnel excavation based on the comparison results and to trigger the corresponding early warning signal.

[0053] This invention provides an intelligent identification method for highway construction safety risks. By constructing a physically constrained dynamic Bayesian network model driven by physical mechanisms, and embedding fundamental principles of geotechnical mechanics and engineering experience as prior constraints into the reasoning process of the model, the interpretability and extrapolation reliability of the model are fundamentally improved, making the risk identification results more physically meaningful. Furthermore, by using the posterior probability values ​​of the risks output by the physically constrained dynamic Bayesian network model as a basis, and integrating an adaptive safety threshold adjustment mechanism based on multi-source dynamic environmental data, this invention can significantly improve the accuracy and intelligence level of tunnel collapse risk identification. It can also autonomously adjust its sensitivity according to changes in the current construction environment, effectively reducing false alarms and missed alarms, and providing more forward-looking and adaptive decision support for highway tunnel construction safety. Attached Figure Description

[0054] Figure 1 This is a flowchart illustrating the intelligent identification method for highway construction safety risks provided in an embodiment of the present invention.

[0055] Figure 2 This is a flowchart illustrating step S100 in the intelligent identification method for highway construction safety risks provided in this embodiment of the invention.

[0056] Figure 3 This is a flowchart illustrating step S200 in the intelligent identification method for highway construction safety risks provided in an embodiment of the present invention.

[0057] Figure 4 This is a flowchart illustrating step S300 in the intelligent identification method for highway construction safety risks provided in an embodiment of the present invention.

[0058] Figure 5This is a flowchart illustrating step S400 in the intelligent identification method for highway construction safety risks provided in an embodiment of the present invention.

[0059] Figure 6 A schematic diagram of the intelligent identification system for highway construction safety risks provided in an embodiment of the present invention.

[0060] Figure 7 This is a schematic diagram of the structure of the model building module provided in an embodiment of the present invention.

[0061] Figure 8 This is a schematic diagram of the risk reasoning module provided in an embodiment of the present invention.

[0062] Figure 9 This is a schematic diagram of the threshold adaptive adjustment module provided in an embodiment of the present invention.

[0063] Figure 10 This is a schematic diagram of the risk identification module provided in an embodiment of the present invention. Detailed Implementation

[0064] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0065] like Figure 1 As shown, in one embodiment of the present invention, a method for intelligent identification of safety risks during highway construction is provided, comprising the following steps:

[0066] S100. Based on the geological conditions and construction parameters of tunnel engineering, a physically constrained dynamic Bayesian network model is constructed to predict the risk of collapse ahead of tunnel excavation.

[0067] S200: Based on a physically constrained dynamic Bayesian network model, real-time risk probability inference is performed based on real-time monitoring data to obtain the posterior probability value of the risk.

[0068] S300. Based on multi-source dynamic data related to the construction environment, adaptively adjust the safety probability early warning threshold to obtain the dynamic safety probability early warning threshold.

[0069] S400: Based on the dynamic safety probability early warning threshold, the risk of collapse ahead of tunnel excavation is identified according to the risk posterior probability value.

[0070] In practical applications, tunnel engineering geological conditions and construction parameters include the stress state of the surrounding rock, the range of the plastic zone, the equivalent stiffness of the support structure, the groundwater seepage pressure, the convergence displacement around the tunnel, the settlement displacement of the arch crown, the axial force of the anchor bolts, the stress of the initial support concrete, the results of advanced geological prediction, and the tunneling parameters.

[0071] like Figure 2 As shown, in a preferred embodiment of the present invention, the step of constructing a physically constrained dynamic Bayesian network model for predicting the risk of collapse ahead of tunnel excavation based on the tunnel engineering geological conditions and construction parameters, namely step S100, specifically includes:

[0072] S110. Based on the geological conditions and construction parameters of the tunnel project, determine the key physical state variables and monitoring variables that affect the collapse;

[0073] Specifically, the stress state of the surrounding rock, the range of the plastic zone, the equivalent stiffness of the support structure, and the groundwater seepage pressure are identified as key physical state variables.

[0074] The stress state of the surrounding rock is generally represented by the stress ratio, which is a continuous variable. The stress ratio is defined as the ratio of the actual maximum shear stress of the surrounding rock to the shear strength of the rock mass. A stress ratio close to or exceeding 1 indicates impending failure. The actual maximum shear stress of the surrounding rock can be calculated using three-dimensional numerical simulation (such as FLAC3D) combined with the current excavation step (the current excavation step refers to the latest complete excavation and support unit segment being analyzed or simulated in the tunnel excavation cycle). The shear strength of the rock mass can be determined based on the geological survey report (including the rock mass cohesion c and the internal friction angle). ) and the current stress state (such as the normal stress acting on the potential sliding surface) The calculation is as follows:

[0075] ;

[0076] The extent of the plastic zone is generally represented by the radius of the plastic zone; the radius of the plastic zone is a continuous variable, specifically referring to the radius of the area where the rock mass around the tunnel enters the plastic state, and can be estimated by theoretical formulas (such as the Kästner equation);

[0077] The equivalent stiffness of the support structure is a continuous variable, which comprehensively reflects the stiffness of the combined support system of anchor bolts, shotcrete, steel arches, etc. Its theoretical value can be calculated through design parameters and material constitutive models, or obtained through back analysis of field deformation data.

[0078] Groundwater seepage pressure refers to the groundwater seepage pressure in front of the tunnel face. It is a continuous variable that affects the effective stress and strength of the surrounding rock. Specifically, it can be obtained through advance borehole water pressure measurement or seepage coupling numerical analysis.

[0079] In addition, the tunnel perimeter convergence displacement, arch crown settlement displacement, anchor bolt axial force, initial support concrete stress, advanced geological prediction results, and tunneling parameters were identified as monitoring variables associated with key physical state variables; monitoring variables refer to the actual observed data, which serve as evidence for inference.

[0080] Among them, the tunnel perimeter convergence displacement is a continuous variable, which refers to the amount of reduction in the horizontal or inclined distance between two fixed measuring points on the tunnel cross section profile after the tunnel is excavated. It is the most direct and intuitive indicator reflecting the overall inward deformation (convergence) of the surrounding rock, and can be measured by a convergence meter or total station.

[0081] The settlement displacement of the tunnel crown is a continuous variable, which refers to the vertical displacement of the center point of the tunnel crown relative to a stable reference point (usually buried outside the tunnel or in a deep stable stratum). It reflects the stability of the surrounding rock above the tunnel and can be measured by a level or total station.

[0082] The axial force of the anchor bolt is a continuous variable, which refers to the axial tensile force borne by the anchor bolt or anchor cable installed in the surrounding rock. It directly reflects the magnitude of the load borne by the anchor bolt in the plastic zone of the surrounding rock and can be measured by a steel gauge installed on a representative anchor bolt.

[0083] The initial support concrete stress is a continuous variable, referring to the compressive or tensile stress (usually mainly compressive stress) borne by the shotcrete support layer. It reflects the load shared by the shotcrete layer as the main load-bearing body and can be measured by embedded concrete strain gauges.

[0084] Advanced geological prediction results are discrete variables, referring to the geological condition rating predicted using geophysical exploration, drilling, and other methods in front of the tunnel face before excavation. The aim is to predict unfavorable geological bodies ahead (such as faults, fracture zones, water-rich areas, etc.), which can be determined by methods such as tunnel seismic wave method and ground-penetrating radar.

[0085] The tunneling parameters are continuous variables, including tunneling rate, cutterhead torque, and cutterhead thrust, which can be obtained from the construction records; tunneling rate refers to the cutterhead penetration or cycle advance speed; cutterhead torque and cutterhead thrust refer to the torque and total thrust required to overcome the resistance of the surrounding rock, respectively.

[0086] S120. Based on the basic principles of rock and soil mechanics and engineering experience, establish a causal relationship network among various key physical state variables, which serves as the node and structural topology of the dynamic Bayesian network.

[0087] Specifically, based on the Mohr-Coulomb yield criterion or the Hawke-Brown criterion, a causal relationship network is established to extend the stress state of the surrounding rock to the range of the plastic zone;

[0088] Based on the theory of interaction between surrounding rock and support structure, a causal relationship network is established to show that the range of plastic zone and the equivalent stiffness of support structure jointly affect the deformation of surrounding rock (including the convergence displacement around the tunnel and the settlement displacement of the arch).

[0089] Based on the principles of seepage mechanics, a causal relationship network is established between groundwater seepage pressure and the stress state and plastic zone expansion of the surrounding rock.

[0090] In practical applications, the surrounding rock stress ratio at current time t is... The surrounding rock stress ratio at the previous time t-1 and radius of plastic zone And the advanced geological prediction results (geological condition rating) at the current time t. Groundwater seepage pressure Equivalent stiffness of support structure The effect, i.e., the ratio of surrounding rock stress at the current time t. For about , , , , The function is shown in the following formula:

[0091] ;

[0092] Radius of the plastic zone at current time t For about The function is shown in the following formula:

[0093] ;

[0094] In the formula, a is the tunnel excavation radius; p0 is the original rock stress; p r The radial resistance provided to the support is related to the equivalent stiffness of the support structure;

[0095] Hole perimeter convergence displacement at time t Or the settlement and displacement of the arch. Equal deformation is related to the radius of the plastic zone Equivalent stiffness of support structure The function can be determined through the surrounding rock characteristic curve and the support characteristic curve; specifically, the surrounding rock characteristic curve is used to describe the relationship between the required support resistance and the radial displacement of the tunnel wall. The larger the radius of the plastic zone, the higher and rightward the surrounding rock characteristic curve, which means that a larger support force is required for the same displacement, or a larger displacement will occur for the same support force.

[0096] The support characteristic curve is used to describe the relationship between the resistance provided by the support structure and its own deformation. Its slope is closely related to the equivalent stiffness of the support structure; the greater the equivalent stiffness of the support structure, the steeper the curve.

[0097] The intersection of the surrounding rock characteristic curve and the support characteristic curve represents the displacement and support resistance when the interaction between the surrounding rock and the support structure reaches equilibrium. The tunnel perimeter convergence displacement and the arch crown settlement displacement are both positively correlated with the displacement at equilibrium, and the proportionality coefficient is determined by the tunnel shape and stress field.

[0098] Based on the principles of seepage mechanics, as groundwater seepage pressure increases, effective stress decreases, leading to a reduction in the shear strength of the rock mass. This directly results in the yield criterion being more easily satisfied, i.e., an increase in the surrounding rock stress ratio. Simultaneously, the decrease in rock mass shear strength and the change in effective stress cause the yield surface to shift, making the plastic zone more prone to expansion. Furthermore, for surrounding rock containing weak argillaceous interlayers, groundwater inflow softens and muddies it, causing a sharp loss of shear strength (a significant decrease in rock mass cohesion and internal friction angle), which is reflected in this embodiment of the invention through updating the geological condition rating.

[0099] S130. Transform the physical laws in the causal relationship network into constraints on the node state transition probability and conditional probability distribution, and complete the construction of the physical constraint dynamic Bayesian network model.

[0100] Specifically, the basic principles and empirical formulas of geotechnical mechanics are embedded in the parameters of the conditional probability table or Gaussian probability distribution function of the nodes in the form of soft constraints, so that the probability relationship between nodes has a higher weight in the interval dominated by physical laws.

[0101] In practical applications, the core of the physically constrained dynamic Bayesian network model is the conditional probability distribution, whose inference objective is the risk state of a collapse ahead of tunnel excavation. The risk state is a discrete variable, which can include states such as low risk and high risk. In this embodiment of the invention, a parameterized conditional probability distribution is used, and the physical laws in the causal relationship network are embedded as prior information. For nodes with continuous variables, a linear Gaussian model or a Gaussian mixture model is used, wherein the mean in the linear Gaussian model or the Gaussian mixture model can be obtained by the above function. The simplified linear form is used to determine the value; for nodes of discrete variables, such as risk states, which are constrained by the physical states of parent nodes such as surrounding rock stress states and plastic zone ranges, the value can be determined by mapping through preset physical rules; for example, if and If the risk state is high, then the risk state is high; otherwise, the risk state is low. The probability of "high risk" corresponding to the combination of parent node states that satisfies this kind of physical rule will be given a higher initial value (prior), and then Bayesian update learning will be performed through some labeled data.

[0102] In this embodiment of the invention, by encoding the fundamental principles of geotechnical mechanics into probabilistic constraints between network nodes, the physical constraint dynamic Bayesian network model is established on explicit physical causal relationships, overcoming the shortcomings of traditional data models, such as lack of interpretability and poor extrapolation. Furthermore, this embodiment successfully integrates deterministic physical laws with uncertain engineering experience and monitoring data, forming a unified probabilistic reasoning framework that both follows objective laws and accommodates uncertainty, laying a solid and flexible foundation for subsequent intelligent analysis.

[0103] like Figure 3 As shown, in a preferred embodiment of the present invention, the step of obtaining the posterior probability value of risk by performing real-time risk probability inference based on real-time monitoring data using a physically constrained dynamic Bayesian network model, i.e., step S200, specifically includes:

[0104] S210. Collect monitoring data in real time, obtain real-time monitoring data, and use the real-time monitoring data as evidence input for the corresponding monitoring variables in the physical constraint dynamic Bayesian network model.

[0105] In practical applications, monitoring data include tunnel perimeter convergence displacement, arch crown settlement displacement, anchor bolt axial force, initial support concrete stress, geological condition rating, tunneling rate, cutterhead torque, and cutterhead thrust.

[0106] S220. Based on the evidence input, use the Bayesian inference algorithm to update the conditional probability distribution of the state of each node in the physical constraint dynamic Bayesian network model, and output the posterior probability value of the risk of collapse in front of the tunnel at the current moment.

[0107] Specifically, the objective of the physically constrained dynamic Bayesian network model is to calculate the posterior probability distribution of the target variable (risk state) given all current and past evidence. Since this physically constrained dynamic Bayesian network model is a hybrid (continuous + discrete) model and may contain cyclic dependencies (through time), the embodiments of this invention use particle filtering algorithm or joint tree algorithm as Bayesian inference algorithm.

[0108] In this embodiment of the invention, taking the particle filter algorithm as an example, the reasoning process is as follows:

[0109] 1. Initialization: At t=0, generate N particles according to the prior distribution (i.e., assign values ​​to N complete network state variables).

[0110] 2. Prediction: Based on the state of the particles at time t-1 and the transition probabilities of the physical constraint dynamic Bayesian network model (defined by the conditional probability distribution), predict the new state of each particle at time t.

[0111] 3. Update: Compare the real-time monitoring data at time t with the predicted state of each particle, and calculate the weight of each particle. The weight is proportional to the probability of observing actual data in the current particle state (calculated from the conditional probability distribution of the observation node).

[0112] 4. Resampling: Particles are resampled according to their weights, low-weight particles are eliminated, and high-weight particles are replicated. The resampled particle set approximately represents the posterior distribution.

[0113] 5. Output: The proportion of particles with a high-risk state in the particle set after statistical resampling is the posterior probability value of the risk of collapse ahead of the tunnel at the current moment.

[0114] like Figure 4 As shown, in a preferred embodiment of the present invention, the step of adaptively adjusting the safety probability early warning threshold based on multi-source dynamic data related to the construction environment to obtain the dynamic safety probability early warning threshold, i.e., step S300, specifically includes:

[0115] S310. Simultaneously acquire multi-source dynamic environmental data related to the construction environment at the current moment; the multi-source dynamic environmental data includes: advanced geological forecast results, blasting vibration monitoring data, temperature and humidity data inside the tunnel, rainfall data, etc.

[0116] S320. Establish a mapping relationship model between multi-source dynamic environmental data and early warning sensitivity adjustment coefficients;

[0117] Specifically, historical multi-source dynamic environmental data are analyzed, and the correspondence between the risk posterior probability values ​​output by the physical constraint dynamic Bayesian network model and the actual occurrence of engineering hazards under different combinations of environmental conditions is statistically analyzed to obtain statistical results.

[0118] Based on statistical results, a mapping relationship model is constructed using fuzzy logic rules or lightweight machine learning models. This model takes multi-source dynamic data as input and an early warning sensitivity adjustment coefficient as output. The early warning sensitivity adjustment coefficient is used to lower the safety probability early warning threshold in high-risk environments and to appropriately raise the safety probability early warning threshold in low-risk environments.

[0119] S330. Based on the current environmental multi-source dynamic data, calculate the warning sensitivity adjustment coefficient for the current moment through a mapping relationship model; wherein, the warning sensitivity adjustment coefficient is a positive number less than or equal to 1.

[0120] S340. Based on the warning sensitivity adjustment coefficient at the current moment, dynamically adjust the preset benchmark safety probability warning threshold to generate a dynamic safety probability warning threshold suitable for the current construction environment.

[0121] Specifically, the preset baseline security probability warning threshold is defined as follows: , The value of is generally between 0.7 and 0.9; therefore, the dynamic safety probability warning threshold at the current time t is... In the formula, This is the adjustment coefficient for the early warning sensitivity at the current time t.

[0122] In a preferred embodiment of the present invention, the step of establishing a mapping relationship model between multi-source dynamic environmental data and early warning sensitivity adjustment coefficient, namely step S320, specifically includes:

[0123] First, comprehensive environmental factors are determined based on multi-source dynamic environmental data;

[0124] Among them, the terrain deterioration trend factor can be determined through the geological condition rating in the advanced geological prediction results. Specifically, the geological condition rating is quantified into a numerical index (such as the RMR value, which refers to the CSIR rock mass geomechanical classification index value). The calculation formula for the terrain deterioration trend factor is as follows:

[0125] ;

[0126] In the formula, E1 is the terrain deterioration trend factor; RMR t This represents the RMR value of the geology at the current moment. The average RMR value of geology over a predetermined historical period; The standard deviation of geological RMR values ​​over a pre-defined historical period;

[0127] The blasting vibration intensity factor can be determined using blasting vibration monitoring data (including peak particle velocity data from cyclic blasting). Specifically, the calculation formula for the blasting vibration intensity factor is as follows:

[0128] ;

[0129] In the formula, E2 is the blast vibration intensity factor; PPV t The peak particle velocity (PPV) of the previous explosion at the current moment; s A preset safety threshold for the peak particle velocity of cyclic blasting;

[0130] Temperature and humidity influencing factors can be determined using temperature and humidity data within the tunnel (including temperature and humidity data). Specifically, the formula for calculating the temperature influencing factor is as follows:

[0131] ;

[0132] In the formula, E3 is the temperature influence factor; T t T represents the current temperature inside the tunnel. s The preset suitable construction temperature;

[0133] The formula for calculating the humidity impact factor is as follows:

[0134] ;

[0135] In the formula, E4 is the humidity influence factor; RHt The current humidity level inside the tunnel; RH s The preset suitable construction humidity;

[0136] Rainfall impact factors can be determined using rainfall data (including rainfall amount, etc.). Specifically, the calculation formula for rainfall impact factors is as follows:

[0137] ;

[0138] In the formula, E5 is the rainfall impact factor; R t R represents the rainfall at the current moment. s The preset safety threshold for rainfall;

[0139] The formula for calculating comprehensive environmental factors is as follows:

[0140] ;

[0141] In the formula, w1, w2, w3, w4, and w5 are weighting coefficients. Generally, w1 takes a value of 0.3-0.5, w2 takes a value of 0.2-0.3, w3 takes a value of 0.05-0.1, w4 takes a value of 0.05-0.1, and w5 takes a value of 0.2-0.3. It should be noted that the above weighting coefficients can be optimized and determined based on historical statistical results using fuzzy logic rules or lightweight machine learning models (such as linear regression models).

[0142] Then, based on the improved Sigmoid function and considering comprehensive environmental factors, a mapping relationship model for the early warning sensitivity adjustment coefficient is determined; the expression of the mapping relationship model is as follows:

[0143] ;

[0144] In the formula, Z is the early warning sensitivity adjustment coefficient; Z is the comprehensive environmental factor, whose value is not less than zero; k is the steepness coefficient greater than zero, which is generally taken as 0.1-1.

[0145] This invention overcomes the limitations of fixed early warning thresholds, enabling safety standards to be dynamically and adaptively adjusted based on real-time environmental conditions at the construction site. Specifically, when the construction environment deteriorates, the value of the comprehensive environmental factors increases, causing the early warning sensitivity adjustment coefficient to decrease. Consequently, the dynamic safety probability early warning threshold decreases, reducing the conditions for triggering safety risk warnings and improving the sensitivity of safety risk identification to prevent missed alarms and enhance construction safety. When the construction environment is favorable, restoring or increasing the dynamic safety probability early warning threshold can reduce false alarm interference.

[0146] like Figure 5As shown, in a preferred embodiment of the present invention, the step of identifying the risk of collapse ahead of tunnel excavation based on a dynamic safety probability early warning threshold and a risk posterior probability value, specifically step S400, includes:

[0147] S410. Compare the posterior probability value of the risk with the dynamic safety probability warning threshold to obtain the comparison result;

[0148] S420. Based on the comparison results, determine the risk level of collapse ahead of the tunnel excavation and trigger the corresponding early warning signal.

[0149] In practical applications, multiple risk probability warning intervals can be set based on dynamic safety probability warning thresholds, with each interval corresponding to a risk level. The risk level corresponding to the risk probability warning interval into which the aforementioned posterior probability value falls is determined as the current risk level. Risk levels include, but are not limited to, four levels: safe, caution, warning, and danger. For example, if the posterior probability value... If the risk level is 0, then the current risk level is determined to be safe; if the posterior probability value of the risk is 0. If the risk level is determined to be "attention," then the current risk level is determined to be "attention." If the posterior probability value of the risk is... If the risk level is determined to be a warning, then the posterior probability value of the risk is... If so, the current risk level is determined to be dangerous.

[0150] like Figure 6 As shown, in another embodiment of the present invention, a smart identification system for highway construction safety risks is also provided to implement the above-mentioned smart identification method for highway construction safety risks, comprising:

[0151] Model building module 10 is used to build a physically constrained dynamic Bayesian network model based on the geological conditions and construction parameters of the tunnel project to predict the risk of collapse ahead of tunnel excavation.

[0152] Risk reasoning module 20 is used to perform real-time risk probability reasoning based on physical constraint dynamic Bayesian network model and real-time monitoring data to obtain risk posterior probability values.

[0153] The threshold adaptive adjustment module 30 is used to adaptively adjust the safety probability early warning threshold based on multi-source dynamic data related to the construction environment, so as to obtain the dynamic safety probability early warning threshold.

[0154] The risk identification module 40 is used to identify the risk of collapse ahead of tunnel excavation based on the dynamic safety probability warning threshold and the risk posterior probability value.

[0155] like Figure 7 As shown, in a preferred embodiment of the present invention, the model building module 10 specifically includes:

[0156] The variable determination unit 11 is used to determine the key physical state variables and monitoring variables that affect the collapse based on the tunnel engineering geological conditions and construction parameters.

[0157] Causal relationship establishment unit 12 is used to establish a causal relationship network between various key physical state variables based on the basic principles of geotechnical mechanics and engineering experience, which serves as the node and structural topology of the dynamic Bayesian network.

[0158] Model generation unit 13 is used to transform the physical laws in the causal relationship network into constraints on the node state transition probability and conditional probability distribution, thereby completing the construction of a physical constraint dynamic Bayesian network model.

[0159] like Figure 8 As shown, in a preferred embodiment of the present invention, the risk reasoning module 20 specifically includes:

[0160] The monitoring data acquisition unit 21 is used to acquire monitoring data in real time, obtain real-time monitoring data, and use the real-time monitoring data as evidence input for the corresponding monitoring variables in the physical constraint dynamic Bayesian network model.

[0161] Bayesian inference unit 22 is used to update the conditional probability distribution of the state of each node in the physical constraint dynamic Bayesian network model based on evidence input and using Bayesian inference algorithm, and output the posterior probability value of the risk of collapse in front of the tunnel at the current moment.

[0162] like Figure 9 As shown, in a preferred embodiment of the present invention, the threshold adaptive adjustment module 30 specifically includes:

[0163] The environmental data acquisition unit 31 is used to synchronously acquire multi-source dynamic environmental data related to the construction environment at the current moment.

[0164] The mapping relationship establishment unit 32 is used to establish a mapping relationship model between multi-source dynamic environmental data and early warning sensitivity adjustment coefficients;

[0165] The adjustment coefficient calculation unit 33 is used to calculate the early warning sensitivity adjustment coefficient at the current moment based on the current environmental multi-source dynamic data through a mapping relationship model.

[0166] The threshold dynamic adjustment unit 34 is used to dynamically adjust the preset benchmark safety probability warning threshold based on the warning sensitivity adjustment coefficient at the current moment, and generate a dynamic safety probability warning threshold suitable for the current construction environment.

[0167] like Figure 10 As shown, in a preferred embodiment of the present invention, the risk identification module 40 specifically includes:

[0168] Risk probability comparison unit 41 is used to compare the posterior risk probability value with the dynamic safety probability warning threshold to obtain the comparison result;

[0169] Risk level determination unit 42 is used to determine the risk level of collapse ahead of tunnel excavation based on the comparison results and trigger the corresponding early warning signal.

[0170] It should be noted that the above modules and units can be implemented as a computer program, which can run on a computer device. The computer device's memory can store the computer program that makes up the modules or units, enabling the processor to execute the various steps of the above method.

[0171] It should be understood that although the steps in the flowcharts of the embodiments of the present invention are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in each embodiment may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0172] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods.

[0173] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A highway construction safety risk intelligent identification method, characterized in that, The method comprises the following steps: Based on the tunnel engineering geological conditions and construction parameters, a physically constrained dynamic Bayesian network model for predicting the collapse risk in front of the tunnel excavation is constructed; Based on the physically constrained dynamic Bayesian network model, real-time risk probability reasoning is performed according to real-time monitoring data to obtain a risk posterior probability value; According to the multi-source dynamic data related to the construction environment, the safety probability early warning threshold is adaptively adjusted to obtain a dynamic safety probability early warning threshold; Based on the dynamic safety probability early warning threshold, the risk of collapse in front of the tunnel excavation is identified according to the risk posterior probability value. 2.The intelligent identification method for highway construction safety risks according to claim 1, characterized in that, The step of constructing a physically constrained dynamic Bayesian network model for predicting the collapse risk in front of the tunnel excavation based on the tunnel engineering geological conditions and construction parameters specifically comprises: According to the tunnel engineering geological conditions and construction parameters, the key physical state variables and monitoring variables affecting the collapse are determined; Based on the basic principles of rock mechanics and engineering experience, a causal relationship network among the key physical state variables is established, which is used as the node and structure topology of the dynamic Bayesian network; The physical laws in the causal relationship network are converted into constraint conditions for the node state transition probability and conditional probability distribution, and the construction of the physically constrained dynamic Bayesian network model is completed. 3.The intelligent identification method for highway construction safety risks according to claim 2, characterized in that, The key physical state variables include the stress state of surrounding rock, the plastic zone range, the equivalent stiffness of the supporting structure, and the seepage pressure of groundwater; the monitoring variables include the convergence displacement around the hole, the settlement displacement of the vault, the anchor rod axial force, the initial support concrete stress, the advanced geological prediction result, and the excavation parameter. 4.The intelligent identification method for highway construction safety risks according to claim 1, characterized in that, The step of performing real-time risk probability reasoning based on the physically constrained dynamic Bayesian network model according to real-time monitoring data to obtain a risk posterior probability value specifically comprises: Real-time monitoring data is obtained by real-time acquisition of monitoring data, and the real-time monitoring data is input as evidence into the corresponding monitoring variables in the physically constrained dynamic Bayesian network model; Based on the evidence input, the conditional probability distribution of each node state in the physically constrained dynamic Bayesian network model is updated using the Bayesian inference algorithm, and the risk posterior probability value of the collapse in front of the tunnel excavation at the current time is output. 5.The intelligent identification method for highway construction safety risks according to claim 1, characterized in that, The step of adaptively adjusting the safety probability early warning threshold based on the multi-source dynamic data related to the construction environment to obtain a dynamic safety probability early warning threshold specifically comprises: Synchronously acquiring environment multi-source dynamic data related to the construction environment at the current time; Establishing a mapping relationship model between the environment multi-source dynamic data and the early warning sensitivity adjustment coefficient; According to the environment multi-source dynamic data at the current time, the early warning sensitivity adjustment coefficient at the current time is calculated through the mapping relationship model; Based on the early warning sensitivity adjustment coefficient at the current time, the preset baseline safety probability early warning threshold is dynamically adjusted to generate a dynamic safety probability early warning threshold suitable for the current construction environment. 6.The intelligent identification method for highway construction safety risks according to claim 5, characterized in that, The step of establishing a mapping relationship model between the environment multi-source dynamic data and the early warning sensitivity adjustment coefficient specifically comprises: Determine a comprehensive environmental factor according to environmental multi-source dynamic data; the environmental multi-source dynamic data includes super-geological prediction results, blasting vibration monitoring data, tunnel internal temperature and humidity data, and rainfall data; the comprehensive environmental factor is determined by weighted linear fusion of terrain degradation trend factor, blasting vibration intensity factor, temperature influence factor, humidity influence factor, and rainfall influence factor; the terrain degradation trend factor is determined by the geological condition rating in the super-geological prediction results; the blasting vibration intensity factor is determined by the blasting vibration monitoring data; the temperature influence factor and the humidity influence factor are determined by the tunnel internal temperature and humidity data; and the rainfall influence factor is determined by the rainfall data; Based on the improved Sigmoid function, a mapping relationship model of the early warning sensitivity adjustment coefficient is determined according to the comprehensive environmental factor. 7.The intelligent identification method for highway construction safety risks according to claim 6, characterized in that, The expression of the mapping relationship model is as follows: ； In the formula, is a pre-warning sensitivity adjustment coefficient; Z is the comprehensive environmental factor, and its value is not less than zero; k is an inclination coefficient greater than zero. 8.The intelligent identification method for highway construction safety risks according to claim 1, characterized in that, Based on the dynamic safety probability early warning threshold, according to the risk posterior probability value, the steps of identifying the risk of collapse in front of tunnel excavation are as follows: The risk posterior probability value is compared with the dynamic safety probability early warning threshold to obtain a comparison result; According to the comparison result, the risk level of collapse in front of tunnel excavation is determined, and a corresponding early warning signal is triggered.

9. A highway construction safety risk intelligent identification system for implementing the highway construction safety risk intelligent identification method of any one of claims 1-8, characterized in that, It includes: A model construction module for constructing a physically constrained dynamic Bayesian network model for predicting the collapse risk in front of tunnel excavation based on the tunnel engineering geological conditions and construction parameters; A risk reasoning module for performing real-time risk probability reasoning based on the physically constrained dynamic Bayesian network model according to real-time monitoring data to obtain a risk posterior probability value; A threshold self-adaptive adjustment module for self-adaptively adjusting the safety probability early warning threshold according to multi-source dynamic data related to the construction environment to obtain a dynamic safety probability early warning threshold; A risk identification module for identifying the risk of collapse in front of tunnel excavation based on the dynamic safety probability early warning threshold according to the risk posterior probability value. 10.The intelligent identification system for highway construction safety risks according to claim 9, characterized in that, The model construction module specifically includes: A variable determination unit for determining key physical state variables and monitoring variables affecting collapse according to the tunnel engineering geological conditions and construction parameters; A causal relationship establishment unit for establishing a causal relationship network among the key physical state variables based on the basic principles of rock-soil mechanics and engineering experience, and using the network as the node and structure topology of the dynamic Bayesian network; A model generation unit for converting the physical laws in the causal relationship network into constraint conditions of the node state transition probability and conditional probability distribution, and completing the construction of the physically constrained dynamic Bayesian network model; The risk reasoning module specifically includes: A monitoring data acquisition unit for acquiring real-time monitoring data by real-time monitoring data acquisition, and inputting the real-time monitoring data as evidence into the corresponding monitoring variables in the physically constrained dynamic Bayesian network model; A Bayesian inference unit for updating the conditional probability distribution of each node state in the physically constrained dynamic Bayesian network model by using the Bayesian inference algorithm based on the evidence input, and outputting the risk posterior probability value of collapse in front of tunnel excavation at the current time. The threshold adaptive adjustment module specifically comprises: An environment data acquisition unit, configured to synchronously acquire environment multi-source dynamic data related to a construction environment at a current time; A mapping relationship establishment unit, configured to establish a mapping relationship model between the environment multi-source dynamic data and an early warning sensitivity adjustment coefficient; An adjustment coefficient calculation unit, configured to calculate the early warning sensitivity adjustment coefficient at the current time through the mapping relationship model according to the environment multi-source dynamic data at the current time; A threshold dynamic adjustment unit, configured to dynamically adjust a preset reference safety probability early warning threshold based on the early warning sensitivity adjustment coefficient at the current time, and generate a dynamic safety probability early warning threshold suitable for a current construction environment; The risk identification module specifically comprises: A risk probability comparison unit, configured to compare the risk posterior probability value with the dynamic safety probability early warning threshold to obtain a comparison result; A risk grade determination unit, configured to determine a risk grade of collapse occurring in front of the tunnel excavation according to the comparison result, and trigger a corresponding early warning signal.

Images