Multi-source surveying and mapping fusion disaster early warning method for civil air defense engineering surrounding rock

Abstract

The application relates to the technical field of civil engineering, and discloses a multi-source surveying and mapping fusion civil air defense engineering surrounding rock disaster warning method. The application establishes a cascade transmission chain model of three levels of surrounding rock level, interface level and sensing level, and simultaneously estimates the three levels of states in the same inversion problem, so that the following effects are achieved: the three levels of states are mutually constrained through two levels of transmission relationship, the complementarity of point observation and surface observation is improved from data layer to model layer coupling, the observation distortion caused by interface decoupling is transformed from implicit assumption into identifiable model parameters, and a disaster incubation area can be located; sensing degradation is transformed from error source into precursor signal source, the spatiotemporal sequence fitting failure wave front of the degradation event directly gives the extension direction and disaster time window of the incubation area; the order of the three levels of precursors is ensured by the physical transmission direction of the transmission chain, the warning grading is free from the experience threshold, and the warning time window is overall moved forward.

Description

Technical Field

[0001] This application relates to the field of civil engineering technology, and in particular to a method for early warning of surrounding rock disasters in civil defense projects using multi-source mapping fusion. Background Technology

[0002] Monitoring and disaster early warning of surrounding rock in underground engineering is an important technical direction in the field of geotechnical engineering safety, and it is widely used in projects such as traffic tunnels, mine roadways, and urban underground spaces. As an important component of urban underground spaces, civil defense projects experience a gradual evolution of damage to the surrounding rock under the long-term effects of high ground stress, groundwater, and material aging. This damage progresses from local incubation to expansion and ultimately to disaster. Therefore, early warning of surrounding rock disasters has become a key technical issue for the safe operation and maintenance of civil defense projects. Currently, multi-point sensing methods such as multi-point displacement gauges and anchor stress gauges, along with area mapping methods such as 3D laser scanning, are commonly used to conduct multi-source monitoring of the surrounding rock in underground structures.

[0003] For multi-source monitoring data, existing methods mainly assess the surrounding rock condition based on thresholds after data fusion. For example, Chinese patent application CN117514357A discloses a method for early warning of surrounding rock instability in tunnel deformation monitoring, which determines the deformation warning value based on the longitudinal deformation curve of the tunnel excavation and the ultimate strain of the rock mass, and then judges the stability of the surrounding rock by combining the monitoring data; Chinese patent application CN122574571A discloses a tunnel deformation monitoring method based on multi-source data fusion, which fuses the internal mechanical data of the surrounding rock collected by the sensor network with the surface image data collected by the camera through Kalman filtering to obtain the internal state data of the surrounding rock. However, the above methods have the following shortcomings: First, point observations and area observations are only superimposed on the data layer. They are projections of the same mechanical state of the surrounding rock at different scales, which cannot locate the disaster-inducing zone. Second, the observation object of area mapping is the lining surface. The void behind the lining causes its deformation to be decoupled from the actual deformation of the surrounding rock, resulting in a systematic distortion of the evaluation criteria. Third, the sensing elements undergo adhesion degradation as deformation develops. Existing methods treat this as noise and discard the precursory information carried by the degradation. Fourth, the warning level is divided according to empirical thresholds, which lacks physical basis and results in a late warning time window. Summary of the Invention

[0004] It has the advantages of advanced identification, precise positioning and physical classification, which can solve the problems in the existing technology that point sensing and area mapping are only superimposed on the data layer, the observation constraints of the lining surface are distorted, and the sensor degradation information is mistakenly discarded, resulting in the difficulty in locating the disaster incubation area and the late warning time window.

[0005] To achieve the above objectives, this application adopts the following technical solution: a method for early warning of surrounding rock disasters in civil defense engineering through multi-source mapping fusion, the specific steps of which are as follows: S1. A chain-type co-located observation network is set up along the caverns of the civil defense project to acquire and solve multi-source observation data, and obtain the planar deformation field and point multi-channel time series. S2. Establish a cascaded transmission chain model, which includes three levels of state variables: surrounding rock level, interface level, and sensing level, and two levels of transmission relationships that express the hierarchical mapping of the three levels of states. S3. Implement chain-style joint inversion, establish a joint objective function with the planar deformation field as the boundary constraint and the point-like multi-channel time series as the internal constraint, simultaneously estimate the third-level state variables in the same inversion problem, and output the mechanical state field and transmission efficiency distribution inside the surrounding rock. S4. Construct a three-level precursor criterion according to the physical hierarchy of the cascaded transmission chain model, and determine the warning level based on the triggering state; S5. Fit the spatiotemporal sequence of sensor channel degradation events to obtain the propagation direction and speed of the failure wavefront, and calibrate the expansion direction of the disaster incubation zone and the disaster time window. S6. Spatial superposition of the field anomaly zone of the mechanical state field inside the surrounding rock with the envelope zone of the failure wavefront, the intersection is determined as the disaster incubation zone and its location result and early warning are output.

[0006] Furthermore, in step S1, the chain-type co-located observation network is composed of several chain links arranged in an array. Each chain link has a co-located observation group at the same location, which is a three-level carrier that penetrates the surrounding rock, the interface, and the sensing level. The observation group includes multi-point displacement gauges, anchor stress gauges, interface contact pressure sensors, and acoustic emission probes, as well as distributed sensing optical cables laid along the lining circumferential direction and surface mapping elements covering the chain link. A chain link index is established to record the co-located correspondence of the three-level observation channels within each chain link. The spatial distance between adjacent chain links is less than the expected expansion scale of the disaster-inducing zone. The chain links in geologically weak sections and structurally complex sections are densely arranged.

[0007] Furthermore, in step S1, the multi-source observation data are uniformly reduced to the same coordinate system and the same time reference, and are collected according to the link index to form an observation data package with the link as the unit. The three-level observation channel data in the same observation data package carry the same link identifier and time scale.

[0008] Furthermore, in step S2, the surrounding rock level state includes the surrounding rock damage field and stress-strain field; the interface level state is expressed by the contact stiffness and fit between the lining and the surrounding rock; the sensing level state is expressed by the transmission efficiency of each sensing channel, where the transmission efficiency is the ratio of the sensor element output value to the actual deformation of the matrix; the two-level transmission relationship includes the first-level transmission relationship expressing the mapping from the surrounding rock level state to the interface level state and the second-level transmission relationship expressing the mapping from the interface level state to the sensing level state; the measured values ​​of each sensing channel in the observation equation are converted according to the transmission efficiency of that channel and then used in the calculation; the planar deformation field is converted into boundary constraints by surface element according to the interface level state.

[0009] Furthermore, in step S2, the first-level transmission relationship is determined by the contact stiffness and friction coefficient between the lining and the surrounding rock, and the second-level transmission relationship is determined by the bond stiffness between the sensing element and the matrix. Before the inversion, the initial value of the surrounding rock level state is taken from the mechanical parameters of the corresponding level in the surrounding rock classification table, the initial value of the interface level state is set according to the complete fit between the lining and the surrounding rock, and the initial value of the sensing level state is taken from the nominal transmission coefficient of the sensing element. The parameters of the first-level transmission relationship and the second-level transmission relationship are iteratively updated along with the three-level state variables in the chain joint inversion in step S3.

[0010] Furthermore, in step S3, the constraint weights of each sensing channel in the joint objective function are taken as the normalized values ​​of the channel's transmission efficiency. After the inversion converges, the transmission efficiencies of each measuring point and each surface element corresponding to the sensing channel are collected position by position to form an observation confidence field. The transmission efficiency values ​​of the degraded sensing channels are output along with the inversion results.

[0011] Furthermore, in step S4, the three-level precursor criteria include: a sensor-level precursor is triggered when the transmission efficiency of the sensing channel within the link decays beyond the drift threshold of that channel; an interface-level precursor is triggered when the contact pressure at the interface within the link decays below the theoretical sharing value or when the boundary of the contact state partition migrates; and a surrounding rock-level precursor is triggered when a strain energy density accumulation zone, a gradient abrupt change zone in the deformation field, or a cluster of acoustic emission events appears in the mechanical state field inside the surrounding rock. The identification is carried out step by step in the order of sensor-level precursor, interface-level precursor, and surrounding rock-level precursor.

[0012] Furthermore, in step S5, the spatiotemporal sequence is composed of the occurrence time and spatial coordinates of the sensing level precursors and sensing channel failure events of each chain segment recorded. The fitting includes: calculating the propagation direction and propagation speed of the failure wavefront according to the time difference of the arrival of the degradation event at each chain segment.

[0013] Furthermore, in step S6, the superimposed objects of spatial superposition also include the rupture location results of the acoustic emission event. The location results include the spatial coordinates, range, expansion direction and disaster time window of the disaster incubation zone. The credibility of each location is marked in the early warning conclusion, and the credibility is taken as the transmission efficiency of the corresponding location.

[0014] Furthermore, after the early warning is output in step S6, the mapping acquisition frequency and acoustic emission sampling density of the chain segment array where the disaster-incubating zone is located are increased according to the early warning level and the location of the disaster-incubating zone. The observation task of the degraded chain segment is taken over by the observation channel and surface mapping channel of the adjacent chain segment. The updated observation data is reinjected into step S3 for re-inversion. When the deformation rate of the disaster-incubating zone meets the acceleration criterion and the transmission efficiency of the chain segment is lower than the preset threshold, all observation tasks of the chain segment are switched to the non-contact observation channel.

[0015] The beneficial effects of this invention are as follows: This application provides a multi-source mapping fusion method for early warning of rock-surrounding disasters in civil defense engineering. By establishing a cascaded transmission chain model of three levels of states—rock-surrounding, interface, and sensor—and simultaneously estimating the three levels of states in the same inversion problem, the following effects are achieved: the three levels of states are mutually constrained by the two-level transmission relationship; the complementarity of point observations and area observations is improved from data layer superposition to model layer coupling; the observation distortion caused by interface decoupling is transformed from implicit assumptions into identifiable model parameters, and the disaster incubation zone can be located; sensor degradation is transformed from an error source into a precursor signal source; the failure wavefront of the spatiotemporal sequence fitting of degradation events directly gives the expansion direction of the incubation zone and the disaster time window; the order of the three levels of precursors is guaranteed by the physical transmission direction of the transmission chain; the early warning classification is freed from empirical thresholds; and the overall early warning time window is shifted forward; the observation credibility field is endogenously generated by the chain state; the early warning conclusion carries location-specific credibility; observation resources are closed-loop scheduled according to the location of the incubation zone; and the supply of precursor information remains continuous during the pre-disaster stage. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort: Figure 1 This is a diagram illustrating the method steps of this application. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Civil defense projects operate long-term in underground rock and soil environments. The surrounding rock and lining together form a load-bearing system, and their safety and stability directly affect the project's protective and peacetime functions. During long-term service, the surrounding rock is subjected to continuous effects of high ground stress, groundwater, construction disturbance, and material aging. Its damage is a gradual process that begins locally, expands progressively, and eventually leads to overall disaster. Currently, the condition of the surrounding rock is mainly assessed through three methods: manual inspection, point sensor monitoring, and periodic surface mapping. Manual inspection is insufficient to detect the accumulation of damage within the lining. Point sensors such as displacement gauges and stress gauges can only provide readings at isolated points, lacking spatial correlation between measurement points and failing to delineate the anomaly range on the surface. Mapping methods such as 3D laser scanning and photogrammetry can acquire planar deformation of the tunnel interior surface, but their observation object is the lining surface. When there is a gap or slippage between the lining and the surrounding rock, the deformation of the lining surface cannot represent the true deformation of the surrounding rock. More importantly, as the surrounding rock deforms, the interface contact state and the bonding state of the sensing elements deteriorate simultaneously, leading to gradual distortion of the sensor readings. Current processing methods typically involve simply removing distorted data, resulting in less usable observational information as the disaster approaches. These intertwined problems make it difficult to pinpoint the location of the disaster-prone zone and predict the timing of the disaster. However, with the development of distributed fiber optic sensing, 3D laser scanning, and multi-source data fusion technologies, it has become possible to treat the surrounding rock, interface, and sensing elements as a single transmission chain for holistic observation and joint calculation. This allows for the location of the potential zone and the prediction of the disaster's time window before it occurs, assisting managers in evacuating personnel, relocating equipment, and reinforcing facilities in advance, providing a scientific basis for engineering maintenance decisions.

[0019] refer to Figure 1 This application provides a multi-source mapping fusion method for early warning of rockfall disasters in civil defense engineering, including the following steps: Step S1 involves deploying a chain-like co-located observation network along the caverns of the civil defense project to acquire and process multi-source observation data, obtaining the planar deformation field and point-like multi-channel time series. The chain-like co-located observation network consists of several chain links arranged in an array. A chain link is a basic observation unit formed by deploying multiple observation methods at the same location. The gestation process of surrounding rock disasters is transmitted cascaded along three levels: surrounding rock, lining-surround rock interface, and sensing elements. That is, the development of surrounding rock damage alters the force transmission conditions at the lining-surround rock interface, and the deterioration of the interface force transmission conditions further alters the strain transmission conditions of the sensing elements. These three levels constitute the same causal transmission chain. If various sensors are deployed separately according to monitoring items such as displacement, stress, and strain, the three levels of states at the same location cannot be compared temporally and spatially, and the transmission process cannot be observed. Therefore, each chain link deploys an observation group at the same location that penetrates the three levels of the surrounding rock, interface, and sensing elements. The observation group includes a multi-point displacement gauge, an anchor stress gauge, an interface contact pressure sensor and an acoustic emission probe, as well as a distributed sensing optical cable laid along the circumferential direction of the lining and a surface mapping element covering the link section.

[0020] The data sources and formats for each observation channel are as follows: Deep displacement data of the surrounding rock is acquired using multi-point displacement gauges. These gauges are installed along boreholes perpendicular to the lining surface, with multiple measuring points at different depths within the borehole. The output is the displacement reading of each measuring point relative to the borehole opening, in millimeters, reflecting the displacement distribution within the surrounding rock along the borehole direction. Support load data is acquired using anchor bolt stress gauges. These gauges are installed on the body of the system anchor bolts, and their output is the axial load of the anchor bolt, in kilonewtons, reflecting the interaction between the surrounding rock and the support structure. Interface contact pressure data is acquired using interface contact pressure sensors. These sensors are pre-embedded at the contact surface between the lining and the surrounding rock, and their output is the normal pressure at the contact surface, in megapascals, directly characterizing the interface force transmission conditions. Acoustic emission event data is acquired using acoustic emission probes. These probes are installed on the inner surface of the lining or in shallow holes, and their output includes characteristic parameters such as the occurrence time, amplitude, energy, and ring count of each acoustic emission event, reflecting the intensity and frequency of fracture activity within the surrounding rock. Circumferential strain data of the lining is acquired through distributed sensing optical cables, such as those based on Brillouin scattering. These cables are continuously laid along the circumferential direction of the lining and bonded to the lining structure. The output is a continuously distributed strain sequence along the cable, measured in microstrain, with a spatial resolution of, for example, 0.5m to 1m. Point cloud data of the tunnel interior surface is acquired through 3D laser scanning or close-range photogrammetry. For example, a 3D laser scanner is used to scan each link section in stages at fixed stations, dividing the point cloud into several surface elements, each corresponding to a surface mapping observation point. These six types of data differ in their measurement objects, spatial scale, and sampling frequency: point-type sensing channels have a high sampling frequency and point spatial coverage, while mapping channels have surface spatial coverage and a low sampling frequency, thus complementing each other in observation.

[0021] After deployment, a link index is established. The link index records the co-location correspondence of the three-level observation channels within each link, and also records the spatial correspondence between each measuring point and each mapping element, allowing any observation data to be traced back to its corresponding link and location. The spatial spacing between adjacent links is less than the expected expansion scale of the disaster-inducing zone, for example, one-third to one-half of the expected expansion scale. The expected expansion scale of the inducing zone can be determined statistically based on the surrounding rock grade, cavern span, and disaster cases of similar projects. The spacing value here is constrained by the wavefront fitting requirements of subsequent steps: if the link spacing is too large, the propagation order of degradation events on the link array will be indistinguishable; if the link spacing is too small, the deployment cost will be too high. Using the above range, a balance can be achieved between wavefront resolution and deployment cost. For geologically weak sections such as fault fracture zones, sections with weak interlayers, and cavern intersections, as well as structurally complex sections such as the junction of the arch, sidewalls, and invert, the link arrangement is densified.

[0022] In step S1, the multi-source observation data are uniformly converted to the same coordinate system and the same time reference. Point clouds and various point sensors each have independent coordinate references and sampling clocks. If they are directly mixed, observations at the same physical location will be spatially misaligned, and observations of the same physical process will be temporally misaligned. Therefore, control reference points are set up inside the cavern, using these reference points as a common constraint to convert the coordinates of each period's point cloud and each measuring point to the same coordinate system. The timestamps of each observation device are unified to the same clock source, for example, by using network time synchronization or a clock synchronization device to synchronize the time of each acquisition terminal. The converted observation data is collected according to the link index, forming observation data packages with links as units. The three-level observation channel data within the same observation data package carry the same link identifier and timestamp, allowing any set of data retrieved later to be compared in spatial location and acquisition time. Data processing also includes outlier removal and filtering: readings exceeding the physical range are removed, and point-like time series are subjected to moving averages or low-pass filtering to suppress high-frequency noise caused by environmental disturbances. The solution process for the planar deformation field is as follows: Point clouds on the inner surfaces of adjacent tunnel sections are registered, and surface elements with registration residuals exceeding the limit are marked as surface elements to be checked; a difference operation is performed on the registered point clouds of the two sections to obtain the three-dimensional deformation vector of each surface element; the three-dimensional deformation vectors of each surface element form the overall planar deformation field and its deformation rate distribution. The point-based multi-channel time series includes the deep displacement of the surrounding rock at each link, anchor bolt load, interface contact pressure, lining circumferential strain, and acoustic emission event time series.

[0023] Step S2, referencing the model, establish a cascaded transmission chain model. The cascaded transmission chain model includes three levels of state variables: surrounding rock level, interface level, and sensor level, as well as two levels of transmission relationships expressing the hierarchical mapping of these three levels. The surrounding rock level state includes the surrounding rock damage field and stress-strain field, expressed using a numerical mechanics model: the model geometry is imported from the three-dimensional geometric model of the cavern, and after meshing, a surrounding rock computational domain is formed. Boundary conditions and material parameters are updated iteratively with inversion. The interface level state is expressed by the contact stiffness and fit between the lining and the surrounding rock, where the fit characterizes the proportion of the effective force transmission area of ​​the contact surface; the fit at the detached part approaches zero. The sensor level state is expressed by the transmission efficiency of each sensor channel, which is the ratio of the sensor element output value to the actual deformation of the substrate. For example, if the actual deformation of the substrate where the anchor stress gauge is located in a certain link is 2.0 mm, and the deformation reading corresponding to the anchor stress gauge output value is 1.6 mm, then the transmission efficiency of this sensor channel is 1.6 / 2.0 = 0.8. A transmission efficiency of 1 indicates that the strain is completely transmitted. The greater the transmission efficiency deviates from 1, the more severe the distortion of the channel.

[0024] The two-level transmission relationship includes a first-level transmission relationship and a second-level transmission relationship. The first-level transmission relationship expresses the mapping from the surrounding rock state to the interface state, determined by the contact stiffness and friction coefficient between the lining and the surrounding rock: the interface shear stress equals the product of the contact stiffness and the relative displacement on both sides of the interface, and does not exceed the upper limit of friction determined by the product of the friction coefficient and the interface normal pressure; when the surrounding rock deformation develops to the point where the interface shear stress reaches the upper limit of friction, or the interface normal pressure drops to zero, the contact stiffness decreases, the fit decreases, and the lining surface deformation decouples from the surrounding rock deformation. The second-level transmission relationship expresses the mapping from the interface state to the sensing state, determined by the bonding stiffness between the sensing element and the substrate: the sensing element output equals the product of the bonding transfer coefficient and the substrate deformation; the bonding stiffness deteriorates as the substrate deformation increases, and the transmission efficiency decreases accordingly. From the above transmission relationships, it can be seen that the lining surface deformation is the result of the surrounding rock deformation transmitted through the interface, and the sensing reading is the result of the interface deformation transmitted through the bonding process. If the deformation of the lining surface is directly equated with the deformation of the surrounding rock, and the sensor readings are directly equated with the deformation of the matrix, the observed values ​​will systematically deviate from the true deformation at locations where the interface is decoupled or the bond is deteriorated. The observation constraints established in this way will cause the overall inversion results to be distorted. Therefore, the measured values ​​of each sensor channel in the observation equation are converted according to the transmission efficiency of that channel before participating in the calculation, and the planar deformation field is converted into boundary constraints by converting it into surface elements according to the interface-level state.

[0025] Before the inversion begins, the cascaded transfer chain model requires initial values ​​for the three-level state variables and the two-level transfer relationship parameters. Due to significant differences in surrounding rock grades, lining forms, and construction techniques among various civil defense projects, initial values ​​that deviate from the actual engineering conditions will increase the number of inversion iterations and may even lead to convergence errors. Therefore, in one embodiment, the initial values ​​for the surrounding rock level state are taken from the mechanical parameters of the corresponding level in the surrounding rock classification table. For example, for a surrounding rock segment classified as Level III according to the surrounding rock classification standard, the elastic modulus, Poisson's ratio, and strength parameters corresponding to that level are taken as initial values. The initial values ​​for the interface level state are set based on the assumption that the lining and surrounding rock are completely in contact, i.e., the fit is taken as 1, and the contact stiffness is taken as the equivalent contact stiffness of the lining and surrounding rock materials. The initial values ​​for the sensing level state are taken from the nominal transfer coefficient of the sensing element. The contact stiffness and friction coefficient in the first-level transfer relationship can be calibrated for their range through interface direct shear tests or in-situ field tests; the bonding stiffness in the second-level transfer relationship can be determined for its initial value through sensing element bonding calibration tests. Since the parameters of the transfer relationship are difficult to accurately calibrate one by one during the deployment stage, and they change continuously with the deformation of the surrounding rock and the evolution of the interface state, the parameters of the two-level transfer relationship are iteratively updated along with the three-level state variables in the chain joint inversion in step S3, so that the transfer relationship is refined online with the accumulation of observations.

[0026] Step S3: Perform chain-style joint inversion. A joint objective function is established using the planar deformation field as boundary constraints and the point-based multi-channel time series as internal constraints. Three-level state variables are estimated simultaneously within the same inversion problem, outputting the internal mechanical state field and transmission efficiency distribution of the surrounding rock. The joint objective function consists of two parts: boundary constraint terms and internal constraint terms. The boundary constraint terms are the sum of squares of the residuals between the model surface deformation calculated from the interface-level state at each surface element and the measured planar deformation. The internal constraint terms are the weighted sum of the residuals between the model response calculated from the two-level transmission relationship at each sensor channel and the measured readings. The constraint weight for each sensor channel is the normalized value of the channel's transmission efficiency. The inversion is solved using iterative optimization algorithms, such as constrained least squares iteration, extended Kalman filtering, or Bayesian inversion. In each iteration, the three-level state variables are mutually constrained through two-level transitive relationships: the surrounding rock state is mapped to the interface state through the first-level transitive relationship, and the interface state is mapped to the sensor state through the second-level transitive relationship. The mapping results and the residuals from the measured observations are used to inversely correct the three-level state variables. Inversion convergence is determined when the change in state variables between two adjacent iterations is less than the convergence threshold. After convergence, the distribution of the internal mechanical state field of the surrounding rock and the transmission efficiency of each sensor channel is output. The internal mechanical state field of the surrounding rock includes the displacement field, stress field, and strain field.

[0027] The effect of the aforementioned weight settings can be illustrated through a numerical example. For instance, the transmission efficiencies of the three sensor channels within the same link are 0.9, 0.6, and 0.3, respectively, and the normalized constraint weights are 0.50, 0.33, and 0.17, respectively. The higher the degree of degradation of a channel, the smaller the impact of its observation on the inversion results. It can be seen from the above that if the observation weights of the degraded channels are not adjusted, distorted observations will enter the inversion with the same weight as normal observations, directly contaminating the estimation results of the surrounding rock state. Therefore, transmission efficiency must be used as an observation weight simultaneously; and the transmission efficiency itself is estimated synchronously by the inversion, with the two mutually constraining and converging synchronously in the same inversion problem. This setting is consistent with the causal relationship of the transmission chain: sensor degradation is driven by matrix deformation, and the degree of degradation and the surrounding rock state are manifestations of the same catastrophic evolution process on different carriers. Placing both in the same inversion problem for synchronous solution avoids distorted data contaminating the inversion results and allows degradation information to be preserved as a precursor.

[0028] After the inversion converges, the transmission efficiency of each measuring point and the corresponding channel of each surface element is collected position by position to form an observation reliability field; the transmission efficiency value of the degraded sensing channel is output along with the inversion result. The observation reliability field refers to the observation reliability field distributed spatially, measured by transmission efficiency. The higher the value, the more complete the observation chain at that location and the more reliable the observation result. In this way, the reliability of the observation data no longer depends on an independent sensor diagnostic module, but is endogenously given by the inversion result. Thus, the output of step S3 provides data for subsequent steps: the internal mechanical state field of the surrounding rock is used by step S4 to identify surrounding rock-level precursors and by step S6 to delineate the field anomaly zone; the interface-level state is used by step S4 to identify interface-level precursors; the transmission efficiency distribution is used by step S4 to identify sensor-level precursors and by step S5 to construct the spatiotemporal sequence of degradation events; and the observation reliability field is used by step S6 to label the position-by-position reliability of the warning conclusion.

[0029] Step S4: Construct three-level precursor criteria according to the physical hierarchy of the cascaded transmission chain model, and determine the warning level based on the triggering state. The identification object of the three-level precursor criteria is the output of step S3. When the transmission efficiency of the sensing channel within the chain link decays beyond the drift threshold of that channel, a sensing-level precursor is triggered; the drift threshold is set, for example, as a decay amount of 10% to 20% of the nominal transmission coefficient, and can be set according to the channel type. When the interface contact pressure within the chain link decays below the theoretical sharing value, or when the boundary of the contact state partition migrates, an interface-level precursor is triggered; the theoretical sharing value refers to the theoretical value of the interface contact pressure calculated according to the load sharing ratio jointly borne by the lining and the surrounding rock. When the mechanical state field within the surrounding rock exhibits a strain energy density accumulation zone, a sudden change in deformation field gradient, or a cluster of acoustic emission events, a surrounding rock-level precursor is triggered. Strain energy density is determined by the corresponding product of stress and strain. A strain energy density accumulation zone refers to a continuous spatial region where the strain energy density exceeds a corresponding threshold. A cluster of acoustic emission events refers to a region where the number of acoustic emission events per unit time and unit space exceeds a preset multiple of the background level. Identification is performed step-by-step in the order of sensor-level precursors, interface-level precursors, and surrounding rock-level precursors. Since downstream levels in the transmission chain can produce identifiable responses at lower deformation levels than upstream levels, sensor-level precursors precede interface-level precursors, and interface-level precursors precede surrounding rock-level precursors. These three levels of precursors constitute a sequential warning sequence with progressively increasing severity. The warning level is determined based on the triggering state of the three levels of precursors. For example, triggering only a sensor-level precursor is classified as a low-level warning; further triggering an interface-level precursor upgrades it to a medium-level warning; and further triggering a surrounding rock-level precursor upgrades it to a high-level warning. The migration of early warning levels corresponds one-to-one with the objective process of disaster progression along the transmission chain, thus providing a physical basis for early warning classification and avoiding the subjectivity of artificially classifying levels based on experience thresholds.

[0030] Step S5: Referencing and fitting the spatiotemporal sequence of sensor channel degradation events, obtain the propagation direction and velocity of the failure wavefront, and calibrate the expansion direction of the disaster incubation zone and the disaster time window. The spatiotemporal sequence consists of the recorded sensor-level precursors and the occurrence time and spatial coordinates of sensor channel failure events for each link. Specifically, an event record is established for each degradation event, including the link identifier, channel identifier, event type, occurrence time, and spatial coordinates; the event type is distinguished into two categories: sensor-level precursor triggering and complete channel failure. All event records are arranged according to their occurrence time, forming the spatiotemporal sequence of degradation events. The failure wavefront refers to the spatiotemporal envelope formed by the sequential occurrence of sensor channel degradation events on the link array. When the disaster incubation zone expands, it drives the links on the expansion path to degrade in sequence. Therefore, the spatiotemporal distribution of degradation events carries the direction and rate information of the incubation zone expansion, and sensor degradation is thus transformed from observation error into a precursor signal.

[0031] The fitting algorithm is as follows. The failure wavefront is approximated as a planar wavefront on the scale of the chain link array. Let θ be the angle between its propagation direction and the reference direction, and let v be the propagation speed. Then, the arrival time of the degradation event at link i satisfies ti = t0 + (xi·cosθ + yi·sinθ) / v, where xi and yi are the spatial coordinate components of link i, and t0 is the wavefront's starting time. The least-squares fit is performed on this equation using the measured arrival times of degradation events at each link to solve for the propagation direction θ and propagation speed v. For example, links A, B, and C, equally spaced along the longitudinal direction of the cavern, are spaced 6m apart. The arrival times of the sensor-level precursors at the three links are the 10th, 13th, and 15th days, respectively. Time difference fitting shows that the propagation direction is along the longitudinal direction of the cavern from link A to link C, with a propagation speed of approximately 2m / d. If the leading edge of the failure wavefront is still 8m from the downstream link D at this time, then based on the current propagation speed, the disaster time window is estimated to be approximately 4 days. Furthermore, the propagation direction of the failure wavefront is used to determine the expansion direction of the disaster incubation zone, and the propagation speed and link spacing are used to determine the disaster time window. As the expansion of the incubation zone accelerates during the pre-disaster stage, the density of link degradation events increases accordingly. The number of event samples that can be used for fitting increases as the disaster approaches, and the increase in precursor information complements the decrease in single-channel observation accuracy.

[0032] Step S6, referencing the above, spatially superimposes the field anomaly zone of the mechanical state field within the surrounding rock with the envelope zone of the failure wavefront. The intersection is determined as the catastrophic incubation zone, and its location result and early warning are output. The field anomaly zone refers to the spatial region in the mechanical state field within the surrounding rock where the strain energy density, deformation gradient, or damage variable exceeds a corresponding threshold. The envelope zone of the failure wavefront refers to the spatial region bounded by the position of the failure wavefront leading edge and its propagation direction as the axis. The superimposed objects also include the fracture location results of acoustic emission events, which are obtained, for example, through time difference positioning of multiple acoustic emission probes. The spatial superposition is performed in the same coordinate system, and the intersection of each superimposed object is taken as the catastrophic incubation zone. The field anomaly zone delineates the anomaly range from the perspective of mechanical state, while the failure wavefront envelope zone delineates the anomaly range from the perspective of degradation event propagation. The formation mechanisms of the two types of ranges are independent of each other. If the incubation zone is delineated using only a single object, the location result is easily affected by the local error of a single observation method; therefore, the intersection of each superimposed object is determined as the catastrophic incubation zone, so that the location result is cross-verified by multiple independent pieces of evidence. The location results include the spatial coordinates, extent, direction of expansion, and time window of the disaster incubation zone. The confidence level is marked for each location in the early warning conclusion, with the confidence level determined by the transmission efficiency at that location.

[0033] The correspondence between the above output results and the technical effects is as follows: The spatial coordinates and range of the incubation zone enable personnel to directly locate the treatment position and affected area, solving the problem in existing technologies where anomalies are perceptible but their locations cannot be pinpointed; the expansion direction allows treatment forces to be prioritized in the forward areas of the incubation zone expansion; the disaster time window provides a clear time limit for personnel evacuation, equipment relocation, and reinforcement treatment; and the location-by-location credibility allows personnel to simultaneously know the reliability of the location conclusions, arranging the priority of verification and treatment according to the credibility level, avoiding over-responding to low-credibility conclusions.

[0034] The multi-source mapping fusion method for early warning of surrounding rock disasters in civil defense engineering is designed as follows: within the same link, the three-level precursors are triggered sequentially in the order of sensor level, interface level, and surrounding rock level. When the link is simultaneously located within the intersection of the anomaly zone and the failure wavefront envelope zone, the location of the disaster incubation zone is output, and the early warning level is upgraded. Based on this, personnel can quickly pinpoint the location of the incubation zone, view its expansion direction and disaster time window, and arrange for personnel evacuation, equipment relocation, and reinforcement before the disaster occurs, thereby solving to some extent the problems of delayed detection of surrounding rock disasters and difficulty in locating incubation zones. Of course, there may be situations where the triggering order of the three-level precursors within the same link is inconsistent with the above order. For example, if a sensor channel is suddenly damaged due to collision, water seepage, or other reasons, directly triggering the sensor-level precursor, while the interface-level and surrounding rock-level precursors do not respond, it indicates that the degradation of this link is not driven by the disaster incubation process. This link can be marked as a link to be verified, and cross-checked using observation data from adjacent links. If only the sensor-level precursor is triggered, and neither the interface-level precursor nor the surrounding rock-level precursor is triggered within the predetermined time, it indicates that the degradation is still at the sensor level and the surrounding rock body has not yet shown an identifiable catastrophic response. In this case, the current warning level should be maintained and observation should continue.

[0035] During the pre-disaster phase, sensing elements located in the disaster-incubating zone may physically fail before the disaster occurs. If the observation tasks are not adjusted, the precursor information chain will be interrupted during the most critical pre-disaster phase. Therefore, in one embodiment, after the early warning is output in step S6, the mapping acquisition frequency and acoustic emission sampling density of the chain segment array where the disaster-incubating zone is located are increased according to the early warning level and the location of the disaster-incubating zone; the observation tasks of the degraded chain segment are taken over by the observation channel and surface mapping channel of the adjacent chain segment; the updated observation data is reinjected into step S3 for re-inversion. When the deformation rate of the disaster-incubating zone meets the acceleration criterion and the transmission efficiency of the chain segment is lower than a preset threshold, all observation tasks of the chain segment are switched to non-contact observation channels. The acceleration criterion refers to the criterion that the deformation rate continuously increases over time, such as judging by the increment of the deformation rate exceeding a preset multiple of its current value; non-contact observation channels include observation methods that do not require physical contact with the lining or surrounding rock, such as three-dimensional laser scanning, close-range photogrammetry, and microwave interferometry. In this way, the switching of observation functions is completed before the physical failure of sensing elements, ensuring a continuous supply of precursory information during the disaster-prone phase; observation resources are concentrated on the disaster-inducing areas, improving the utilization efficiency of limited observation resources. Thus, the observation data, through inversion in step S3, identification in step S4, fitting in step S5, and superposition in step S6, forms an early warning conclusion. This conclusion drives the redistribution of observation resources, and the updated observation data re-enters the inversion process. The entire process forms a closed loop of observation, inversion, identification, and re-observation, with the cascaded transmission chain model being refined round by round as observations accumulate.

[0036] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for early warning of rockfall disasters in civil defense engineering through multi-source mapping fusion, characterized in that, The specific steps are as follows: S1. A chain-type co-located observation network is set up along the caverns of the civil defense project to acquire and solve multi-source observation data, and obtain the planar deformation field and point multi-channel time series. S2. Establish a cascaded transmission chain model, which includes three levels of state variables: surrounding rock level, interface level, and sensing level, and two levels of transmission relationships that express the hierarchical mapping of the three levels of states. S3. Implement chain-style joint inversion, establish a joint objective function with the planar deformation field as the boundary constraint and the point-like multi-channel time series as the internal constraint, simultaneously estimate the third-level state variables in the same inversion problem, and output the mechanical state field and transmission efficiency distribution inside the surrounding rock. S4. Construct a three-level precursor criterion according to the physical hierarchy of the cascaded transmission chain model, and determine the warning level based on the triggering state; S5. Fit the spatiotemporal sequence of sensor channel degradation events to obtain the propagation direction and speed of the failure wavefront, and calibrate the expansion direction of the disaster incubation zone and the disaster time window. S6. Spatial superposition of the field anomaly zone of the mechanical state field inside the surrounding rock with the envelope zone of the failure wavefront, the intersection is determined as the disaster incubation zone and its location result and early warning are output.

2. The method for early warning of rockfall disasters in civil defense engineering based on multi-source mapping fusion according to claim 1, characterized in that, In step S1, the chain-type co-located observation network is composed of several chain links arranged in an array. Each chain link has a co-located observation group at the same location that penetrates the surrounding rock, interface, and sensor levels. The observation group includes multi-point displacement gauges, anchor stress gauges, interface contact pressure sensors, and acoustic emission probes, as well as distributed sensing optical cables laid along the lining circumferential direction and surface mapping elements covering the chain link. A chain link index is established to record the co-located correspondence of the three-level observation channels within each chain link. The spatial distance between adjacent chain links is less than the expected expansion scale of the disaster-inducing zone. The chain links in geologically weak sections and structurally complex sections are densely arranged.

3. The method for early warning of rockfall disasters in civil defense engineering based on multi-source mapping fusion according to claim 2, characterized in that, In step S1, multi-source observation data are uniformly converted to the same coordinate system and the same time reference, and collected according to the link index to form observation data packages with links as units. The three-level observation channel data in the same observation data package carry the same link identifier and time scale.

4. The method for early warning of rockfall disasters in civil defense engineering based on multi-source mapping fusion according to claim 3, characterized in that, In step S2, the surrounding rock level state includes the surrounding rock damage field and stress-strain field; the interface level state is expressed by the contact stiffness and fit between the lining and the surrounding rock; the sensing level state is expressed by the transmission efficiency of each sensing channel, where the transmission efficiency is the ratio of the sensor element output value to the actual deformation of the matrix; the two-level transmission relationship includes the first-level transmission relationship expressing the mapping from the surrounding rock level state to the interface level state and the second-level transmission relationship expressing the mapping from the interface level state to the sensing level state; the measured values ​​of each sensing channel in the observation equation are converted according to the transmission efficiency of that channel and then used in the calculation; the planar deformation field is converted into boundary constraints by surface element according to the interface level state.

5. The method for early warning of rockfall disasters in civil defense engineering based on multi-source mapping fusion according to claim 4, characterized in that, In step S2, the first-level transmission relationship is determined by the contact stiffness and friction coefficient between the lining and the surrounding rock, and the second-level transmission relationship is determined by the bonding stiffness between the sensing element and the matrix. Before inversion, the initial value of the surrounding rock level state is taken from the mechanical parameters of the corresponding level in the surrounding rock classification table, the initial value of the interface level state is set according to the lining and the surrounding rock being completely in contact, and the initial value of the sensing level state is taken from the nominal transmission coefficient of the sensing element. The parameters of the first-level transitivity and the second-level transitivity are iteratively updated along with the third-level state variables in the chain joint inversion of step S3.

6. The method for early warning of rockfall disasters in civil defense engineering based on multi-source mapping fusion according to claim 5, characterized in that, In step S3, the constraint weights of each sensing channel in the joint objective function are taken as the normalized values ​​of the channel's transmission efficiency. After the inversion converges, the transmission efficiencies of each measuring point and each surface element corresponding to the sensing channel are collected position by position to form an observation confidence field. The transmission efficiency values ​​of the degraded sensing channels are output along with the inversion results.

7. The method for early warning of rockfall disasters in civil defense engineering based on multi-source mapping fusion according to claim 6, characterized in that, In step S4, the three-level precursor criteria include: a sensor-level precursor is triggered when the transmission efficiency of the sensing channel within the link decays beyond the drift threshold of that channel; an interface-level precursor is triggered when the contact pressure at the interface within the link decays below the theoretical sharing value or when the boundary of the contact state partition migrates; and a surrounding rock-level precursor is triggered when a strain energy density accumulation zone, a gradient abrupt change zone in the deformation field, or a cluster of acoustic emission events appears in the mechanical state field inside the surrounding rock. The identification is carried out step by step in the order of sensor-level precursor, interface-level precursor, and surrounding rock-level precursor.

8. The method for early warning of rockfall disasters in civil defense engineering based on multi-source mapping fusion according to claim 7, characterized in that, In step S5, the spatiotemporal sequence is composed of the occurrence time and spatial coordinates of the sensing level precursors and sensing channel failure events of each chain segment recorded. The fitting includes: calculating the propagation direction and propagation speed of the failure wavefront according to the time difference of the arrival of the degradation event at each chain segment.

9. The method for early warning of rockfall disasters in civil defense engineering based on multi-source mapping fusion according to claim 8, characterized in that, In step S6, the superimposed objects of spatial superposition also include the rupture location results of the acoustic emission event. The location results include the spatial coordinates, range, expansion direction and disaster time window of the disaster incubation zone. The credibility of each location is marked in the early warning conclusion, and the credibility is taken as the transmission efficiency of the corresponding location.

10. The method for early warning of rockfall disasters in civil defense engineering based on multi-source mapping fusion according to claim 9, characterized in that, After the warning is output in step S6, the mapping acquisition frequency and acoustic emission sampling density of the chain segment array where the disaster-incubating zone is located are increased according to the warning level and the location of the disaster-incubating zone. The observation task of the degraded chain segment is taken over by the observation channel and surface mapping channel of the adjacent chain segment. The updated observation data is reinjected into step S3 for re-inversion. When the deformation rate of the disaster-incubating zone meets the acceleration criterion and the transmission efficiency of the chain segment is lower than the preset threshold, all observation tasks of the chain segment are switched to the non-contact observation channel.

Images