A method for modeling a segmented mixed channel under a mine explosion working condition

Abstract

The application discloses a kind of segmented mixed channel modeling methods under mine explosion working condition, it is related to the technical field of underground environment channel propagation.First, the scene is modeled by obtaining mine roadway, explosion and obstacle data, and the scene is divided into three types of scene blocks, straight roadway, curved roadway and explosion area.Second, the main and auxiliary model mixed modeling is used for each scene block, the waveguide mode model is mainly used for straight roadway, and the ray tracing is auxiliary, the ray tracing is mainly used for curved roadway, and the propagation chart is auxiliary, the propagation chart is mainly used for explosion area, and the local ray tracing is auxiliary, and the delay-doppler domain preliminary propagation kernel is output.After physical consistent processing, the interface coupling operator is constructed, and the total propagation kernel is obtained by twisted convolution to realize the inter-segment propagation kernel splicing.Finally, the model parameters are optimized by combining the measured signal inversion.The application realizes the accurate matching of multiple propagation mechanisms, considers the modeling accuracy and calculation efficiency, guarantees the physical authenticity, and provides reliable channel support for underground explosion scene communication and detection.

Description

Technical Field

[0001] This invention relates to the field of underground environment channel propagation technology, specifically to a segmented hybrid channel modeling method under mine explosion conditions. Background Technology

[0002] Under explosive conditions, mine roadways are typically composed of alternating straight roadways, curved roadways, and localized areas affected by the explosion. The shape of the roadway cross-section, the electromagnetic parameters of the wall material, and the distribution of supports and equipment obstacles all affect electromagnetic propagation.

[0003] For electromagnetic propagation and channel modeling in underground roadways, existing technologies often employ straight roadway propagation models based on waveguide / modal theory, ray tracing models based on geometric optics / diffraction mechanisms, and statistical or empirical models for complex scattering environments. In practical engineering, there is also the practice of dividing the roadway into structural segments and modeling them separately, then combining the results of each segment in the time or frequency domain to obtain the end-to-end channel response. However, the above methods still have shortcomings when applied to mine explosion conditions: 1. A single mechanism model cannot simultaneously take into account the waveguide propagation in straight segments, the modal coupling / diffraction in curved segments, and the strong scattering and abrupt changes in medium parameters in the explosion-affected segments, thus limiting the applicability of the model.

[0004] 2. When the segmented results are directly spliced ​​in the time / frequency domain, there is often a lack of unified processing of physical constraints (such as energy conservation, reciprocity, phase continuity, modal coupling relationship, etc.) at the interface between segments, which can easily lead to interface discontinuities and error accumulation.

[0005] 3. Changes in geometric and material parameters caused by explosions, as well as environmental uncertainties such as obstacles / smoke, make it difficult to accurately set model parameters. Existing methods mostly rely on manual experience for parameter tuning and lack a mechanism for closed-loop calibration and inversion using measured data.

[0006] Therefore, there is an urgent need to develop a method that can achieve segmented hybrid modeling and physical consistency splicing between segments under any roadway block sequence conditions, and can combine measured data for parameter calibration and electromagnetic inversion. Summary of the Invention

[0007] The purpose of this invention is to propose a segmented hybrid channel modeling method under mine explosion conditions. The method uses three types of scene blocks as atomic units to construct a main model + auxiliary model correction chain, outputs uniformly in the delay-Doppler domain and uses twisting to achieve cross-block splicing, and finally achieves parameter calibration through electromagnetic inversion of measured propagation kernels.

[0008] The technical solution adopted in this invention is: a segmented hybrid channel modeling method under mine explosion conditions, including the following steps: Step S1, scene modeling: obtain the geometric dimensions of the mine roadway, the electrical parameters of the wall material, the explosion location and energy level, and the shape and location of significant obstacles to form the basic data of the mine explosion scene model.

[0009] Step S2, Scene Segmentation: Divide the scene under the explosion condition into a sequence of scene blocks of three types: straight tunnels, curved tunnels, and tunnels in the explosion area, and form a block sequence according to the actual spatial connection order. , where N is a positive integer representing the total number of scene blocks.

[0010] Step S3, Intra-segment primary and secondary correction modeling: For each scene block ( The modeling employs a hybrid approach combining a primary model and an auxiliary model. For straight roadway blocks, the waveguide mode propagation model is used as the primary model, with ray tracing used as an auxiliary model to generate corrections for the mode parameters. For curved roadway blocks, ray tracing is used as the primary model, with propagation maps used as an auxiliary model to provide prior angular domain and path search constraints. For explosion area blocks, propagation maps are used as the primary model, with local ray tracing used as an auxiliary model to calibrate the path cluster parameters. Each scene block... Output its corresponding delay-Doppler domain propagation kernel ,in For delay, This is the Doppler frequency shift.

[0011] Step S4, Physical consistency processing: Propagation kernel for each scene block Physical consistency constraints are applied to obtain a propagation kernel that satisfies the preset physical constraints. .

[0012] Step S5, Inter-segment Interface Coupling and Whole-Field Propagation Kernel Stitching: For each pair of adjacent scene blocks and Constructing interface coupling operators ,in In the time-delay-Doppler domain, warped convolution operations are used to propagate the kernel of all block scenes. Interface coupling operator By cascading according to spatial connection order, the total propagation kernel of the entire mine roadway is obtained. Its expression is: .

[0013] As a further improvement of the present invention, in step S3, the straight tunnel block is modeled with master-slave correction, specifically including the following steps: Step S301, master model waveguide mode propagation: select Construct an M-dimensional complex mode amplitude vector from several effective waveguide modes. Its position along the axial direction of the tunnel The evolution satisfies the following differential equation: in, Its elements Indicates the first Each waveguide mode is at position ,frequency and time The following complex amplitude; for The modal attenuation constant matrix characterizes the amplitude attenuation of each mode; for The modal phase constant matrix characterizes the phase change of each mode; for The modal coupling matrix characterizes the energy exchange between different modes; for The equivalent injection source vector of dimension represents the external stimulus (such as an antenna) at position. Energy injected into each mode; It is the imaginary unit.

[0014] Step S302, Auxiliary model ray tracing correction: Within the straight tunnel block, with axial position Centered on a length of Local analysis window Perform ray tracing within this window to obtain... Each sample of X-rays Corresponding to an incident direction angle and a statistical weight Based on these samples, construct a frequency The equivalent scattering angle spectrum below The expression is: in, For kernel function, Its bandwidth parameters, The normalization coefficient is... .

[0015] The parameters of the main model are calculated using a predefined mapping function. The correction amount includes the modal attenuation correction amount. Modal phase correction amount Modal coupling correction amount and equivalent source correction amount .

[0016] Step S303, Parameter Write-back and Propagation Kernel Generation: First, the correction values ​​calculated in step S302 are written back to the corresponding parameters of the main model: The differential equation in step S301 is solved again using the updated parameters to obtain the outlet section of the straight tunnel block. Modal amplitude vector at .

[0017] Then, a projection vector related to the characteristics of the receiving antenna is used. Calculate the equivalent port frequency response of this scene block: in, This represents the equivalent frequency response of the scene block under unit port excitation.

[0018] Finally, the two-dimensional Fourier transform is used to... Mapping to the time-delayed Doppler domain yields the initial intra-segment propagation kernel for this straight tunnel block. : .

[0019] As a further improvement of the present invention, in step S3, the curved tunnel block is modeled using a primary and secondary correction method, specifically including the following steps: Step S311, Construction of the secondary model propagation diagram: Based on the centerline geometry, radius of curvature, and wall parameters of the curved tunnel, a propagation diagram model is constructed. , where vertex set Represents key scattering points, edge set This represents the visible and reachable propagation paths between vertices.

[0020] Clustering reachable paths using a propagation graph yields a series of candidate path clusters. and for each cluster Output the following prior information: feasible emission angle cone Maximum number of reflections allowed And the statistical priors of clusters, including average latency. Mean Doppler And the cluster delay and Doppler dispersion , .

[0021] Step S312, the main model is constrained ray tracing: using the prior information output in step S304, the main model is guided and constrained.

[0022] Step S313, Propagation Nuclear Convergence: For all valid ray paths that pass through the constraint and reach the receiving point. Calculate the delay for each path. Doppler and complex gain .

[0023] In the time-delay-Doppler two-dimensional plane, the energy distribution of each path is aggregated by a windowing function to generate the initial propagation kernel of the curved tunnel block. : ;in, and This is a normalized window function.

[0024] As a further improvement of the present invention, step S3 involves performing master-slave correction modeling on the explosion area block, specifically including the following steps: Step S321, master model propagation map construction and path weight calculation: constructing a propagation map in the explosion-affected area. , where the vertex Characterizing discrete scatterers, wall patches, diffraction edges, and entry / exit interface nodes within the explosion zone. Characterizes the visibility and reachability of nodes.

[0025] For each candidate path on the propagation graph Calculate a parameter affected by an explosion disturbance Path weights affected ,in The attenuation coefficient is given by the path weight formula: ;in, The arc length along the path; The equivalent attenuation coefficient characterizes the energy loss caused by absorption and scattering due to explosion disturbance parameters. The impact; This is the product of the amplitude and phase coefficients of reflection, transmission, and diffraction along this path.

[0026] Step S322, Clustering and Key Cluster Selection: Cluster candidate paths according to key geometric event sequences and time-delay-Doppler proximity relationships to obtain a set of path clusters. .

[0027] Calculate each path cluster Total weight , and according to Statistical results for each cluster As the cluster prior, the one with the largest total weight is selected. These clusters constitute a set of key clusters to be calibrated. .

[0028] Step S323, Local Calculation and Parameter Calibration of the Auxiliary Model: For the set of key clusters For each cluster, high-precision local ray tracing is performed within its corresponding physical space region to obtain the calculated complex amplitude for that cluster. .

[0029] The main model is based on macroscopic equivalent parameters. Predicted complex amplitude of the same cluster With actuarial value By comparing the two, an optimization problem is constructed with the objective of minimizing the difference between them, and the solution is obtained for the parameters of the main model. and Calibration amount and .

[0030] Step S324, Parameter Write-back and Propagation Kernel Generation: The calibrated parameters... and The path weight calculation function of the main model is written back to recalculate the complex amplitude of each cluster, and finally aggregated to generate the initial propagation kernel of the exploded region block. : in, and This is an interpolation and windowing function that normalizes per unit area.

[0031] As a further improvement of the present invention, in step S4, the physical unification process refers to mapping the initial propagation kernel of each scene block through a mapping process. Converted into a propagation kernel that satisfies preset physical constraints .

[0032] As a further improvement of the present invention, in step S5, the step refers to each pair of adjacent scene blocks. and Constructed interface coupling operator It is a linear operator composed of the following three cascaded operators, specifically including: step S51, basis alignment transformation operator. Used to move the previous block The output signal representation is transformed into the next block. The required representation of the input signal; this operator is implemented by calculating the inner product matrix of the two basis functions on the interface section.

[0033] Step S52, Phase Compensation Operator Used to compensate for the additional phase delay caused by changes in wavefront curvature or reference plane movement when a signal passes through the physical interface between two scene blocks. This phase delay is a time delay. A quadratic function.

[0034] Step S53, Local Interaction Operator : Used to characterize the local physical processes that occur at the interface of electromagnetic waves, and its coefficients are determined by Fresnel's formula or uniform geometric diffraction theory coefficients.

[0035] As a further improvement of the present invention, in step S5, the twisted convolution operation is a generalized two-dimensional convolution defined in the delay-Doppler domain; for two functions and Its twisted convolution operation Defined by the following integral: Among them, the integral variable and Representing time delay and Doppler shift respectively, the exponent term It characterizes the coupling phase between Doppler frequency shift and relative time delay.

[0036] As a further improvement to the present invention, it also includes step S6, model parameter calibration and inversion optimization: based on measured detection signals. With received signal Estimate the measured propagation nucleus ;

[0037] Construct to minimize the total propagation kernel With measured propagation nucleus The objective function is the difference between them. in It is a non-negative weight matrix. For Hadamard's product.

[0038] The parameters in the segmented hybrid channel model are adjusted by optimizing the algorithm to minimize the objective function, thereby achieving the calibration and inversion of the model parameters.

[0039] As a further improvement of the present invention, in step S312, the guidance refers to the guidance based on cluster statistical priors. Construct an importance sampling function to concentrate the ray emission direction more on the high-probability corner cone. The constraint refers to forcibly limiting the number of reflections for each path to no more than a certain number during the tracing process. And prune paths whose path loss exceeds the threshold.

[0040] As a further improvement of the present invention, the mapping process of physical uniformization in step S4 enforces one or more of the following constraints: Step S41, Support domain constraint: the propagation kernel is within a preset time delay range. and Doppler range All values ​​other than zero are set to zero.

[0041] Step S42, Energy Constraint: Scaling the amplitude of the propagation kernel to reduce its total energy. Not exceeding the maximum permissible energy derived from the transmit power and system losses. .

[0042] Step S43, Reciprocity Constraint: If the channel model is a reciprocal system, then the propagation kernel must satisfy... ,in This indicates that according to the reciprocity principle, from The kernel obtained by transformation.

[0043] Step S44, Phase Smoothing Constraint: Phase of the propagation kernel Smoothing filters are applied to suppress abrupt, non-physical phase transitions.

[0044] Compared with the prior art, the present invention has the following technical advantages: 1. The present invention divides the roadway scene under the mine explosion condition into three types of atomic scene blocks: straight roadway, curved roadway, and explosion area roadway. For each type of scene block, a hybrid modeling architecture of main model + auxiliary model is customized to achieve accurate matching of different propagation mechanisms: straight roadway uses waveguide mode propagation model to capture the main energy propagation characteristics, and combines local ray tracing to correct scattering and mode coupling disturbances; curved roadway uses restricted ray tracing to explicitly characterize multipath reflection / diffraction, and provides prior constraints through propagation graph to improve the hit rate of effective paths; explosion area uses propagation graph to efficiently enumerate large-scale path clusters, and combines local high-precision ray tracing to accurately calculate the energy-dominant clusters. This not only covers the complex characteristics of strong scattering and parameter mutation in the explosion area, but also avoids the adaptation deviation of a single model to different scenes, and achieves accurate characterization of electromagnetic propagation characteristics under all working conditions.

[0045] 2. On the one hand, this invention employs a bidirectional correction mechanism between the primary and secondary models. The secondary model provides parameter corrections, prior constraints, or high-precision calculated values ​​to the primary model, compensating for the inherent errors of a single model. This includes modal parameter correction for straight lanes and attenuation / disturbance parameter calibration for explosion regions, making the model output more consistent with actual electromagnetic propagation laws. On the other hand, through physical uniformity processing, four constraints—support domain, energy, reciprocity, and phase smoothing—are applied to the propagation kernel of each scene block, forcing the model to satisfy the basic physical laws of electromagnetic propagation and suppressing the generation of non-physical errors from the source. Simultaneously, the interface coupling operator constructed between adjacent scene blocks achieves accurate representation of basis alignment, phase compensation, and local physical processes. Combined with the twisted convolution operation in the delay-Doppler domain, the splicing is completed, solving the error accumulation problems caused by energy non-conservation, phase discontinuity, and lack of modal coupling relationships in traditional segmented splicing, significantly improving the overall accuracy of the entire channel modeling.

[0046] 3. This invention effectively controls computational load through localized computation, constrained tracing, and lightweight enumeration: For straight tunnels, ray tracing is performed only within a local window, avoiding the high computational cost of global ray tracing; for curved tunnels, the ray tracing range is constrained by the prior angular domain cone and upper bound of the reflection order in the propagation graph, and an importance sampling function is constructed to improve the hit rate of effective paths and reduce the computation of invalid rays; in the explosion region, path clusters are quickly enumerated using the propagation graph, and Top-K energy-dominant clusters are selected, with high-precision local ray tracing performed only on key clusters, avoiding the exponential complexity of high-precision computation across the entire region. While ensuring modeling accuracy, this invention significantly reduces the computational load under complex explosion conditions, enabling the model to run efficiently on conventional engineering computing platforms and improving engineering feasibility.

[0047] 4. This invention includes a model parameter calibration and inversion optimization stage. Based on the measured detection and received signals, the measured propagation kernel is estimated. An optimization function is constructed with the goal of minimizing the difference between the model propagation kernel and the measured propagation kernel. The model parameters and interface coupling operator parameters are dynamically adjusted through the optimization algorithm, forming a closed-loop mechanism of modeling-measurement-inversion-calibration. This stage overcomes the limitations of traditional methods that rely on manual experience for parameter tuning. It can adaptively compensate for changes in tunnel geometry and material parameters caused by explosions, as well as environmental uncertainties caused by the distribution of smoke and obstacles. This allows the model to dynamically adapt to real-time changes in explosion conditions, ensuring the modeling effectiveness under different explosion energy levels and different disturbance degrees.

[0048] 5. This invention maps the modeling results of all scene blocks to a time-delay-Doppler domain output propagation kernel, achieving isomorphism in the modeling results of different scene blocks and laying the foundation for cross-block stitching. Simultaneously, it employs a standardized interface coupling operator construction method and a twisted convolution stitching algorithm, supporting the stitching of any number and combination of scene block sequences, adapting to the complex topological structures of different mine roadways (such as alternating straight and curved sections, and multiple explosion zone distributions). The modular and standardized design of the model allows each functional module to be independently optimized and flexibly expanded. Subsequently, it can be adapted to different mine geological characteristics and blasting conditions by simply adjusting the model parameters or constraints of the corresponding scene blocks, without needing to reconstruct the overall modeling framework, significantly improving the model's versatility and engineering application value.

[0049] 6. The whole-field propagation kernel output by this invention can accurately characterize key channel characteristics of electromagnetic signals under mine explosion conditions, such as time delay, Doppler frequency shift, and complex gain. Furthermore, it strictly matches the input-output relationship of the actual received signal, providing accurate and reliable channel model support for the design of wireless communication systems, detection signal processing, and the development of positioning and navigation algorithms in mine explosion scenarios. This model allows for the simulation of channel propagation characteristics under different explosion conditions in advance, guiding the deployment of underground communication equipment, the selection of modulation and coding methods, and the optimization of detection system parameters. This improves the stability and effectiveness of wireless communication and detection in mine explosion emergency rescue and safety production monitoring, possessing significant engineering application value. Attached Figure Description

[0050] Figure 1 This is a flowchart illustrating the overall process of the segmented hybrid channel modeling method for mine explosion conditions according to the present invention.

[0051] Figure 2 This is a schematic diagram of waveguide parameter rewriting in the ray tracing method for straight tunnels.

[0052] Figure 3 A diagram showing the propagation pattern in a curved tunnel → a schematic diagram of the feasible angular region and restricted tracing for the ray tracing method. Figure 4This is a schematic diagram of sensitivity calibration and parameter write-back for the ray tracing method in the explosion area → propagation map.

[0053] Figure 5 This is a schematic diagram of the interface coupling operator structure and the twisting and splicing process. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention are within the scope of protection of the present invention.

[0055] The segmented hybrid channel modeling method for mine explosion conditions in this invention ultimately aims to solve for the whole-field propagation kernel S, which characterizes the electromagnetic propagation properties of a mine explosion scenario. tot (τ,ν), the propagation kernel and the actual detected and received signals satisfy the following overall input-output relationship: the received signal y(t) and the entire propagation kernel satisfy: in The detected signal is known.

[0056] The interchangeability between impulse response and propagation kernel follows a fixed convention: .

[0057] The fixed convention for frequency domain—delay / Doppler domain is: Let Let be the equivalent frequency response at point t in slow time, then: .

[0058] Specifically, such as Figure 1 As shown, this method includes the following steps: Step S1 Scene modeling: Obtain the center line, cross-sectional dimensions, radius of curvature, and bifurcation position of the tunnel; mark the location and energy level of the explosion area, and record the dielectric constant / conductivity and roughness of the wall material, as well as the parameters of obstacles such as supports and equipment, and the ventilation environment.

[0059] Step S2 divides the scene into three types of scene blocks: straight alleyway A, curved alleyway B, and explosion zone C, forming a block sequence according to the actual spatial order. Each block establishes a priori (Geometric, material properties, roughness spectrum, suspected reflection / diffraction set, etc.).

[0060] Step S3: Modeling of primary and secondary corrections within the segment.

[0061] For each scene block ( The modeling process employs a hybrid approach combining the primary and secondary models. Specifically, for straight roadway blocks, the waveguide mode propagation model is used as the primary model, with ray tracing used as the secondary model to generate corrections for the mode parameters. For curved roadway blocks, ray tracing is used as the primary model, with propagation maps used as the secondary model to provide prior angular domain and path search constraints. For explosion zone blocks, propagation maps are used as the primary model, with local ray tracing used as the secondary model to calibrate the path cluster parameters. Each block outputs to a time-delay-Doppler domain propagation kernel. The main / auxiliary model functions by having the auxiliary model generate correction values ​​or priors / constraints, which are then written back to the main model's generation mechanism.

[0062] Figure 2 This is a schematic diagram of the A-ray tracing method for straight tunnels, followed by waveguide parameter back-writing. The cross-section and boundary conditions of a straight tunnel section change slowly within the section, and the principal energy propagates along a finite number of effective waveguide modes. Using a waveguide mode propagation model allows for obtaining the principal propagation term with relatively low computational cost. However, wall roughness, support components, and equipment obstacles induce scattering and mode coupling. Directly using global ray tracing would incur high computational costs. Therefore, ray tracing is performed only in a local window to estimate disturbances and back-write the principal model parameters, thus achieving a balance between computational cost and accuracy. Specifically, the process of establishing the principal and auxiliary models is as follows: 1. Principal Model (Mode Propagation Equations): Select... There are 1 effective mode, and the mode magnitude vector is... Satisfy the following differential equations: in, Its elements Indicates the first Each waveguide mode is at position ,frequency and time The following complex amplitude; for The modal attenuation constant matrix characterizes the amplitude attenuation of each mode; for The modal phase constant matrix characterizes the phase change of each mode; for The modal coupling matrix characterizes the energy exchange between different modes; for The equivalent injection source vector of dimension represents the external stimulus (such as an antenna) at position. Energy injected into each mode; It is the imaginary unit.

[0063] 2. Auxiliary Model (Local Ray Tracing Correction): Within a local window, the correction is applied to... Ray tracing is performed on the ray sample based on the incident angle sample. and their weights Formation of scattering angle spectrum Its expression is: in, For kernel function, Its bandwidth parameters, The normalization coefficient is... .

[0064] By weight function calculate , ; by scattering / reflection nuclei calculate Equivalent secondary scattered electric field obtained by local superposition Obtained through modal projection Modal attenuation correction amount Modal phase correction amount Modal coupling correction amount and equivalent source correction amount The expression is as follows: in It is a non-negative weighting function used to characterize the contribution of angle / frequency samples to modal attenuation and phase perturbation; This is a scattering / reflection mapping or kernel function from the incident angle domain to the exit angle domain; As an auxiliary model in the cross section The equivalent secondary scattered electric field obtained by superposition at the location; For the first Transverse field distribution of each mode; and The electrical parameters, roughness, and modal field distribution of the wall material can be obtained through analytical derivation, offline electromagnetic simulation, table lookup, or calibration. It is obtained by superposition of ray fields.

[0065] 3. Write-back and propagation kernel generation: [This will...] Write back to the master model and re-solve to obtain the receiver cross section. Modal amplitude at The equivalent frequency response of the block is constructed by port projection: in The projection vector of the receiving port can be obtained through analytical derivation. This represents the equivalent frequency response of the scene block under unit port excitation.

[0066] Further according to the fixed agreement Mapping to the time-delay-Doppler domain yields the intra-segment propagation kernel. And it is consistent with the overall input-output relationship.

[0067] Figure 3This diagram illustrates the propagation map of a curved tunnel (B) and the feasible angular domain and constrained tracing using ray tracing. Curved sections typically exhibit line-of-sight obstruction and multiple reflection / diffraction sequences caused by curvature. The principal energy is no longer stably confined to a few waveguide modes. Therefore, ray tracing can explicitly characterize the geometry and reflection / diffraction physics of multipath propagation. However, unconstrained tracing across the entire angular domain and high reflection orders leads to a rapid increase in the number of rays and a low effective hit rate, resulting in high computational cost and high estimation variance. To address this, a propagation map is used as an auxiliary model, outputting the feasible angular domain, reflection order, and path family statistical priors. These are used to construct the importance sampling distribution and pruning criteria, thereby achieving a balance between computational cost and accuracy.

[0068] 1. Auxiliary Model (Propagation Graph and Cluster Prior): Constructing the Propagation Graph Vertices represent key geometric event nodes (entrance / exit sections, candidate reflective surfaces, diffraction edges, or their discrete nodes) within the curved segment, while edges represent visibility and reachability connections.

[0069] The propagation graph clusters reachable paths according to the sequence of key nodes to obtain candidate path clusters. And output for each cluster: feasible emission angular cone Upper bound of the reflection order With family statistical priors .

[0070] in The center of the cluster represents the equivalent delay of the path. Discreteness of candidate path delay within cluster For the corresponding center and dispersion of the Doppler; in the implementation of the static geometric approximation, it can be taken as and will Set to a preset small value to reflect slow changes.

[0071] 2. Master Model (Importance Sampling of Constrained Ray Tracing): Corner Cone in the Propagation Map Output Inner launch angle Importance sampling, with the sampling distribution as follows: ;in For the first The path cluster weight coefficients satisfy the following conditions: This is the normalization constant; This is an indicator function that returns 1 if the condition is true and 0 otherwise. These represent the emission angle under the current geometric and wall parameters, respectively. The equivalent delay and Doppler of the generated candidate paths; the scale parameter is determined by the cluster dispersion: ;in To prevent division by zero by a preset small amount; and to apply pruning during the tracking process: limiting the reflection order to no more than And when path cost Exceeding the preset threshold The extension will terminate at that time.

[0072] In specific implementation, the pruning threshold and the upper bound of the order The pruning process is set according to the principle of monotonic path cost reduction and support for early stopping, in order to maximize pruning efficiency.

[0073] This allows sampling to achieve higher probabilities in angular domains with smaller delay / Doppler bias, thereby improving effective path hit rate and reducing estimation variance. (Pruning, Aggregation, and Output) During the tracking process, the reflection order is limited to no more than [a certain value]. And set the path cost threshold. Pruning is carried out; among them We take an additive cost that monotonically does not decrease as the path expands, in order to support early online stopping.

[0074] Calculation of the effective path captured by the receiving end With complex gain The propagation core within the curved roadway section was obtained by time-delay-Doppler mesh aggregation. Simultaneously record the radius of curvature. With the angle of incidence The relevant curvature phase parameters are used for the secondary phase compensation term in the subsequent interface coupling operator.

[0075] Figure 4 This diagram illustrates the sensitivity calibration and parameter write-back of the C-ray tracing method in the explosion region, specifically the propagation map. The explosion-affected area often exhibits characteristics such as dense obstacles, a sharp increase in the number of scatterers, abrupt changes in wall / medium parameters, and enhanced anisotropy. If high-precision ray tracing is performed directly over the entire region, the number of candidate paths increases exponentially, and many paths contribute very little to the reception, leading to excessive computation and increased estimation variance. Therefore, a primary / secondary calibration chain is employed: the propagation map is used to enumerate and sort large-scale candidate path clusters in a compact manner (ensuring coverage and scalability), while local ray tracing performs high-precision calculations only on the energy-dominant Top-K clusters (ensuring accuracy and controlling computational load).

[0076] 1. Main Model (Path Weights and Cluster Priors in the Propagation Graph): For candidate paths on the propagation graph... Calculate path weights: ;in, The arc length along the path; The equivalent attenuation coefficient (absorption and scattering loss) characterizes the energy loss caused by absorption and scattering due to explosion disturbances. The impact; This is the product of the reflection / transmission / diffraction amplitude and phase coefficients along this path. Form a set of path clusters and with Statistical results for each cluster As cluster priors, cluster weights are also calculated. And based on this, Top-K clusters are selected. Entering local precision calculation. The Top-K cluster size is used to balance computational load and accuracy, and can be adjusted according to environmental complexity and error level.

[0077] 2. Auxiliary Model (Local Ray Tracing): For each Perform ray tracing actuarial calculations within its local area and output cluster actuarial energy. With actuarial phase And define the actuarial complex amplitude. .

[0078] The main model in parameters The corresponding cluster complexity amplitude is given below. Solving by least squares , ;in This is the increment correction amount for the attenuation parameter. This is a set of disturbance parameters for the explosion zone. Its incremental correction amount is the regularization coefficient.

[0079] 3. Parameter write-back and propagation kernel generation: derived from the accompanying sensitivity renew And write the updated parameters back to and In the calculations, the nuclear propagation within the explosion zone is formed. : in, , This is an interpolation / windowing function that normalizes per unit area.

[0080] Step S4: Physical uniformity and projection within the segment.

[0081] Let the initial delay-Doppler propagation kernel of the main / auxiliary correction chain output within the segment be denoted as... To ensure physical feasibility and suppress error accumulation during multi-segment cascading, the following measures are taken: Perform physical constraint uniformity and projection processing, and define the constraint set: ;in To support domain constraints Energy non-increase / passive constraint ( For reciprocity constraints; For intra-segment phase continuity constraints (adjacent sampling points satisfy after phase unwrapping) ).

[0082] Projection operator It can be achieved through one or more truncation, normalization, symmetry, phase smoothing, etc., or by solving constrained optimization problems.

[0083] Received It serves as the intra-segment output propagation kernel, used for subsequent cross-block interface coupling operators and twisted roll splicing.

[0084] Step S5: Inter-segment interface coupling and whole-field propagation kernel splicing.

[0085] Adjacent scene blocks may differ in their modeling mechanisms and equivalent bases for output kernels within a segment. For example, the output of a straight tunnel may be dominated by waveguide modes, while the output of a curved tunnel may be dominated by ray clusters. Furthermore, geometric abrupt changes, curvature variations, and local scattering / diffraction exist at the interface, making direct interaction with... and Twisting and splicing introduces errors such as energy non-conservation, phase discontinuity, and reciprocity violation.

[0086] Therefore, this invention is based on the interface section of adjacent blocks. Upper construction interface coupling operator Used to put blocks The outgoing representation is mapped to a block. After the incident representation is obtained, twisting and splicing are then performed, such as Figure 5 As shown, the details are as follows.

[0087] 1. Construct the interface coupling operator: Among them, the basis alignment transformation operator Basis alignment is achieved through the power inner product overlap matrix on the interface cross section.

[0088] The chirp coefficient compensates for group delay dispersion caused by curvature / coordinate transformation through a quadratic phase term. and Proportional and Positive correlation; The equivalent complex amplitude coefficients of reflection / transmission / diffraction at the interface are calculated based on Fresnel / UTD and a linear amplitude-phase mapping is formed.

[0089] constitute Physical consistency can then be further corrected through scaling, symmetry, regularization, or projection to limit energy amplification and / or maintain reciprocal symmetry.

[0090] according to The cascaded structure allows for numerical regularization or amplitude-phase correction when needed to improve cascade stability.

[0091] In this step, the twisted convolution is defined as: By splicing the block sequence together, the total propagation kernel of the entire mine roadway is obtained.

[0092] The numerical implementation employs phase preprocessing—2D fast convolution—phase backcompensation—energy normalization; the mesh satisfies... .

[0093] The aforementioned twisted convolution formula mathematically defines a precise cascaded operation in a continuous domain. To efficiently implement this in a computer and avoid its extremely high computational complexity, the numerical implementation in this step employs a process of phase preprocessing – two-dimensional fast convolution – phase backcompensation – energy normalization. Simultaneously, to discretize this continuous integral, the mesh satisfies… .

[0094] Specifically, let the upper bound parameter of the support domain within the segment be... (For projection constraint sets), the number of discrete grid points can be taken as follows: 1.

[0095] Step S6: Estimation and Inversion of Measured Propagation Kernel Closure: Measured Propagation Kernel Depend on get.

[0096] Inversion objective: ;in It is a non-negative weight matrix, which can be set by information such as the support domain, measurement signal-to-noise ratio, or region of interest; For Hadamard product; update the block parameters and interface coupling operator parameters under parameter range constraints, so that Decrease or converge; the gradient can be obtained by the adjoint operator of twisted convolution or numerical difference, and the update can be achieved by projected gradient or alternating optimization.

[0097] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes that can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention are all within the protection scope of the claims of the present invention.

Claims

1. A segmented hybrid channel modeling method under mine explosion conditions, characterized in that, Includes the following steps: Step S1, Scene Modeling: Obtain the geometric dimensions of the mine roadway, the electrical parameters of the wall material, the explosion location and energy level, and the shape and location of significant obstacles to form the basic data of the mine explosion scene model; Step S2, Scene Segmentation: Divide the scene under the explosion condition into a sequence of scene blocks of three types: straight tunnels, curved tunnels, and tunnels in the explosion area, and form a block sequence according to the actual spatial connection order. , where N is a positive integer representing the total number of scene blocks; Step S3, Intra-segment primary and secondary correction modeling: For each scene block ( The modeling employs a hybrid approach combining a primary model and an auxiliary model. For straight roadway blocks, the waveguide mode propagation model is used as the primary model, with ray tracing used as an auxiliary model to generate corrections for the mode parameters. For curved roadway blocks, ray tracing is used as the primary model, with propagation maps used as an auxiliary model to provide prior angular domain and path search constraints. For explosion area blocks, propagation maps are used as the primary model, with local ray tracing used as an auxiliary model to calibrate the path cluster parameters. Each scene block... Output its corresponding delay-Doppler domain propagation kernel ,in For delay, For Doppler frequency shift; Step S4, Physical consistency processing: Propagation kernel for each scene block Physical consistency constraints are applied to obtain a propagation kernel that satisfies the preset physical constraints. ; Step S5, Inter-segment Interface Coupling and Whole-Field Propagation Kernel Stitching: For each pair of adjacent scene blocks and Constructing interface coupling operators ,in In the time-delay-Doppler domain, warped convolution operations are used to propagate the kernel of all block scenes. Interface coupling operator By cascading according to spatial connection order, the total propagation kernel of the entire mine roadway is obtained. Its expression is: ； The above refers to each pair of adjacent scene blocks and Constructed interface coupling operator is a linear operator consisting of the following three cascaded operators, specifically: Step S51, Basis Alignment Transformation Operator Used to move the previous block The output signal representation is transformed into the next block. The required representation of the input signal; this operator is implemented by calculating the inner product matrix of the two basis functions on the interface section; Step S52, Phase Compensation Operator Used to compensate for the additional phase delay caused by changes in wavefront curvature or reference plane movement when a signal passes through the physical interface between two scene blocks. This phase delay is a time delay. A quadratic function; Step S53, Local Interaction Operator : Used to characterize the local physical processes that occur at the interface of electromagnetic waves, and its coefficients are determined by Fresnel's formula or uniform geometric diffraction theory coefficients.

2. The segmented hybrid channel modeling method under mine explosion conditions according to claim 1, characterized in that, In step S3, the main and auxiliary correction modeling of the straight tunnel block is performed, which specifically includes the following steps: Step S301, Main Model Waveguide Mode Propagation: Select Construct an M-dimensional complex mode amplitude vector from several effective waveguide modes. Its position along the axial direction of the tunnel The evolution satisfies the following differential equation: ； in, Its elements Indicates the first Each waveguide mode is at position ,frequency and time The following complex amplitude; for The modal attenuation constant matrix characterizes the amplitude attenuation of each mode; for The modal phase constant matrix characterizes the phase change of each mode; for The modal coupling matrix represents the energy exchange between different modes; for The equivalent injection source vector of dimension represents the external stimulus (such as an antenna) at position. Energy injected into each mode; The imaginary unit; Step S302, Auxiliary model ray tracing correction: Within the straight tunnel block, with axial position Centered on a length of Local analysis window Perform ray tracing within this window to obtain... Each sample of X-rays Corresponding to an incident direction angle and a statistical weight Based on these samples, construct a frequency The equivalent scattering angle spectrum below The expression is: ； in, For kernel function, Its bandwidth parameters, The normalization coefficient is... ;Calculate the parameters of the main model using a predefined mapping function. The correction amount includes the modal attenuation correction amount. Modal phase correction amount Modal coupling correction amount and equivalent source correction amount ; Step S303, Parameter Write-back and Propagation Kernel Generation: First, the correction values ​​calculated in step S302 are written back to the corresponding parameters of the main model: ； The differential equation in step S301 is solved again using the updated parameters to obtain the outlet section of the straight tunnel block. Modal amplitude vector at ; Then, a projection vector related to the characteristics of the receiving antenna is used. Calculate the equivalent port frequency response of this scene block: ； in, This represents the equivalent frequency response of the scene block under unit port excitation; Finally, the two-dimensional Fourier transform is used to... Mapping to the time-delayed Doppler domain yields the initial intra-segment propagation kernel for this straight tunnel block. : 。 3. The segmented hybrid channel modeling method under mine explosion conditions according to claim 1, characterized in that, In step S3, the curved roadway block is modeled using a primary and secondary correction method, which specifically includes the following steps: Step S311, Construction of the auxiliary model propagation diagram: Based on the centerline geometry, radius of curvature, and wall parameters of the curved tunnel, construct the propagation diagram model. , where vertex set Represents key scattering points, edge set Represents the visible and reachable propagation paths between vertices; Clustering reachable paths using a propagation graph yields a series of candidate path clusters. and for each cluster Output the following prior information: feasible emission angle cone Maximum number of reflections allowed And the statistical priors of clusters, including average latency. Mean Doppler And the cluster delay and Doppler dispersion , ; Step S312, the main model is constrained ray tracing: using the prior information output in step S304, the main model is guided and constrained; Step S313, Propagation Nuclear Convergence: For all valid ray paths that pass through the constraint and reach the receiving point. Calculate the delay for each path. Doppler and complex gain ; On the time-delay-Doppler two-dimensional plane, the energy distribution of each path is aggregated by applying a windowing function to generate the initial propagation kernel of the curved tunnel block. : ； in, and This is a normalized window function.

4. The segmented hybrid channel modeling method under mine explosion conditions according to claim 1, characterized in that, In step S3, the primary and secondary correction modeling of the explosion area block is performed, which specifically includes the following steps: Step S321: Construction of the main model propagation graph and calculation of path weights: Construct a propagation map within the explosion-affected area. , where the vertex Characterizing discrete scatterers, wall patches, diffraction edges, and entry / exit interface nodes within the explosion zone. Characterizes the visibility and reachability of nodes; For each candidate path on the propagation graph Calculate a parameter affected by an explosion disturbance Path weights affected ,in The attenuation coefficient is given by the path weight formula: ； in, The arc length along the path; The equivalent attenuation coefficient characterizes the energy loss caused by absorption and scattering due to explosion disturbance parameters. The impact; This is the product of the amplitude and phase coefficients of reflection, transmission, and diffraction along this path; Step S322, Clustering and Key Cluster Selection: Cluster candidate paths according to key geometric event sequences and time-delay-Doppler proximity relationships to obtain a set of path clusters. ; Calculate each path cluster Total weight , and according to Statistical results for each cluster As the cluster prior, the one with the largest total weight is selected. These clusters constitute a set of key clusters to be calibrated. ; Step S323, Local Calculation and Parameter Calibration of the Auxiliary Model: For the set of key clusters For each cluster, high-precision local ray tracing is performed within its corresponding physical space region to obtain the calculated complex amplitude for that cluster. ; The main model is based on macroscopic equivalent parameters. Predicted complex amplitude of the same cluster With actuarial value By comparing the two, an optimization problem is constructed with the objective of minimizing the difference between them, and the solution is obtained for the parameters of the main model. and Calibration amount and ; Step S324, Parameter write-back and propagation kernel generation: calibrated parameters and The path weight calculation function of the main model is written back to recalculate the complex amplitude of each cluster, and finally aggregated to generate the initial propagation kernel of the exploded region block. : ； in, and This is an interpolation and windowing function that normalizes per unit area.

5. The segmented hybrid channel modeling method under mine explosion conditions according to claim 1, characterized in that, In step S4, the physical unification process refers to mapping the initial propagation kernel of each scene block through a mapping process. Converted into a propagation kernel that satisfies preset physical constraints .

6. The segmented hybrid channel modeling method under mine explosion conditions according to claim 1, characterized in that, In step S5, the twisted convolution operation is a generalized two-dimensional convolution defined in the delay-Doppler domain; for two functions and Its twisted convolution operation Defined by the following integral: ； Among them, the integral variable and Representing time delay and Doppler shift respectively, the exponent term It characterizes the coupling phase between Doppler frequency shift and relative time delay.

7. The segmented hybrid channel modeling method under mine explosion conditions according to claim 1, characterized in that, It also includes step S6, model parameter calibration and inversion optimization: Based on measured detection signals With received signal Estimate the measured propagation nucleus ; Construct to minimize the total propagation kernel With measured propagation nucleus The objective function is the difference between them. in It is a non-negative weight matrix. For Hadamard product; The parameters in the segmented hybrid channel model are adjusted by optimizing the algorithm to minimize the objective function, thereby achieving the calibration and inversion of the model parameters.

8. The segmented hybrid channel modeling method under mine explosion conditions according to claim 3, characterized in that, In step S312, the guidance refers to based on cluster statistical priors. Construct an importance sampling function to concentrate the ray emission direction more on the high-probability corner cone. The constraint refers to forcibly limiting the number of reflections for each path to no more than a certain number during the tracing process. And prune paths whose path loss exceeds the threshold.

9. The segmented hybrid channel modeling method under mine explosion conditions according to claim 5, characterized in that, The mapping process in step S4, which involves physical uniformity, enforces one or more of the following constraints: Step S41, Support Domain Constraint: Ensure the propagation kernel is within a preset time delay range. and Doppler range All values ​​except zero are set to zero; Step S42, Energy Constraint: Scaling the amplitude of the propagation kernel to reduce its total energy. Not exceeding the maximum permissible energy derived from the transmit power and system losses. ; Step S43, Reciprocity Constraint: If the channel model is a reciprocal system, then the propagation kernel must satisfy... ,in This indicates that according to the reciprocity principle, from The kernel obtained by transformation; Step S44, Phase Smoothing Constraint: Phase of the propagation kernel Smoothing filters are applied to suppress abrupt, non-physical phase transitions.

Images