Complex environment railway intelligent line selection gravitational field dynamic modeling and adaptive guiding optimization method

Abstract

The invention relates to the technical field of railway engineering line selection design, and discloses a complex environment railway intelligent line selection gravitational field dynamic modeling and self-adaptive guiding optimization method. The method comprises the following steps: constructing a multi-criterion index system, establishing an adaptation degree space field hierarchical model, and identifying high and low adaptation environment patches by using a DBSCAN density clustering algorithm; constructing an intelligent line selection environment static field model; a kinetic equation of gravitational propagation is established and solved, and intelligent line selection environment dynamic field modeling is achieved; and a Transform neural network model is introduced, gravitational field feature situation awareness is completed, a dynamic adaptive alignment method is put forward, and an optimal line scheme for railway line selection is output. According to the method, by constructing a line selection environment adaptation degree space field and converting the line selection environment adaptation degree space field into a gravitational field and combining dynamic evolution mechanism analysis and a self-adaptive oriented intelligent search strategy, efficient optimization design of a railway line in a dense constraint environment is achieved, the strain capacity of a line determination decision is improved, and the line selection efficiency is improved. And the dynamic optimization requirement of engineering line selection in a complex environment can be met.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of railway engineering route selection design, and particularly relates to a complex environment railway intelligent route selection gravity field dynamics modeling and adaptive guidance optimization method. BACKGROUND

[0002] In the field of railway construction, the complexity of the construction environment has become a key factor restricting the efficiency and quality of the project. This complexity mainly reflects the interweaving of multi-dimensional environmental factors: on the one hand, the natural geographical conditions show significant differences, including complex topography (such as mountains, valleys, marshes), variable geological structure (such as faults, karst caves, loose rock layers), and climate and hydrology influence (such as heavy rain, permafrost, fluctuation of underground water level); on the other hand, the interference of human and artificial environment is also prominent, involving sensitive areas such as existing building groups, traffic networks, underground pipeline dense areas, and ecological protection areas, resulting in multiple constraints of safety, economy and environmental coordination that need to be considered in engineering construction.

[0003] Based on the above-mentioned complex environmental characteristics, railway engineering route selection faces serious dynamic adaptation problems. Traditional route selection schemes often rely on static survey data and experience judgment, and are difficult to respond to dynamic changes in environmental parameters in real time. For example, in areas with unstable geological conditions, the stress distribution of rock-soil mass may be dynamically adjusted with the construction process, and if the route selection scheme fails to adapt to this change in time, it may easily cause safety risks such as landslides and subsidence; at the same time, when the engineering line passes through an ecologically sensitive area, the dynamic factors such as the migration of animal and plant activity range and the fluctuation of hydrological rhythm caused by seasonal changes also put higher requirements on the ecological adaptability of route selection. In addition, under the background of urban renewal, the demand for upgrading and upgrading of existing infrastructure (such as underground pipeline expansion and road widening) may cause sudden changes in the surrounding environmental load, further exacerbating the adaptation contradiction between the route selection scheme and the dynamic environment, and the traditional static planning mode is difficult to meet the dynamic adjustment needs of the whole life cycle of the project.

[0004] In view of the above problems, the existing technical solutions mostly adopt the "line searching environment" mode, that is, first generate a line scheme, and then evaluate the adaptability of the line to the environment, which has the problem of limited application in complex environments. In recent years, the "environment attracting line" strategy for complex environment has been proposed, which attracts the line through the environment and establishes a gravity field model, but most of the description of the environmental gravity field is based on static or quasi-static conditions, and the core defect is the neglect of environmental dynamics. The static force field analysis assumes that the environmental parameters remain constant within the engineering period, and the route selection evaluation is carried out by establishing a fixed mechanical equilibrium equation. However, in actual engineering, the dynamic changes of environmental factors (such as the creep effect of geological bodies, the instantaneous impact of external loads, and the periodic fluctuation of hydrological conditions) will break the balance state of the static force field hypothesis, resulting in significant deviation of the route selection scheme based on this model.

[0005] Therefore, the existing static field technology is difficult to support the dynamic optimization demand of engineering route selection in a complex environment, and it is urgent to design a method for supporting engineering route selection in a complex environment to solve the problems existing in the prior art. SUMMARY

[0006] The main purpose of the present application is to provide a complex environment railway intelligent route selection gravitational field dynamics modeling and adaptive guidance optimization method to solve the problem that the existing static field technology is difficult to support the dynamic optimization demand of engineering route selection in a complex environment. The specific technical scheme is as follows: A complex environment railway intelligent route selection gravitational field dynamics modeling and adaptive guidance optimization method, comprising the following steps: Step S1, a multi-criteria index system is constructed, a hierarchical model of the adaptation degree space field is established, and a DBSCAN density clustering algorithm is used to identify high-adaptation environment patches and low-adaptation environment patches; an intelligent route selection environment static field model is constructed; Step S2, the dynamic equation of gravitational propagation is established and solved, and the intelligent route selection environment dynamic field modeling is realized; Step S3, a Transformer neural network model is introduced, the gravitational field feature situation awareness is completed, a dynamic adaptation route setting method is proposed, and the optimal route scheme of railway route selection is output.

[0007] Preferably, step S1 specifically comprises: Step S1.1, the FP-tree association rule mining algorithm is used for screening the key adaptation degree indexes for the multi-source heterogeneous data involved in engineering route selection, and the core indexes strongly related to the route adaptation degree are identified; a multi-criteria comprehensive evaluation model is constructed by combining the analytic hierarchy process and the entropy weight method, and the subjective weight and the objective weight of the core indexes are obtained; the comprehensive constraint weight of each core index is obtained through a coupling algorithm, and the construction of the multi-criteria index system is completed; Step S1.2, the research area is standardized grid divided, and the grid unit is obtained; based on the structure chromatography principle and the isotropic assumption, the grid unit is relied on, the engineering area is processed in layers according to the vertical depth gradient, and the hierarchical model of the adaptation degree space field is constructed; the spatial interpolation technology of Kriging interpolation is used to obtain the adaptation degree numerical field relying on the grid unit; Step S1.3, the hot spot analysis algorithm is combined with the adaptation degree numerical field obtained in step S1.2 to perform spatial statistical analysis; the high-adaptation environment patches and the low-adaptation environment patches are obtained; Step S1.4, in the high-adaptation environment patches and the low-adaptation environment patches obtained in step S1.3, the high-adaptation environment patches are defined as gravitational sources, and the low-adaptation environment patches are defined as repulsive sources, and the intelligent route selection environment static field model is constructed.

[0008] Preferably, the layering process in step S1.2 specifically involves classifying engineering structures from top to bottom, including: inaccessible layers, ultra-high bridge layers, high-pier bridge layers, ordinary bridge layers, embankment layers, road cutting layers, shallow buried tunnel layers, deep buried tunnel layers, ultra-deep buried tunnel layers, and inaccessible layers.

[0009] Preferably, in step S1.3, the DBSCAN density clustering algorithm is used for automatic extraction. By setting the minimum number of neighborhood points and the scanning radius, discrete points that are spatially adjacent and whose fitness meets the threshold condition are aggregated into independent patches, and high-fitness spatial hotspot areas, i.e., high-fitness environmental patches, and low-fitness spatial hotspot areas, i.e., low-fitness environmental patches, are identified.

[0010] Preferably, the spatial distribution characteristics and multimodal attributes of the identified high-fit environmental patches and low-fit environmental patches are analyzed respectively; the spatial distribution characteristics include at least one of area, morphology and spatial correlation; the multimodal attributes include at least one of geological stability, engineering feasibility and environmental friendliness.

[0011] Preferably, in step S1.4, when constructing the static field model of the intelligent route selection environment, the following formula is used to calculate the impact of the line element at any spatial point P on the first... The force of each patch : ; Wherein: force When the value is positive, it represents gravity; force. When the value is negative, it represents a repulsive force. The gravitational field proportionality constant; For the first The source intensity of each patch; The sensitivity coefficient of the line element; For spatial point P to the th The straight-line distance between the centers of the patches; For from the first The unit vector pointing from the center of each patch to a spatial point P.

[0012] Preferably, step S2 includes the following steps: Step 2.1: Based on the theory of system propagation dynamics, partial differential equations are introduced to characterize the propagation mechanism of gravitational increments, and the dynamic equations of gravitational propagation are constructed as follows: ; in: It is a time-varying term; For spatial diffusion, It is the diffusion coefficient. It is the Laplace operator; It is a decay term; is an attenuation coefficient; is a gravitational source term, representing the influence of the change of the gravitational source on the gravitational field, is a function of the gravitational source strength , position and time ; Step 2.2, in the numerical simulation of gravitational propagation, the dynamics equation of gravitational propagation is solved step by step by using an iterative recursive method to realize the intelligent route selection environment dynamic field modeling.

[0013] Preferably, the step S3 specifically comprises the following steps: Step S3.1, based on the gravitational field dynamic mechanical characteristics and route information obtained in step S2, a multi-dimensional training data set is constructed; Step S3.2, a multi-modal attention neural network model is constructed; based on the multi-dimensional training data set of step S3.1, the multi-modal attention neural network model is trained and dynamically adapted and optimized to obtain a route scheme intelligent route selection model; the multi-modal attention neural network model is a multi-modal attention model constructed based on a Transformer architecture; Step S3.3, the route scheme intelligent route selection model obtained in step S3.2 is applied to each step of route selection action decision-making from the starting point to the ending point, and finally the optimal route scheme of railway route selection is generated.

[0014] Preferably, the multi-modal attention model comprises an encoder module and a decoder module; the encoder module extracts gravitational field features; the decoder module is used for spatial line feature extraction and cross-modal feature interaction decision-making; the cross-modal feature interaction decision-making is to introduce a gravitational field-line heterogeneous attention mapping network in the decoder, calculate the correlation weight of the gravitational field features and the spatial line features, generate a gravitational field state attention feature map, and output the optimal route selection action suggestion adapted to the current gravitational field.

[0015] Preferably, the training and dynamic adaptation optimization is based on a gradient descent optimization strategy and a back propagation algorithm, and the model parameter iterative optimization is realized through the following process: Defining a loss function, calculating the result of the loss function; Parameter adjustment, specifically: based on the result of the loss function, dynamically adjusting the weight and bias parameters of the Transformer encoder and decoder; Parallel acceleration, specifically: using GPU parallel computing technology to process the operations in model training.

[0016] The application discloses a complex environment railway intelligent route selection gravity field dynamics modeling and adaptive guiding optimization method, which comprises the following steps: constructing a multi-criteria index system, establishing an adaptive degree space field layered model, and identifying high and low adaptive environment patches by using a DBSCAN density clustering algorithm; constructing an intelligent route selection environment static field model; establishing and solving a dynamics equation of gravity propagation to realize intelligent route selection environment dynamic field modeling; introducing a Transformer neural network model to complete gravity field feature situation awareness, proposing a dynamic adaptive route setting method, and outputting a railway route selection optimal route scheme. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained from the structures shown in the drawings without creative labor.

[0018] Figure 1 It is an adaptive degree space field layered model schematic diagram in the embodiment of the present application. Figure 2 It is a gravity field situation awareness to adaptive route setting action mapping mechanism schematic diagram in the embodiment of the present application. Figure 3 It is a railway route selection optimal route scheme schematic diagram output by the route scheme intelligent route selection model in the embodiment of the present application.

[0019] The implementation, functional features and advantages of the present application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments only constitute some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the present application.

[0021] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present application are only used to explain the relative positional relationship, movement condition, etc. between components in a certain specific posture (as shown in the drawings), and if the specific posture changes, the directional indications also change accordingly.

[0022] In addition, the descriptions such as "first", "second", etc. in the present application are only for the purpose of description, and cannot be understood as indicating or implying the relative importance of the indicated technical features or implying the number of the indicated technical features. Therefore, the features defined as "first", "second" can be explicitly or implicitly included at least one of the features. In the description of the present application, the meaning of "a plurality of" is at least two, for example, two, three, etc., unless otherwise explicitly specified and limited.

[0023] In the present application, unless otherwise explicitly specified and limited, the terms "connection", "fixing", etc. should be understood in a broad sense, for example, "fixing" can be fixed connection, or detachable connection, or integral; can be mechanical connection, or electrical connection; can be direct connection, or indirect connection through an intermediate medium, can be the internal communication of two elements or the interaction relationship of two elements, unless otherwise explicitly limited. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0024] In addition, the technical solutions of various embodiments of the present application can be combined with each other, but it must be based on the fact that a person skilled in the art can realize it, when the combination of technical solutions appears contradictory or unachievable, it should be considered that the combination of technical solutions does not exist, nor is it within the scope of protection required by the present application.

[0025] A complex environment railway intelligent route selection gravity field dynamics modeling and adaptive guidance optimization method, comprising the following steps: Step S1, constructing a multi-criteria index system, establishing an adaptive degree space field hierarchical model, and using DBSCAN density clustering algorithm to identify high-adaptive environment patches and low-adaptive environment patches; constructing an intelligent route selection environment static field model; Step S2, establishing and solving the dynamics equation of gravity propagation, realizing intelligent route selection environment dynamic field modeling; Step S3, introducing a Transformer neural network model, completing the feature situation awareness of the gravity field, proposing a dynamic adaptive route setting method, and outputting the optimal route scheme of railway route selection.

[0026] In the present embodiment, step S1 specifically comprises: Step S1.1, for the multi-source heterogeneous data involved in engineering route selection (covering geological exploration data, engineering design parameters, environmental monitoring data), the FP-tree association rule mining algorithm is used to screen the key adaptive degree indexes (specifically: the algorithm performs deep association analysis on multi-dimensional data by constructing a frequent pattern tree), and the core indexes with strong correlation with route selection adaptive degree are efficiently identified.

[0027] The multi-criteria comprehensive evaluation model is constructed by combining the analytic hierarchy process and the entropy weight method. The subjective weight of the core index is assigned by the expert experience through the analytic hierarchy process. The objective weight of the core index is calculated based on the data dispersion by the entropy weight method. The comprehensive constraint weight of each index is obtained by the coupling algorithm, the organic integration of subjective judgment and objective data is realized, and the construction of the multi-criteria index system is completed.

[0028] Step S1.2, the research area is standardized grid division, and the grid unit is obtained. A unified spatial analysis unit (the plane grid size is set according to the multi-source data precision and the engineering structure scale, and the vertical direction is matched with the subsequent layered gradient) is determined. The subsequent structure tomography related processing is based on the grid.

[0029] Based on the structure tomography principle and the isotropic assumption (that is, the line passes through the structure arrangement type corresponding to a designed elevation in any direction is only determined by the passing elevation and the ground surface elevation difference), the engineering area can be processed in layers according to the vertical depth gradient in the case of unknown line scheme, and the engineering area is divided into several levels such as the overground overhead layer, the ground surface covering layer, the shallow rock-soil layer and the deep geological structure layer, and the adaptability spatial field layered model is constructed. In this embodiment, the engineering structure types are divided from top to bottom, including: unreachable layer, super-high bridge layer, high-pier bridge layer, ordinary bridge layer, embankment layer, cutting layer, shallow-buried tunnel layer, deep-buried tunnel layer, super-deep-buried tunnel layer and unreachable layer, which will be described in detail in Figure 1 .

[0030] In order to convert the discrete monitoring data into continuous spatial distribution, the spatial interpolation technology of Kriging interpolation is adopted based on the grid unit to obtain the adaptability numerical field. In this embodiment, the spatial interpolation technology of Kriging interpolation is as follows: the spatial autocorrelation of data in the grid is quantified by the semi-variogram function, and the adaptability numerical field with the minimum estimation error is generated, so as to intuitively present the spatial gradual change characteristics of the adaptability index in different depth levels, and to provide continuous data support for subsequent hot spot identification.

[0031] Step S1.3, combining The hotspot analysis algorithm performs spatial statistical analysis on the fitness numerical field obtained in step S1.2 by calculating the local Getis-Ord statistic for each spatial grid and determining the statistical significance of hotspot areas through Z-score testing. The DBSCAN density clustering algorithm is used for automatic extraction. By setting the minimum number of neighborhood points and the scanning radius, spatially adjacent discrete points that meet the fitness threshold are aggregated into independent patches, effectively eliminating interference from isolated noise points and identifying high-fitness spatial hotspot areas (i.e., high-fitness environmental patches) and low-fitness spatial hotspot areas (i.e., low-fitness environmental patches). For the extracted high-fitness and low-fitness environmental patches, their spatial distribution characteristics (including at least one of area, morphology, and spatial correlation) and multimodal attributes (covering at least one of geological stability, engineering feasibility, and environmental friendliness) are further analyzed, providing multi-dimensional decision-making basis for the dynamic adaptation of subsequent route selection schemes.

[0032] Step S1.4: Traverse the high-fitting and low-fitting environmental patches obtained in step S1.3, define the high-fitting environmental patches as gravity sources and the low-fitting environmental patches as repulsion sources, and construct an intelligent route selection environment static field model.

[0033] In this preferred embodiment, when constructing the static field model of the intelligent route selection environment, the following formula is used to calculate the effect of the first... on the line micro-element at any spatial point P. The force of each patch : ; Wherein: force When the value is positive, it represents gravity; force. When the value is negative, it represents a repulsive force. This is the gravitational field scaling constant (calibrated according to the engineering scenario). For the first The source intensity of each patch (positive values ​​for high-fit patches and negative values ​​for low-fit patches; the absolute value is positively correlated with the patch fit). This is the sensitivity coefficient of the line element (characterizing the strength of the line's response to patch constraints). For spatial point P to the th The straight-line distance between the centers of the patches; For from the first The center of each patch points to a unit vector at a spatial point P (with the gravitational force in the opposite direction).

[0034] Therefore, a mathematical expression model of attraction and repulsion can be constructed to obtain a static field model of intelligent route selection environment.

[0035] Since the environmental change can easily cause the change of the size, position and strength of the gravity source, further disturbing the static field, step S2 extends the static field to the dynamic field based on the static field, depicting the propagation process of the gravity increment.

[0036] Step S2 in this embodiment includes the following steps: Step 2.1, depicting the space-time propagation process of the gravity increment, simulating the influence of the environmental change on the field distribution, specifically: According to the system propagation dynamics theory, partial differential equations are introduced to depict the propagation mechanism of the gravity increment to construct the dynamics equation of the gravity propagation, as follows: ; Among them: is the time change term; is the spatial diffusion term, D is the diffusion coefficient (reflecting the rate of gravity propagation), is the Laplace operator (representing the space gravity diffusion process); is the attenuation term; is the attenuation coefficient, describing the energy loss of the gravity field in the propagation process; is the gravity source term, indicating the influence of the change of the gravity source on the gravity field, which is a function of the gravity source strength , position and time ; Step 2.2, in the numerical simulation of the gravity propagation, the iterative recursion method is used to solve the dynamics equation of the gravity propagation step by step, to realize the intelligent route selection environmental dynamic field modeling, specifically: Based on the initial gravity source strength and spatial position information , the initial gravity field state is constructed. At intervals of a small time step , in each time step, according to the gravity field state at the previous time, combined with the dynamic change of the gravity source strength , position , the gravity field increment at the current time is calculated . These increments are substituted into the dynamics equation of the gravity propagation to update the solution of the gravity field at the current time . Starting from the initial state, through continuous iteration and recursion, the propagation process of the gravity is accurately simulated in the time and space dimensions, the dynamic evolution of the gravity field is tracked in real time, and the dynamics propagation range, path and space-time evolution law of the gravity are revealed.

[0037] Thus, the establishment and solution process of the dynamics equation of the gravity propagation is completed, the dynamic evolution of the gravity field caused by the environmental change can be tracked in real time, and the intelligent route selection environmental dynamic field modeling is realized.

[0038] After the dynamic process of gravity propagation is described, step S3 proposes a dynamic adaptation route determination method; the entire route determination process is divided into several sequential route determination actions, and a Transformer neural network model is introduced in each route determination action, which is finally iteratively output after training. Details are as follows: Step S3 specifically includes the following steps: Step S3.1, based on the gravity field dynamic mechanical characteristics obtained in step S2 and the route information, a multi-dimensional training data set is constructed. In this embodiment, the gravity field dynamic mechanical characteristics include gravity source intensity, spatial position and gravity intensity; the route information (including historical route selection three-dimensional coordinates and geometric parameters). Here, the above-mentioned gravity field dynamic mechanical characteristics of local gravity field, global gravity field and time series gravity field are further optimized.

[0039] Step S3.2, a multi-modal attention neural network model is constructed; based on the multi-dimensional training data set of step S3.1, the multi-modal attention neural network model is trained and dynamically adapted and optimized to obtain a route scheme intelligent route selection model; the multi-modal attention neural network model is a multi-modal attention model constructed based on the Transformer architecture. Specifically: The multi-modal attention neural network model is constructed and the features are extracted, which is described in detail in Figure 2 The multi-modal attention model is constructed based on the Transformer architecture, and the deep interaction and long-range dependence mining of the gravity field and the route features are realized through a three-level feature fusion mechanism. Details are as follows: the multi-modal attention model includes an encoder module and a decoder module. The encoder module extracts the gravity field features, specifically: a Transformer encoder is used to analyze the local and global spatio-temporal distribution of the gravity field mechanical parameters through a multi-head self-attention mechanism, to capture the long-range dependence features of the gravity field in the time dimension (history and future) and the space dimension (local and global), and to generate a structured gravity field feature vector.

[0040] The decoder module is used for spatial line feature extraction and cross-modal feature interaction decision, wherein: The spatial line feature extraction is specifically: a Transformer decoder is constructed, a time series spatial line model is constructed by introducing an autoregressive attention mechanism, and the high-dimensional implicit features of the spatial line are learned autonomously by analyzing the spatio-temporal continuity of the historical line sequence.

[0041] The cross-modal feature interaction decision introduces a gravity field-line heterogeneous attention mapping network (cross-attention layer) in the decoder to calculate the correlation weight of the gravity field feature and the spatial line position feature, generate a gravity field state attention feature map, and output the optimal line setting action suggestion adapted to the current gravity field. The specific contents include: hidden space feature, K-dimensional gravity field sequence, M-dimensional line setting action sequence, and N-dimensional cross-feature sequence.

[0042] In this embodiment, the training and dynamic adaptation optimization are further preferably based on a gradient descent optimization strategy and a back propagation algorithm, and the model parameter iterative optimization is realized through the following process: Define the loss function Loss Calculate the result of the loss function. The loss function is used to quantify the difference between the model output line setting action and the artificial optimization scheme as the basis for parameter updating.

[0043] Parameter adjustment, specifically: based on the result of the loss function, dynamically adjust the weight and bias parameters of the Transformer encoder and decoder to optimize the mapping relationship between the gravity field feature and the line setting action.

[0044] Parallel acceleration, specifically: use GPU parallel computing technology to process high-complexity operations (complexity O (n²d), where n is the sequence length and d is the embedding dimension) in model training to improve training efficiency.

[0045] Through multiple rounds of iterative training, the model has the ability to accurately map from the gravity field feature to the adaptive line setting action, realizes intelligent optimization of the line scheme under the dynamic change of the gravity field, and improves the real-time performance and adaptability of the line selection decision.

[0046] Step S3.3, model application and line setting decision, specifically: apply the line scheme intelligent line selection model obtained in step S3.2 to each step of line setting action decision from the starting point to the ending point; after several repeated line setting actions, determine the optimized line scheme from the starting point to the ending point, output the optimal line scheme of railway line selection, and propose a dynamic adaptation line setting method to realize "line setting from field".

[0047] Application case: The FP-tree association rule mining algorithm is used to determine 5 key indicators affecting the environment-line adaptation degree evaluation, which are cost, geological disaster risk, stability, carbon emission, and line short straight direction deviation. The analytic hierarchy process and entropy weight method are used to determine the index weight.

[0048] The definition of the unreachable high bridge threshold is 1000m, the super high bridge-high pier bridge threshold is 150m, the high pier bridge-ordinary bridge is 50m, the ordinary bridge-embankment threshold is 20m, the cutting-shallow tunnel depth threshold is-20m, the shallow-deep tunnel depth threshold is-400m, and the deep-super deep tunnel depth threshold is-1000m. According to the above, the spatial structure layer is divided from top to bottom. According to the calculation model of each adaptive key index in different structure layers and the respective weight, the environment-line adaptability value of each spatial unit is calculated, and the continuous adaptability spatial distribution field is formed by using the Kriging spatial interpolation technology.

[0049] The introduction of The Getis-Ord statistics of each spatial unit is calculated by the hotspot analysis method, and the spatial unit with Getis-Ord>0 and Z>1.96 (corresponding to 95% confidence) is defined as the high adaptability spatial hotspot area, and the remaining is the low adaptability spatial hotspot area. The minimum number of domain points of the DBSCAN clustering method is defined as 8, and the scanning radius is 180m. The high adaptability environment patch and the low adaptability environment patch are clustered respectively.

[0050] The spatial grid unit is traversed in turn, and the size and direction (range of 360°) of the attractive force of each spatial grid unit are calculated, wherein the number of attractive and repulsive force sources is 23 and 15 respectively.

[0051] 79 artificial line optimization schemes are selected, and the intelligent line selection environment static field model of step S1 and the intelligent line selection environment dynamic field model of step S2 are modeled. The input vector of the Transformer model is generated by combining the sliding window and data enhancement method.

[0052] The three-layer attention mechanism of the Transformer is used to extract the features of the attractive field state to the optimal line action (including the forward direction and step length of the current intersection point).

[0053] The difference between the Transformer predicted line action and the optimal line action is defined as the loss function, and the input set is continuously trained under the GPU parallel operation framework to obtain the Transformer network structure storing the attractive field state to the optimal line action. The attractive field state is input into the intelligent line selection model of the line scheme, and the forward step length and direction of each intersection point are predicted by using the above intelligent line selection model of the line scheme, and the optimal line scheme of the railway line selection is output, which is described in detail in Figure 3 .

[0054] By applying the technical scheme of the embodiment, the line selection environment adaptability spatial field is constructed and converted into an attractive field, the dynamic evolution mechanism analysis and the self-adaptive guiding intelligent search strategy are combined, the efficient optimization design of the railway line in the dense constraint environment is realized, the strain capacity of the line selection decision is improved, and a new technical approach for the railway line selection in the complex environment is provided.

[0055] The embodiments of the present application are described above with reference to the accompanying drawings, but the present application is not limited to the above-described specific embodiments, and the above-described specific embodiments are merely illustrative, but not restrictive, and a person of ordinary skill in the art can make many forms under the inspiration of the present application without departing from the purpose of the present application and the scope protected by the claims.

Claims

1. A complex environment railway intelligent route selection gravitational field dynamics modeling and adaptive guidance optimization method, characterized in that, Comprise the following steps: Step S1, construct a multi-criteria index system, establish an adaptive space field hierarchical model, and use DBSCAN density clustering algorithm to identify high and low adaptive environment patches; construct an intelligent alignment environment static field model; Step S2, establish and solve the dynamics equation of gravitational propagation to realize intelligent alignment environment dynamic field modeling; Step S3, introduce the Transformer neural network model to complete the gravitational field feature situation awareness, propose a dynamic adaptive alignment method, and output the optimal route scheme of railway alignment.

2. The complex environment railway intelligent route selection gravitational field dynamics modeling and adaptive guidance optimization method of claim 1, wherein, Step S1 specifically includes: Step S1.1, for the multi-source heterogeneous data involved in engineering alignment, use the FP-tree association rule mining algorithm to screen the key adaptive index, identify the core index strongly related to the alignment adaptability; combine the analytic hierarchy process and entropy weight method to construct a multi-criteria comprehensive evaluation model, obtain the subjective weight and objective weight of the core index; get the comprehensive constraint weight of each core index through coupling algorithm, and complete the construction of multi-criteria index system; Step S1.2, standardize the grid division of the study area to obtain grid cells; based on the structure chromatography principle and the isotropic assumption, rely on the grid cells, and process the engineering area in layers according to the vertical depth gradient to construct an adaptive space field hierarchical model; rely on the grid cells to obtain the adaptive numerical field by using the spatial interpolation technology of Kriging interpolation; Step S1.3, combination The hotspot analysis algorithm performs spatial statistical analysis on the fitness value field obtained in step S1.2; and obtains high-fitness environment patches and low-fitness environment patches. Step S1.4, in the high and low adaptive environment patches obtained in step S1.3, define the high adaptive environment patches as gravitational sources and the low adaptive environment patches as repulsive sources to construct an intelligent alignment environment static field model.

3. The complex environment railway intelligent route selection gravitational field dynamics modeling and adaptive guidance optimization method of claim 2, wherein, The hierarchical processing in step S1.2 is: top-down division of engineering structure types, including: inaccessible layer, super-high bridge layer, high-pier bridge layer, ordinary bridge layer, embankment layer, cutting layer, shallow tunnel layer, deep tunnel layer, super-deep tunnel layer and inaccessible layer.

4. The complex environment railway intelligent route selection gravitational field dynamics modeling and adaptive guidance optimization method of claim 2, wherein, In step S1.3, the DBSCAN density clustering algorithm is used for automatic extraction. By setting the minimum number of neighborhood points and the scanning radius, the discrete points that are adjacent in space and meet the threshold condition of adaptability are aggregated into independent patches, and the high adaptive environment patches and the low adaptive environment patches are identified.

5. The complex environment railway intelligent route selection gravitational field dynamics modeling and adaptive guidance optimization method of claim 4, wherein, The spatial distribution characteristics and multi-modal attributes of the identified high and low adaptive environment patches are analyzed respectively; the spatial distribution characteristics include at least one of area, shape and spatial correlation; the multi-modal attributes include at least one of geological stability, engineering feasibility and environmental friendliness.

6. The complex environment railway intelligent route selection gravitational field dynamics modeling and adaptive guidance optimization method of claim 2, wherein, In step S1.4, when constructing the static field model of the intelligent route selection environment, the following formula is used to calculate the impact of the line element at any spatial point P on the first... The force of each patch : ； where: force is positive, it is an attractive force; force is negative, it is a repulsive force. is the attractive field proportionality constant; is the source intensity of the th patch; is the sensitivity coefficient of the line element; is the spatial straight-line distance from the spatial point P to the center of the th patch; is the unit vector pointing from the center of the th patch to the spatial point P.

7. The complex environment railway intelligent route selection gravitational field dynamics modeling and adaptive guidance optimization method of claim 6, wherein, Step S2 includes the following steps: Step 2.1, according to the system propagation dynamics theory, introduce partial differential equation to describe the propagation mechanism of gravitational increment to construct the dynamics equation of gravitational propagation as follows: ； wherein: is a time variation term; is a spatial diffusion term, is a diffusion coefficient, is a Laplacian operator; is an attenuation term; is an attenuation coefficient; is a gravitational source term, representing the effect of the variation of the gravitational source on the gravitational field, is a function of the gravitational source intensity , position and time ; Step 2.2, in the numerical simulation of gravitational propagation, the iterative recursive method is used to solve the dynamics equation of gravitational propagation step by step to realize intelligent alignment environment dynamic field modeling.

8. The complex environment railway intelligent route selection gravitational field dynamics modeling and adaptive guidance optimization method of claim 7, wherein, The step S3 specifically includes the following steps: Step S3.1, based on the dynamic force characteristics of the gravitational field obtained in step S2 and the line information, a multi-dimensional training data set is constructed; Step S3.2, a multi-modal attention neural network model is constructed; based on the multi-dimensional training data set of step S3.1, the multi-modal attention neural network model is trained and dynamically adapted and optimized to obtain a line scheme intelligent alignment model; the multi-modal attention neural network model is a multi-modal attention model constructed based on the Transformer architecture; Step S3.3, the line scheme intelligent alignment model obtained in step S3.2 is applied to each step of the alignment action decision from the starting point to the ending point, and finally the optimal line scheme of railway alignment is generated.

9. The complex environment railway intelligent route selection gravity field dynamics modeling and adaptive guidance optimization method of claim 8, wherein, The multi-modal attention model includes an encoder module and a decoder module; the encoder module extracts gravitational field features; The decoder module is used for spatial line feature extraction and cross-modal feature interaction decision; The cross-modal feature interaction decision is to introduce a gravitational field-line heterogeneous attention mapping network in the decoder, calculate the correlation weight of the gravitational field features and the spatial line features, generate a gravitational field state attention feature map, and output the optimal alignment action suggestion adapted to the current gravitational field.

10. The complex environment railway intelligent route selection gravitational field dynamics modeling and adaptive guidance optimization method of claim 8, wherein, The training and dynamic adaptation optimization is based on the gradient descent optimization strategy and the back propagation algorithm, and the model parameter iterative optimization is realized through the following process: Define the loss function and calculate the result of the loss function; Parameter adjustment, specifically: based on the result of the loss function, dynamically adjust the weight and bias parameters of the Transformer encoder and decoder; Parallel acceleration, specifically: use GPU parallel computing technology to process the operations in model training.

Images