Railway route selection effect evaluation method and system based on artificial intelligence

Abstract

The application discloses a railway line selection effect evaluation method and system based on artificial intelligence, and belongs to the technical field of railway intelligent line selection. The method comprises data preparation, toughness feature extraction, climate toughness scoring and multi-dimensional effect evaluation. The application adopts line climate toughness feature extraction based on multi-source space-time toughness, comprehensively analyzes line structure characteristics, environmental conditions and climate disturbance, quantifies the stability and adaptability of different line sections at different space-time scales, and thus can more accurately reflect the overall toughness level of the line. The climate toughness scoring combined with multi-disaster risk and dynamic correction can comprehensively consider various disaster types and their interaction in the line selection process, identify key vulnerable sections and potential risk points through dynamic correction, and thus provide a scientific basis for line scheme comparison and optimization.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of intelligent railway route selection, and specifically refers to a railway route selection effect evaluation method and system based on artificial intelligence. BACKGROUND

[0002] The railway route selection effect evaluation based on artificial intelligence utilizes artificial intelligence technology to comprehensively analyze railway route schemes, generates climate resilience scores and multi-dimensional effect evaluation results, aims to discover weak links of the line under extreme climate conditions such as floods, landslides, and freeze-thaw in advance, provides targeted risk prevention and control basis for railway route selection, improves the climate adaptability of the whole life cycle of the railway, and promotes the transformation of railway route selection from experience-driven to data-driven and intelligent-driven.

[0003] However, in the existing railway route selection effect evaluation process, the vulnerability of the line under various climate conditions is difficult to be fully quantified, traditional evaluation methods rely on single indicators or static experience, and cannot accurately reflect the comprehensive influence of line structure, geological environment and climate events, and it is difficult to evaluate the dynamic influence of extreme weather on line safety and operation capacity; traditional railway route selection often focuses on topography, cost and engineering feasibility, but ignores the combined influence of disasters under the background of climate change (such as landslides caused by heavy rain, deformation of subgrade caused by freeze-thaw, expansion of rails under high temperature, and flood erosion, etc.), and disasters often have spatio-temporal coupling characteristics, so single indicators cannot accurately reflect the overall risk level of the line, resulting in the technical problem that the route selection decision ignores the potential high-risk area. SUMMARY

[0004] In view of the above situation, in order to overcome the defects of the prior art, the present application provides a railway route selection effect evaluation method and system based on artificial intelligence, which creatively adopts line climate resilience feature extraction based on multi-source spatio-temporal resilience, comprehensively analyzes line structure characteristics, environmental conditions and climate disturbance, and quantifies the stability and adaptability of different line sections under different spatio-temporal scales, so as to more accurately reflect the overall resilience level of the line; creatively adopts climate resilience scoring combined with multi-disaster risk and dynamic correction, which can comprehensively consider multiple disaster types and their interactive influence in the route selection process, and identify key vulnerable sections and potential risk points through dynamic correction, thereby providing a scientific basis for line scheme comparison and optimization.

[0005] The technical scheme adopted by the present application is as follows: the present application provides a railway route selection effect evaluation method based on artificial intelligence, which comprises the following steps:

[0006] Step S1: data preparation;

[0007] Step S2: resilience feature extraction;

[0008] Step S3: climate resilience scoring;

[0009] Step S4: multi-dimensional effect evaluation.

[0010] Further, in step S1, the data preparation, specifically collecting railway line structure data, climate environment data and railway operation data, and performing dimension normalization, time synchronization, space matching and unified storage management, to obtain a railway multi-source data set.

[0011] Further, in step S2, the resilience feature extraction, specifically based on the railway multi-source data set, using line climate resilience feature extraction based on multi-source spatio-temporal resilience, to obtain comprehensive climate resilience features, including the following steps:

[0012] Step S21: climate feature mapping, for quantifying the multi-scale influence pattern of climate events, specifically by improving the Transformer model to map the climate environment data to generate climate multi-scale features;

[0013] The improved Transformer model, specifically introducing a multi-scale perturbation gating mechanism into the standard Transformer model, for dynamically adjusting the response to different climate event intensities;

[0014] The multi-scale perturbation gating mechanism, specifically by performing multi-scale convolution on the climate environment data to obtain climate perturbation response features, and calculating the perturbation gating weight corresponding to each scale, and weighting and fusing the multi-scale climate perturbation response features according to the perturbation gating weight, to obtain climate multi-scale features;

[0015] Step S22: structure environment feature modeling, for constructing the coupling relationship between line structure and external environment, specifically constructing a structure environment bidirectional graph, coupling and fusing the structure node features and environment node features by improving the graph convolution network model, and generating structure environment coupling features after node updating;

[0016] The structure environment bidirectional graph, specifically by constructing a line structure graph and a line environment graph, and establishing bidirectional edges, to obtain the structure environment bidirectional graph;

[0017] The line structure graph, specifically taking the engineering structure units in the railway line as structure nodes, and setting structure node features, establishing edges between structure nodes according to the line topology relationship, forming a line adjacency matrix, and obtaining the line structure graph;

[0018] The line environment graph, specifically taking the regional environment units as environment nodes, and setting environment node features, establishing edges between environment nodes according to spatial proximity, to obtain the line environment graph;

[0019] The bidirectional edge is established, specifically, when the geographical position of the structure node is in the coverage area of the environment node, a bidirectional edge between the structure node and the environment node is established to obtain a structure-environment bidirectional graph.

[0020] The improved graph convolution network model is specifically introduced into a dual-domain attention and a geology gate constraint in a standard graph convolution network model.

[0021] The dual-domain attention is specifically a self-attention mechanism used to calculate a structure domain attention weight and an environment domain attention coefficient, respectively, for adaptively adjusting the contribution size of different types of neighbors to node feature updating.

[0022] The step S23 is a spatiotemporal resilience embedding, specifically, the structure-environment coupling feature and the line adjacency matrix are taken as inputs, and a standard spatiotemporal graph convolution network model is used for spatiotemporal resilience embedding to model the dynamic response of the structure node to climate disturbance at different time and space scales, and spatiotemporal resilience embedding features are obtained.

[0023] The step S24 is a resilience feature fusion, specifically, the climate multi-scale feature, the structure-environment coupling feature and the spatiotemporal resilience embedding feature are weighted and fused, and are standardized by layer normalization to obtain comprehensive climate resilience features, which are used to represent the overall resilience capability of the railway line under climate disturbance.

[0024] Further, in the step S3, the climate resilience score is specifically obtained according to the railway multi-source data set and the comprehensive climate resilience feature, and a climate resilience score combining multi-disaster risk and dynamic correction is used to obtain a line climate resilience score, including the following steps:

[0025] The step S31 is a vulnerability loss mapping, which is used to establish the relationship between climate disaster intensity and line segment performance loss, specifically, a structure vulnerability function is constructed by fitting the comprehensive climate resilience feature to calculate the damage probability of the line segment under different disaster intensities, and then the damage probability is mapped to the performance decline loss by combining the line operation recovery process to obtain the line segment performance loss index under each disaster;

[0026] The step S32 is a multi-disaster risk aggregation, which is used to aggregate the loss influence of each disaster on the same line segment, specifically, the line segment performance loss index under each disaster is weighted and aggregated, and a tail risk constraint is introduced to obtain a line segment aggregated risk index;

[0027] The step S33 is a line overall evaluation, which is used to generate a line-level climate resilience score, specifically, a risk fluctuation factor is calculated by historical disaster simulation, the line segment aggregated risk index is dynamically corrected, and then weighted and summarized to obtain a line climate resilience score.

[0028] Furthermore, in step S4, the multidimensional effect evaluation specifically involves combining the line climate resilience score with multidimensional key evaluation indicators, and using a multi-objective optimization method to comprehensively score and rank the railway route selection schemes to obtain a railway route selection effect evaluation report.

[0029] The railway alignment evaluation system based on artificial intelligence provided by this invention includes a data construction module, a feature construction module, a resilience scoring module, and an alignment evaluation module.

[0030] The data construction module is used for data preparation. Through data preparation, a multi-source railway dataset is obtained, and the multi-source railway dataset is sent to the feature construction module and the resilience scoring module.

[0031] The feature construction module is used for resilience feature extraction. Through resilience feature extraction, a comprehensive climate resilience feature is obtained, and the comprehensive climate resilience feature is sent to the resilience scoring module.

[0032] The resilience scoring module is used for climate resilience scoring. Through the climate resilience scoring, the line climate resilience score is obtained, and the line climate resilience score is sent to the line selection effect evaluation module.

[0033] The route selection effect evaluation module is used for multi-dimensional effect evaluation, and a railway route selection effect evaluation report is obtained through multi-dimensional effect evaluation.

[0034] The beneficial effects achieved by adopting the above solution are as follows:

[0035] (1) In the existing railway alignment effect evaluation process, it is difficult to fully quantify the vulnerability of the line under various climatic conditions. Traditional evaluation methods mostly rely on single indicators or static experience, which cannot accurately reflect the comprehensive impact of line structure, geological environment and climate events, and it is difficult to assess the dynamic impact of extreme weather on line safety and operation capacity. This scheme creatively adopts line climate resilience feature extraction based on multi-source spatiotemporal resilience, comprehensively analyzes line structural characteristics, environmental conditions and climate disturbances, quantifies the stability and adaptability of different line sections under different spatiotemporal scales, and thus can more accurately reflect the overall resilience level of the line.

[0036] (2) In the existing railway route selection effect evaluation process, the traditional railway route selection often focuses on terrain, cost and engineering feasibility, while ignoring the disaster composite influence under the background of climate change (such as landslide caused by heavy rain, roadbed deformation caused by freeze-thaw, high temperature rail expansion and flood erosion, etc.), and the disasters often have spatio-temporal coupling characteristics, and a single index cannot accurately reflect the overall risk level of the line, leading to the technical problem that the route selection decision ignores the potential high-risk area. The scheme creatively uses the climate resilience score combining multi-disaster risk and dynamic correction, which can comprehensively consider multiple disaster types and their interaction in the route selection process, and identify key vulnerable sections and potential risk points through dynamic correction, thereby providing a scientific basis for line scheme comparison and optimization. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 The flowchart of the railway route selection effect evaluation method based on artificial intelligence provided by the present application is shown.

[0038] Figure 2 The module schematic diagram of the railway route selection effect evaluation system based on artificial intelligence provided by the present application is shown.

[0039] Figure 3 The flowchart of step S2 is shown.

[0040] Figure 4 The flowchart of step S3 is shown.

[0041] The accompanying drawings are used to provide a further understanding of the present application, and constitute a part of the specification, together with the embodiments of the present application, to explain the present application, and do not constitute a limitation on the present application. DETAILED DESCRIPTION

[0042] 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 are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0043] In the description of the present application, it should be understood that the terms "upper", "lower", "front", "back", "left", "right", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, therefore cannot be understood as a limitation on the present application.

[0044] Embodiment one, refer to Figure 1The application provides a railway line selection effect evaluation method based on artificial intelligence, which comprises the following steps:

[0045] Step S1: data preparation;

[0046] Step S2: resilience feature extraction;

[0047] Step S3: climate resilience scoring;

[0048] Step S4: multi-dimensional effect evaluation.

[0049] Embodiment two, see Figure 1 This embodiment is based on the above-mentioned embodiment, in step S1, the data preparation, specifically collecting railway line structure data, climate environment data and railway operation data, and performing dimension normalization, time synchronization, space matching and unified storage management, to obtain a railway multi-source data set;

[0050] The railway line structure data includes total length of line, line section division, engineering structure unit and its attributes; the climate environment data includes climate data, geological data, hydrological data and ecological data, by dividing the climate environment data according to the space grid, to form a regional environment unit; the railway operation data includes line transport capacity, transport volume and historical disaster recovery situation.

[0051] Embodiment three, see Figure 1 and Figure 3 This embodiment is based on the above-mentioned embodiment, in step S2, the resilience feature extraction, for quantifying the structural stability and operation adaptability of the railway line under various climate disturbances, specifically according to the railway multi-source data set, adopting line climate resilience feature extraction based on multi-source spatio-temporal resilience, to obtain comprehensive climate resilience features, comprising the following steps:

[0052] Step S21: climate feature mapping, for quantifying the multi-scale influence mode of climate events, specifically by improving the Transformer model, mapping the climate environment data to generate climate multi-scale features;

[0053] The improved Transformer model, specifically introducing a multi-scale disturbance gating mechanism in the standard Transformer model, for dynamically adjusting the response to different climate event intensities;

[0054] The multi-scale disturbance gating mechanism, specifically by performing multi-scale convolution on the climate environment data to obtain climate disturbance response features, and calculating the disturbance gating weight corresponding to each scale, weighting and fusing the multi-scale climate disturbance response features according to the disturbance gating weight, to obtain the climate multi-scale features, the calculation formula is:

[0055] ;

[0056] ;

[0057] where Z t is the multi-scale climate feature, k is the scale index, K is the scale number, g (k) is the disturbance gating weight corresponding to the kth scale, G (k) is the climate disturbance response feature of the kth scale, is the Sigmoid function, is the gating weight of the kth scale, [·,·] is the splicing operation, mean(·) is the mean calculation function, var(·) is the variance calculation function, is the gating bias term of the kth scale.

[0058] Step S22: structure environment feature modeling, for constructing the coupling relationship between the line structure and the external environment, specifically for constructing the structure environment bidirectional graph, coupling and fusing the structure node features and the environment node features through the improved graph convolution network model, generating the structure environment coupling features through node updating;

[0059] The structure environment bidirectional graph is constructed, specifically by constructing a line structure graph and a line environment graph, and establishing bidirectional edges to obtain the structure environment bidirectional graph.

[0060] The line structure graph is specifically that engineering structure units in the railway line are taken as structure nodes, and structure node features are set, edges between structure nodes are established according to the line topological relationship, a line adjacency matrix is formed, and a line structure graph is obtained; the engineering structure units include bridges, tunnels, roadbeds, culverts, retaining walls and embankments, and the structure node features include structure types, lengths, bearing capacities and material properties.

[0061] The line environment graph is specifically that regional environment units are taken as environment nodes, and environment node features are set, edges between environment nodes are established according to spatial proximity, and a line environment graph is obtained; the regional environment units include climate grid units, geological units, hydrological units and ecological coverage units, and the environment node features include climate parameters, geological parameters, topographic parameters and ecological parameters.

[0062] The bidirectional edges are established, specifically when the geographical position of the structure node is in the coverage area of the environment node, the bidirectional edges between the structure node and the environment node are established, and the structure environment bidirectional graph is obtained.

[0063] The improved graph convolution network model is specifically that a dual-domain attention and a geological gating constraint are introduced into a standard graph convolution network model, which is used for adaptively transmitting information between different types of nodes and enhancing the perception of complex geological differences.

[0064] The dual-domain attention, in particular, the structural domain attention weight and the environmental domain attention coefficient are calculated respectively by the self-attention mechanism, and are used for adaptively adjusting the contribution size of different types of neighbors to the node feature update;

[0065] The calculation formula of the geology gating constraint is:

[0066] ;

[0067] In the formula, is the geology gating constraint, W g is the gating weight, is the geology class parameter of the kth environmental node, k is the environmental node index, b g is the gating bias term;

[0068] The calculation formula of the node update is:

[0069] ;

[0070] In the formula, is the i th structural node feature of the lay+1 th layer, lay is the graph convolution layer index, i is the first index of the structural node, j is the second index of the structural node, ReLU(·) is the ReLU activation function, W s is the structural node mapping weight, N s (i) is the neighbor structural node set of the i th structural node, is the structural domain attention weight, is the j th structural node feature of the lay th layer, W e is the environmental node mapping weight, N e (i) is the neighbor environmental node set of the i th structural node, is the environmental domain attention coefficient, F k is the k th environmental node feature;

[0071] Step S23: spatiotemporal resilience embedding, in particular, the structural environmental coupling feature and the line adjacency matrix are taken as inputs, and the standard spatiotemporal graph convolution network model is used for spatiotemporal resilience embedding, so as to realize modeling of the dynamic response of the structural node to the climate disturbance under different time and space scales, and obtain the spatiotemporal resilience embedding feature;

[0072] Step S24: resilience feature fusion, in particular, the climate multi-scale feature, the structural environmental coupling feature and the spatiotemporal resilience embedding feature are weighted and fused, and are standardized by layer normalization, so as to obtain the comprehensive climate resilience feature, which is used to represent the overall resilience capability of the railway line under the climate disturbance.

[0073] By performing the above operations, this solution addresses the technical problems in the existing railway alignment evaluation process, such as the difficulty in comprehensively quantifying the vulnerability of railway lines under various climatic conditions, the reliance of traditional evaluation methods on single indicators or static experience, the inability to accurately reflect the combined impact of line structure, geological environment, and climate events, and the difficulty in assessing the dynamic impact of extreme weather on line safety and operational capacity. This solution creatively adopts line climate resilience feature extraction based on multi-source spatiotemporal resilience, comprehensively analyzes line structural characteristics, environmental conditions, and climate disturbances, and quantifies the stability and adaptability of different line sections at different spatiotemporal scales, thereby more accurately reflecting the overall resilience level of the line.

[0074] Example 4, see Figure 1 and Figure 4 This embodiment is based on the above embodiment. In step S3, the climate resilience score is used to transform the comprehensive climate resilience characteristics into comparable resilience evaluation indicators. Specifically, based on the railway multi-source dataset and comprehensive climate resilience characteristics, a climate resilience score combining multi-hazard risk and dynamic correction is used to obtain the line climate resilience score, including the following steps:

[0075] Step S31: Vulnerability loss mapping, used to establish the relationship between the intensity of climate disasters and the performance loss of line segments. Specifically, by fitting the comprehensive climate resilience characteristics, a structural vulnerability function is constructed, the damage probability of line segments under different disaster intensities is calculated, and then, combined with the line operation recovery process, the damage probability is mapped to performance degradation loss to obtain the line segment performance loss index under each disaster type.

[0076] The formula for calculating the structural vulnerability function is as follows:

[0077] ;

[0078] In the formula, This represents the probability of damage to the r-th line segment at disaster type m and disaster intensity h, where r is the line segment index, the line segment consists of structural nodes within the same spatial range, m is the disaster type index, and h is the disaster intensity. It is the standard normal distribution function, and ln is the natural logarithm operator. It is the median value parameter of the structural vulnerability function, used to represent the median damage intensity. It is the scale parameter of the structural vulnerability function, used to represent the sensitivity to the damage probability distribution. and stated All of these were obtained by fitting the comprehensive climate resilience characteristics using a regression network model;

[0079] The formula for calculating the performance loss index of the line segment is as follows:

[0080] ;

[0081] ;

[0082] wherein, is the performance loss index of the rth line section at the intensity h of the mth hazard, is the time integral of the performance loss of the line section during the recovery period, and t is the post-disaster time index, is the total recovery time after the disaster, is the line section capacity at the post-disaster time t, is the line section capacity before the disaster, is the repair progress curve, which can be set according to historical disaster recovery conditions, including but not limited to a linear repair progress curve, an exponential recovery progress curve, and an S-shaped recovery progress curve;

[0083] Step S32: Multi-hazard risk aggregation, for aggregating the loss effects of each hazard on the same line section, specifically by weighting and aggregating the performance loss indexes of the line section under each hazard, and introducing a tail risk constraint to obtain a line section aggregated risk index, the calculation formula being:

[0084] ;

[0085] ;

[0086] wherein, R r is the rth line section aggregated risk index, and M is the number of hazards, is the hazard weight, is a tail risk penalty factor for adjusting the influence degree of the tail risk, is the tail risk at the confidence level for measuring the average loss under extreme disasters, is the confidence level, and E[·] is the expectation operator, is the quantile of the line section performance loss distribution at the confidence level ;

[0087] Step S33: Line overall evaluation, for generating a line-level climate resilience score, specifically by calculating a risk fluctuation factor through historical disaster simulation, dynamically correcting the line section aggregated risk index, and then weighting and summarizing to obtain a line climate resilience score, the calculation formula being:

[0088] ;

[0089] ;

[0090] wherein, is the corrected rth line section aggregated risk index, is a regulation coefficient for controlling the influence degree of the risk fluctuation factor, max(·) is a maximum value function, is the risk fluctuation factor of the rth line section under the historical disaster simulation, specifically refers to the standard deviation of the aggregated risk index of the rth line section under the historical disaster simulation, is a fluctuation threshold, Score line is the line climate resilience score, N is the number of line sections of the entire line, w r is the weight of the rth line section, R max is the maximum value of the aggregated risk index of the line section.

[0091] By performing the above operation, in the existing railway route selection effect evaluation process, the traditional railway route selection often focuses on the terrain, cost and engineering feasibility, and ignores the disaster composite influence under the background of climate change (such as landslide caused by heavy rain, roadbed deformation caused by freeze-thaw, high temperature rail expansion and flood erosion, etc.), and disasters often have spatio-temporal coupling characteristics, and a single index cannot accurately reflect the overall risk level of the line, resulting in the technical problem that the route selection decision ignores the potential high-risk area. The scheme creatively uses the climate resilience score combining multi-disaster risk and dynamic correction, which can comprehensively consider multiple disaster types and their interactive effects in the route selection process, and identifies key vulnerable sections and potential risk points through dynamic correction, thereby providing a scientific basis for line scheme comparison and optimization.

[0092] Embodiment five, refer to Figure 1 This embodiment is based on the above-mentioned embodiments, in step S4, the multi-dimensional effect evaluation, specifically combining the line climate resilience score with the multi-dimensional key evaluation index, and comprehensively scoring and sorting the railway route selection scheme through a multi-objective optimization method, to obtain a railway route selection effect evaluation report;

[0093] The multi-dimensional key evaluation index includes but is not limited to operation and maintenance cost evaluation index, environmental impact evaluation index and social benefit index; the multi-objective optimization method includes but is not limited to weighted comprehensive scoring method, analytic hierarchy process, TOPSIS method, genetic algorithm and multi-objective programming method.

[0094] Embodiment six, refer to Figure 2 This embodiment is based on the above-mentioned embodiments, and the railway route selection effect evaluation system based on artificial intelligence provided by the present application includes a data construction module, a feature construction module, a resilience scoring module and a route selection effect evaluation module;

[0095] The data construction module is used for data preparation, through data preparation, a railway multi-source data set is obtained, and the railway multi-source data set is sent to the feature construction module and the resilience scoring module;

[0096] The feature construction module is configured to extract the resilience features, obtain comprehensive climate resilience features through the resilience feature extraction, and send the comprehensive climate resilience features to the resilience scoring module.

[0097] The resilience scoring module is configured to score the climate resilience, obtain a line climate resilience score through the climate resilience scoring, and send the line climate resilience score to the line selection effect evaluation module.

[0098] The line selection effect evaluation module is configured to evaluate the multi-dimensional effect, and obtain a railway line selection effect evaluation report through the multi-dimensional effect evaluation.

[0099] It should be noted that, in this document, the terms such as first and second are used merely to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprise", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed or inherent to such a process, method, article or device.

[0100] Although the embodiments of the present application have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present application.

[0101] The above describes the present application and its embodiments, which are not restrictive, and the embodiments shown in the drawings are only one of the embodiments of the present application, and the actual structure is not limited thereto. In general, if a person skilled in the art is inspired by it, without departing from the purpose of the present application, without creative design, similar structure and embodiments of the technical solution can be designed, which should belong to the protection scope of the present application.

Claims

1. A method for evaluating the effect of railway route selection based on artificial intelligence, characterized in that: The method comprises the following steps: Step S1: data preparation, obtaining a railway multi-source data set; Step S2: resilience feature extraction, specifically, based on the railway multi-source data set, a line climate resilience feature extraction based on multi-source space-time resilience is adopted to obtain comprehensive climate resilience features, including the following steps: step S21 climate feature mapping, step S22 structure environment feature modeling, step S23 space-time resilience embedding and step S24 resilience feature fusion; Step S21 climate feature mapping is used to quantify the multi-scale influence mode of climate events, specifically by improving the Transformer model to map the climate environment data to generate climate multi-scale features; the improved Transformer model specifically introduces a multi-scale disturbance gating mechanism into the standard Transformer model to dynamically adjust the response to different climate event intensities; Step S22 structure environment feature modeling is used to build the coupling relationship between line structure and external environment, specifically by building a structure environment bidirectional graph, coupling and fusing structure node features and environment node features through an improved graph convolution network model, and generating structure environment coupling features after node updating; the improved graph convolution network model specifically introduces a dual-domain attention and geological gating constraint into the standard graph convolution network model; The structure environment bidirectional graph is specifically constructed by constructing a line structure graph and a line environment graph, and establishing a bidirectional edge to obtain the structure environment bidirectional graph; The line structure graph is specifically constructed by taking the engineering structure unit in the railway line as a structure node, setting the structure node features, establishing the edges between the structure nodes according to the line topological relationship to form a line adjacency matrix, and obtaining the line structure graph; The line environment graph is specifically constructed by taking the regional environment unit as an environment node, setting the environment node features, and establishing the edges between the environment nodes according to the spatial proximity relationship to obtain the line environment graph; The bidirectional edge is specifically established by establishing a bidirectional edge between the structure node and the environment node when the geographical location of the structure node is within the coverage area of the environment node to obtain the structure environment bidirectional graph; Step S23 space-time resilience embedding, specifically, taking the structure environment coupling features and the line adjacency matrix as inputs, embedding the space-time resilience through a standard space-time graph convolution network model to model the dynamic response of the structure node to climate disturbance at different time and space scales, and obtaining the space-time resilience embedding features; Step S24 resilience feature fusion, specifically, by weighting and fusing the climate multi-scale features, the structure environment coupling features and the space-time resilience embedding features, and standardizing the processing through layer normalization, the comprehensive climate resilience features are obtained to represent the overall resilience capability of the railway line under climate disturbance; Step S3: climate resilience scoring, specifically, based on the railway multi-source data set and the comprehensive climate resilience features, a climate resilience scoring combined with multi-hazard risk and dynamic correction is adopted to obtain the line climate resilience score, including the following steps: step S31 vulnerability loss mapping, step S32 multi-hazard risk aggregation and step S33 line overall evaluation; Step S31 vulnerability loss mapping, for establishing the relationship between climate disaster intensity and line segment performance loss, specifically, by fitting the comprehensive climate resilience characteristics, constructing the structure vulnerability function, calculating the damage probability of the line segment under different disaster intensities, then combining the line operation recovery process, mapping the damage probability to the performance decline loss, obtaining the line segment performance loss index under each disaster; Step S32 multi-disaster risk aggregation, for aggregating the loss influence of each disaster on the same line segment, specifically, by weighting and aggregating the line segment performance loss index under each disaster, and introducing tail risk constraints, obtaining the line segment aggregated risk index; Step S33 line overall evaluation, for generating line-level climate resilience score, specifically, by calculating the risk fluctuation factor through historical disaster simulation, first dynamically correcting the line segment aggregated risk index, then weighting and summarizing, obtaining the line climate resilience score; Step S4: multi-dimensional effect evaluation. 2.The railway route selection effect evaluation method based on artificial intelligence according to claim 1, characterized in that: In step S21, the multi-scale perturbation gating mechanism, specifically, by extracting climate disturbance response features through multi-scale convolution of climate environment data, and calculating the perturbation gating weight corresponding to each scale, weighting and fusing the multi-scale climate disturbance response features according to the perturbation gating weight, obtaining the climate multi-scale features; In step S22, the dual-domain attention, specifically, by using self-attention mechanism to calculate structure domain attention weight and environment domain attention coefficient respectively, for adaptively adjusting the contribution size of different types of neighbors to node feature update. 3.The railway route selection effect evaluation method based on artificial intelligence according to claim 2, characterized in that: In step S4, the multi-dimensional effect evaluation, specifically, combining the line climate resilience score with multi-dimensional key evaluation indexes, and using multi-objective optimization method to comprehensively score and sort the railway route selection scheme, obtaining the railway route selection effect evaluation report.

4. The railway route selection effect evaluation method based on artificial intelligence according to claim 3, characterized in that: In step S1, the data preparation, specifically, collecting railway line structure data, climate environment data and railway operation data, and performing dimension normalization, time synchronization, space matching and unified storage management, obtaining the railway multi-source data set.

5. The railway route selection effect evaluation system based on artificial intelligence, for realizing the railway route selection effect evaluation method based on artificial intelligence as claimed in any one of claims 1-4, characterized in that: It includes a data construction module, a feature construction module, a resilience scoring module and a route selection effect evaluation module. 6.The railway route selection effect evaluation system based on artificial intelligence according to claim 5, characterized in that: The data construction module is used for data preparation, and through data preparation, a railway multi-source data set is obtained, and the railway multi-source data set is sent to the feature construction module and the resilience scoring module; The feature construction module is used for resilience feature extraction, and through resilience feature extraction, comprehensive climate resilience characteristics are obtained, and the comprehensive climate resilience characteristics are sent to the resilience scoring module; The resilience scoring module is used for climate resilience scoring, and through climate resilience scoring, a line climate resilience score is obtained, and the line climate resilience score is sent to the route selection effect evaluation module; The route selection effect evaluation module is used for multi-dimensional effect evaluation, and through multi-dimensional effect evaluation, a railway route selection effect evaluation report is obtained.

Images