A Precise Analysis Method and System for Urban Physical Examinations Based on AI Multimodal Data Collaboration

Abstract

This invention relates to the field of data analysis technology, specifically to a precise analysis method and system for urban health checks based on AI-driven multimodal data collaboration. The method includes the following steps: constructing an adaptive fractal spatiotemporal grid to map urban multimodal data sources into spatiotemporal encoded vectors with fractal dimensions; establishing a physics-driven co-resonance network to extract abnormal co-resonance patterns exceeding normal resonance thresholds; calculating modal entropy chain values ​​based on the real-time metabolic rate of the urban system and generating an optimal weight matrix using a non-equilibrium thermodynamic model; and fusing abnormal co-resonance patterns and the dynamic weight matrix to generate a spatiotemporal causal graph displaying the origin and propagation path of faults using a causal discovery algorithm. This invention, utilizing multimodal data collaboration and causal analysis technology, enhances the emergency response capability of urban systems in the face of sudden faults, ensuring the stability of urban operations. The spatiotemporal causal graph provides comprehensive visualization information for the fault propagation process.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, and in particular to a method and system for precise analysis of urban health checkups based on AI-driven multimodal data collaboration. Background Technology

[0002] With the acceleration of urbanization, the complexity of urban infrastructure and the dynamic nature of its operating environment are gradually increasing, posing challenges to urban management and infrastructure maintenance. Traditional urban system health monitoring and problem point diagnosis are limited to the status assessment of a specific area or equipment, lacking a holistic and cross-domain comprehensive analysis of the entire urban system. This approach often has the following shortcomings:

[0003] During operation, the health status of various infrastructures in an urban system is affected by a variety of factors, including physical characteristics such as vibration spectrum, pressure waves, and thermal radiation. These characteristics often manifest in different forms, and data from a single mode cannot fully reflect the operating status of the system.

[0004] Existing fault diagnosis systems are mainly based on static data analysis models, which ignore the complex coupling relationships between infrastructures and fault propagation paths. Even when a fault occurs, the diagnosis system often has difficulty quickly identifying the fault source and propagation path, resulting in slow response during the handling process, which in turn increases repair costs and risks.

[0005] Traditional urban operation and maintenance management systems typically adopt a static operation and maintenance model, lacking real-time performance and flexibility. They cannot dynamically adjust operation and maintenance strategies based on the operational status of urban infrastructure and changes in the environment. Urban systems may face different risks and loads at different times, which requires operation and maintenance management to have adaptive capabilities and monitor and predict potential failure risks in real time. Summary of the Invention

[0006] This invention provides a method and system for precise analysis of urban physical examinations based on AI-driven multimodal data collaboration.

[0007] A precise analysis method for urban health checkups based on AI-powered multimodal data collaboration includes the following steps:

[0008] S1, Dynamic Fractal Spatiotemporal Coding: Construct an adaptive fractal spatiotemporal grid to map urban multimodal data sources into spatiotemporal coding vectors with fractal dimensions, generating a unified cross-modal representation. Urban multimodal data sources include urban infrastructure vibration spectrum, underground pipe network pressure wave, and surface thermal radiation data.

[0009] S2, Multimodal Co-resonance Detection: Establish a physical field-driven co-resonance network, calculate the data resonance intensity through the intermodal energy transfer equation, and extract abnormal co-resonance modes that exceed the normal resonance threshold;

[0010] S3, Dynamic weighting of metabolic entropy chain: The modal entropy chain value is calculated based on the real-time metabolic rate of the urban system, and the optimal weight matrix is ​​generated through a non-equilibrium thermodynamic model;

[0011] S4. Construction of Causal Heterogeneous Graph: Integrating abnormal collaborative patterns and dynamic weight matrices, a spatiotemporal causal graph that displays the origin and propagation path of faults is generated using a causal discovery algorithm.

[0012] Optionally, S1 specifically includes:

[0013] S11, Dynamic Calculation of Fractal Dimension: Based on the spatiotemporal density distribution of multimodal data sources, the fractal dimension is calculated in real time to describe the self-similarity of data in space and time. The fractal dimension reflects the complexity and structure of the data point distribution. The distribution of the grid is dynamically adjusted by analyzing the density and spatial resolution of the data. The density of the data points determines the calculation of the fractal dimension. The minimum spatial resolution is used to determine the accuracy and scale of the grid, allowing the model to adapt to different data sources and handle data point densities at different scales.

[0014] S12, Fractal Mesh Generation: When generating fractal meshes, the urban geographic coordinate system is used as the basis to construct a fractal mesh structure with self-similar properties. The size of the mesh cells is adjusted according to the fractal dimension, the mesh side length is scaled according to the self-similar characteristics of different data sources, and the network structure is divided into meshes according to the characteristics of different regions.

[0015] S13, Multimodal Data Mapping: In the data mapping phase, different processing is applied to different types of urban data sources to adapt to their unique physical characteristics, specifically including:

[0016] For infrastructure vibration spectrum data, the spatiotemporal distribution characteristics of vibration are characterized by extracting the energy value of its main frequency band. A time decay weighting function is applied to enhance the timeliness of recent data. The data weighted by the time decay weighting function is mapped to the tangential component of the grid to reflect the spatial diffusion characteristics of vibration.

[0017] For underground pipe network pressure wave data, the spatiotemporal convolution value of the pressure gradient is calculated to obtain the variation pattern of the pressure wave in time and space. The Gaussian smoothing method is used to remove noise in the data to ensure the smoothness of the pressure data. The processed pressure data is mapped to the radial component of the grid to reflect the transmission characteristics of underground pressure waves.

[0018] For surface thermal radiation data, high-frequency anomalous temperature variations are extracted using wavelet packet decomposition. These anomalous temperature variations are then supplemented using spatial interpolation to obtain anomalous temperature components, thereby providing the data with more detailed spatial resolution. These anomalous temperature components are mapped to the axial components of the grid to capture the transmission characteristics of thermal radiation.

[0019] S14, Cross-modal fusion coding: This method fuses the tangential components of vibration data, the radial components of pressure wave data, and the axial components of thermal radiation data according to rules. This includes using nonlinear fusion methods to combine the components into a unified spatiotemporal coding vector. During fusion, the relative importance of different modal data sources is considered, and the contribution of each mode is dynamically adjusted using a trainable weight matrix. This fusion method not only ensures that multi-dimensional information of the data is preserved but also constrains the range of the final output vector through nonlinear functions, avoiding excessive growth or explosion of values.

[0020] Optionally, S2 specifically includes:

[0021] S21, Physical Field Coupling Modeling: Using the spatiotemporal encoded vector generated by S1, it is transformed into a multimodal physical field coupling tensor. The construction of the multimodal physical field coupling tensor is based on a three-dimensional structure, where each dimension represents different fractal mesh units, modal types, and spatiotemporal features. Information from different modal data sources is integrated into a unified physical field representation, including mapping the spatiotemporal features extracted from multiple data sources into a high-dimensional tensor structure.

[0022] S22, Energy Transfer Equation Construction: Construct an equation describing the resonant energy transfer between modes. This equation includes the coupling coefficient between modes, spatiotemporal convolution operation, and energy transfer characteristics. The energy transfer between multimodal data sources depends on spatiotemporal characteristics, spatial coupling relationship, energy propagation speed, and the influence of physical properties. It also introduces spatial smoothing under the action of the Laplacian operator and the nonlinear characteristics of the Hadamard product to describe the propagation and interaction of energy between multimodal data sources in spatiotemporal space.

[0023] S23, Dynamic Resonance Threshold Calculation: Set a dynamic threshold to determine which resonance events belong to anomalous multimodal data sources, including training based on historical normal data, calculating the energy mean and standard deviation of each pair of modal data sources under the baseline state, and marking energy events exceeding the standard deviation range as anomalous by using the set dynamic threshold.

[0024] S24. Abnormal Co-operation Mode Extraction: When the resonant energy of a certain mode pair exceeds the set dynamic threshold, the abnormal co-operation mode extraction process will be triggered.

[0025] Optionally, the abnormal collaboration pattern extraction process includes:

[0026] Tracing back the resonance source: Tracing back the resonance source along the spatiotemporal convolution kernel path, tracking the origin and propagation trajectory of energy, and drawing an energy propagation path diagram by tracing back the path;

[0027] Calculate the energy gradient and mark high-risk impact regions: During the backtracking process, the resonant energy gradient is calculated. The resonant energy gradient represents the rate and direction of energy change. Regions where the gradient changes abruptly are marked as "high-risk impact regions".

[0028] Multimodal resonance events are fused together to generate a three-dimensional anomaly co-location map, which includes the spatiotemporal coordinates, energy intensity, and energy conduction direction of each resonance event.

[0029] Optionally, S3 specifically includes:

[0030] S31, Multimodal metabolic flow modeling: Based on the extracted abnormal collaborative patterns, a metabolic flow tensor of an urban system is constructed. The metabolic flow tensor dimensions include multiple fractal grid cells, time slices, and different types of metabolic channels, including energy, matter, and information. Different types of metabolic flows are organized and quantified in a spatiotemporal grid.

[0031] S32, Dynamic calculation of entropy yield: Based on the theory of non-equilibrium thermodynamics, the entropy chain value is calculated for each metabolic mode. The entropy chain value calculation includes metabolic power and metabolic flux. By integrating the dynamic process of the metabolic system, the entropy yield in the future time is evaluated. The entropy chain value reflects the non-equilibrium state of metabolism.

[0032] S33, Weight gradient field generation: Based on the calculation results of the entropy chain value, the weight gradient field is generated by constructing a weight allocation differential equation. The weight allocation is adjusted based on the entropy diffusion coefficient and the benchmark weight regression factor. The cross-modal competition relationship is modeled through coupling parameters to adjust the weight distribution of each metabolic channel.

[0033] S34, Dynamic Matrix Optimization: The weight matrix is ​​optimized by entropy flow constraints, taking into account the diagnostic sensitivity and regularization terms of each mode, to obtain the optimal weight matrix of metabolic flow.

[0034] Optionally, the metabolic flux tensor is represented as: ,in, The number of fractal mesh elements. The number of time slices. Metabolic channel types include three types of metabolic flows: energy, matter, and information. The tangential / radial / axial components of metabolic channel types correspond as follows: energy → vibrational spectrum, matter → pressure wave, and information → thermal radiation.

[0035] Optionally, the dynamic calculation of entropy yield includes modal entropy chain values ​​defined according to non-equilibrium thermodynamics:

[0036] ,in, Indicates the first The entropy of a mode (energy, matter, information) as it changes over time in a dynamic process, through metabolic power. and flux Integrate to calculate the entropy chain value. Represents the gradient. For the first Metabolic power of each modality Characteristic relaxation time This represents the actual metabolic flux. As the equilibrium reference flux, Indicates a tiny increment in time. From resonance energy Obtained by Fourier transform.

[0037] Optionally, S4 specifically includes:

[0038] S41, Heterogeneous Data Fusion: The three-dimensional anomaly co-location map is combined with the generated optimal weight matrix. The three-dimensional anomaly co-location map includes information about spatiotemporal coordinates, energy intensity, and conduction direction. The dynamic weight matrix assigns a weight value to each element. The data is then fused into an enhanced causal feature tensor through tensor operations. Specifically, the spatiotemporal coordinates and weight values ​​are combined through tensor product operations to form a spatiotemporal causal map.

[0039] S42, Spatiotemporal Causality Discovery: Constructing a dual-constraint causal structure learning model, which constrains the inference of causal relationships through two constraints: temporal delay and spatial constraints;

[0040] S43, Spatiotemporal Causal Graph Optimization: The optimization phase employs Bayesian structure learning and reinforcement pruning strategies. The Bayesian structure learning uses the Markov chain Monte Carlo method to sample the posterior probability of causal edges and adaptively adjust the sampling step size. The sampling results are used to filter causal edges with a posterior probability higher than 0.89. The reinforcement pruning strategy includes: after filtering causal edges, performing subgraph isomorphism detection on redundant causal chains, and removing pseudo-causal relationships that do not match the abnormal propagation path graph by comparing the causal chain structure.

[0041] S44, Dynamic Graph Generation: The optimized spatiotemporal causal graph includes multiple attributes, including:

[0042] Node attributes: the spatiotemporal code corresponding to each grid cell and the modal weight entropy of the node. It also includes the node's metabolic power, which represents the energy consumed by the node during causal propagation.

[0043] Edge properties: Edge properties include causal strength (reflecting the strength of the causal relationship), time delay (describing the time delay in the occurrence of the causal relationship), and energy conduction efficiency.

[0044] Optionally, the time delay constraint is used to capture the causal relationship between time series. A cross-modal delay time window is adopted. By calculating the variance change of the causal graph within different time windows, it is determined whether the change of a certain node is affected by other nodes.

[0045] The spatial constraints are introduced during the causal reasoning process. The adjacency matrix is ​​defined using the network topology of the fractal grid to ensure that causal edges are generated only between adjacent grid cells, thus limiting the propagation range of causal relationships.

[0046] The AI-based multimodal data collaboration-based urban health checkup precision analysis system is used to implement the aforementioned AI-based multimodal data collaboration-based urban health checkup precision analysis method, and includes the following modules:

[0047] Dynamic fractal spatiotemporal coding module: Based on urban multimodal data sources, an adaptive fractal spatiotemporal grid is constructed, and the multimodal data sources are mapped into spatiotemporal coding vectors with fractal dimensions;

[0048] Multimodal Co-resonance Detection Module: By establishing a physical field-driven co-resonance network, the module calculates the resonance intensity between different modal data using the energy transfer equation between multimodal data sources, and extracts abnormal co-resonance modes that exceed the normal resonance threshold.

[0049] Metabolic entropy chain dynamic weighting module: By monitoring and analyzing the real-time metabolic rate of the urban system, the corresponding modal entropy chain value is calculated, and the optimal dynamic weight matrix is ​​generated using a non-equilibrium thermodynamic model;

[0050] Causal Heterogeneous Graph Construction Module: By integrating abnormal collaborative patterns and dynamic weight matrices from multimodal data, a spatiotemporal causal graph is established using causal discovery algorithms to display the origins and propagation paths of faults in various facilities within the urban system.

[0051] The beneficial effects of this invention are:

[0052] This invention maps multimodal data sources such as urban infrastructure, underground pipe networks, and surface thermal radiation onto a fractal spatiotemporal grid, forming a unified cross-modal representation. This enhances the comprehensive perception capability of various levels of the city. By using the fractal spatiotemporal grid to process urban data, various types of data are transformed into encoded vectors with spatiotemporal characteristics, capturing the features of different time and spatial scales in the urban system. The fractal structure can handle complex and irregular urban data, providing an efficient data representation method and demonstrating the accuracy of urban health checks.

[0053] This invention, by generating a spatiotemporal causal graph of fault origins and propagation paths, can identify the sources and propagation paths of faults in urban systems. Utilizing multimodal data collaboration and causal analysis techniques, it enhances the emergency response capabilities of urban systems in the face of sudden faults, ensuring the stability of urban operations. The spatiotemporal causal graph provides comprehensive and visualized information on the fault propagation process, supporting decision-makers in making precise resource allocation based on the fault propagation paths, risk points, and weak links in the graph. This avoids resource waste caused by the lack of data support in traditional operation and maintenance, improves the operation and maintenance efficiency and effectiveness of urban infrastructure, ensures timely repair of critical areas, and reduces long-term maintenance costs.

[0054] This invention introduces a calculation method for a dynamic weight coefficient matrix and a metabolic entropy chain, which not only enhances the responsiveness of urban systems to dynamic changes, but also identifies key links and potential risk points in urban systems. This breakthrough enables urban management to make real-time adjustments in complex dynamic environments, improving the accuracy of forecasting and emergency dispatch. Attached Figure Description

[0055] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0056] Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention;

[0057] Figure 2 This is a schematic diagram of the system functional modules according to an embodiment of the present invention. Detailed Implementation

[0058] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. It should also be noted that, to make the embodiments more comprehensive, the following embodiments are the best and preferred embodiments, and those skilled in the art can use other alternative methods to implement some well-known technologies; moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0059] It should be noted that the use of terms such as "an embodiment," "an embodiment," "an exemplary embodiment," and "some embodiments" in the specification indicates that the described embodiment may include a specific feature, structure, or characteristic, but not every embodiment necessarily includes that specific feature, structure, or characteristic. Furthermore, when a specific feature, structure, or characteristic is described in connection with an embodiment, implementing such a feature, structure, or characteristic in conjunction with other embodiments (whether explicitly described or not) should be within the knowledge of those skilled in the art.

[0060] Generally, terms can be understood at least partly from their use in context. For example, depending at least partly on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in a singular sense, or a combination of features, structures, or characteristics in a plural sense. Additionally, the term "based on" can be understood not necessarily to convey an exclusive set of factors, but rather, alternatively, depending at least partly on the context, to allow for the presence of other factors that are not necessarily explicitly described.

[0061] like Figure 1 As shown, the precise analysis method for urban physical examinations based on AI multimodal data collaboration includes the following steps:

[0062] S1, Dynamic Fractal Spatiotemporal Coding: Construct an adaptive fractal spatiotemporal grid to map urban multimodal data sources into spatiotemporal coding vectors with fractal dimensions, generating a unified cross-modal representation. Urban multimodal data sources include urban infrastructure vibration spectrum, underground pipe network pressure wave, and surface thermal radiation data.

[0063] S2, Multimodal Co-resonance Detection: Establish a physical field-driven co-resonance network, calculate the data resonance intensity through the intermodal energy transfer equation, and extract abnormal co-resonance modes that exceed the normal resonance threshold;

[0064] S3, Dynamic weighting of metabolic entropy chain: The modal entropy chain value is calculated based on the real-time metabolic rate of the urban system, and the optimal weight matrix is ​​generated through a non-equilibrium thermodynamic model;

[0065] S4. Construction of Causal Heterogeneous Graph: Integrating abnormal collaborative patterns and dynamic weight matrices, a spatiotemporal causal graph that displays the origin and propagation path of faults is generated using a causal discovery algorithm.

[0066] By mapping multimodal data sources such as urban infrastructure vibration spectrum, underground pipe network pressure wave, and surface thermal radiation data to a fractal spatiotemporal grid, a unified cross-modal representation is formed. This fusion method enhances the comprehensive perception capability of various levels of the city (infrastructure, environment, etc.) and forms the basis for accurate analysis. Using a fractal spatiotemporal grid to process urban data transforms various data types into encoded vectors with spatiotemporal characteristics, effectively capturing the features of different temporal and spatial scales in the urban system. The fractal structure can handle complex and irregular urban data, providing an efficient data representation method and demonstrating the accuracy of urban health checks. A physical field-driven co-resonance network is established to analyze energy transfer between modes and detect abnormal co-resonance patterns. The key to this process is utilizing the energy interaction between modes to reveal potential faults or anomalies in the system, accurately identifying hidden dangers in the urban system and achieving early warning. Entropy chain values ​​are calculated using metabolic rates, and a dynamic weight matrix is ​​generated. Considering the dynamic characteristics of the urban system and adjusting the weights according to a non-equilibrium thermodynamic model, it can adapt to different situations in a changing urban environment, ensuring the real-time nature and accuracy of the analysis results. Abnormal patterns are combined with the dynamic weight matrix, and a spatiotemporal causal map is generated through a causal discovery algorithm. Demonstrating the origin and propagation path of a fault helps to accurately pinpoint the source of the problem and provides a scientific basis for fault repair.

[0067] S1 specifically includes:

[0068] S11, Dynamic Calculation of Fractal Dimension: Based on the spatiotemporal density distribution of multimodal data sources, the fractal dimension is calculated in real time to describe the self-similarity of data in space and time. The fractal dimension reflects the complexity and structure of the data point distribution. The distribution of the grid is dynamically adjusted by analyzing the density and spatial resolution of the data. The density of the data points determines the calculation of the fractal dimension. The minimum spatial resolution is used to determine the accuracy and scale of the grid, allowing the model to adapt to different data sources and handle data point densities at different scales.

[0069] S12, Fractal Mesh Generation: When generating fractal meshes, the urban geographic coordinate system is used as the basis to construct a fractal mesh structure with self-similar properties. The size of the mesh cells is adjusted according to the fractal dimension, and the mesh side length is scaled according to the self-similar characteristics of different data sources. An adaptive method is used to ensure that the spatial performance of each data source is optimally matched. In this way, the network structure is divided into meshes according to the characteristics of different regions, so that complex data distributions can be represented more accurately.

[0070] S13, Multimodal Data Mapping: In the data mapping phase, different processing is applied to different types of urban data sources to adapt to their unique physical characteristics, specifically including:

[0071] For infrastructure vibration spectrum data, the spatiotemporal distribution characteristics of vibration are characterized by extracting the energy value of its main frequency band. A time decay weighting function is applied to enhance the timeliness of recent data. The data weighted by the time decay weighting function is mapped to the tangential component of the grid to reflect the spatial diffusion characteristics of vibration.

[0072] For underground pipe network pressure wave data, the spatiotemporal convolution value of the pressure gradient is calculated to obtain the variation pattern of the pressure wave in time and space. The Gaussian smoothing method is used to remove noise in the data to ensure the smoothness of the pressure data. The processed pressure data is mapped to the radial component of the grid to reflect the transmission characteristics of underground pressure waves.

[0073] For surface thermal radiation data, high-frequency anomalous temperature variations are extracted using wavelet packet decomposition. These anomalous temperature variations are then supplemented using spatial interpolation to obtain anomalous temperature components, thereby providing the data with more detailed spatial resolution. These anomalous temperature components are mapped to the axial components of the grid to capture the transmission characteristics of thermal radiation.

[0074] Through these different types of data mapping processes, the model can comprehensively capture various physical phenomena involved in the urban physical examination process, providing rich spatiotemporal data support for subsequent analysis;

[0075] S14, Cross-modal fusion coding: This method fuses the tangential components of vibration data, the radial components of pressure wave data, and the axial components of thermal radiation data according to rules. This includes using nonlinear fusion methods to combine the components into a unified spatiotemporal coding vector. During fusion, the relative importance of different modal data sources is considered, and the contribution of each mode is dynamically adjusted using a trainable weight matrix. This fusion method not only ensures that multi-dimensional information of the data is preserved but also constrains the range of the final output vector through nonlinear functions, avoiding excessive growth or explosion of values.

[0076] In the dynamic calculation of fractal dimension: the fractal dimension is calculated based on the spatiotemporal density distribution of the multimodal data source. , ,in, For data point density, For minimum spatial resolution, The recommended value is 1.2-1.8 (adjust according to the complexity of the data);

[0077] In fractal mesh generation, a fractal mesh with self-similar properties is generated based on the urban geographic coordinate system, and the edge length of the mesh cells is adjusted. Dynamic scaling is performed, and scaling is represented as: , Using the base side length, The adjustment factor is for the data source type, which adjusts the variation in grid cell size. Different data source types have different adjustment factors: vibration data (high frequency): 0.8; thermal radiation data (low frequency): 1.2; underground pipe network data: 1.0.

[0078] In multimodal data mapping: Extract the dominant frequency band energy integral value from infrastructure vibration spectrum data. ,according to After time decay weighting, it is mapped to the mesh tangential component. It is time. This is a reference time point, so it can be set to 0. It is the time decay constant. Use a 72-hour cycle;

[0079] Calculate the pressure gradient from underground pipe network pressure wave data. The spatiotemporal convolution value, through After Gaussian smoothing, it is mapped to the radial components of the mesh. It is a Gaussian smoothing kernel;

[0080] For surface thermal radiation data, wavelet packet decomposition is used to extract high-frequency anomaly components. ,through Spatial interpolation is then mapped to the grid axial components. It is spatial frequency. It refers to spatial resolution;

[0081] In cross-modal fusion coding: the tangential, radial, and axial components of the same mesh element are categorized as follows: . ) are fused into a spatiotemporal encoded vector, where The modal coupling coefficient is recommended to have the following values: vibration data: 0.7; thermal radiation data: 0.3; underground pipe network data: 0.5. For trainable weight matrix, It is a spatiotemporal encoding vector, which uses the hyperbolic tangent function tanh to constrain the vector value range (-1,1) to avoid numerical explosion caused by the difference in dimensions between modes.

[0082] S2 specifically includes:

[0083] S21, Physical Field Coupling Modeling: Using the spatiotemporal encoded vector generated by S1, it is transformed into a multimodal physical field coupling tensor. The construction of the multimodal physical field coupling tensor is based on a three-dimensional structure, where each dimension represents different fractal mesh units, mode types, and spatiotemporal features. Information from different modal data sources is integrated into a unified physical field representation, including mapping the spatiotemporal features extracted from multiple data sources into a high-dimensional tensor structure. This allows for comprehensive modeling of the interactions and resonance phenomena between various modes, laying the foundation for energy transfer analysis.

[0084] S22, Energy Transfer Equation Construction: Construct an equation describing the resonant energy transfer between modes. This equation includes the coupling coefficient between modes, spatiotemporal convolution operation, and energy transfer characteristics. The energy transfer between multimodal data sources depends on spatiotemporal characteristics, spatial coupling relationship, energy propagation speed, and the influence of physical properties. It also introduces spatial smoothing under the action of the Laplacian operator and the nonlinear characteristics of the Hadamard product to describe the propagation and interaction of energy between multimodal data sources in spatiotemporal space.

[0085] S23, Dynamic Resonance Threshold Calculation: Set a dynamic threshold to determine which resonance events belong to anomalous multimodal data sources, including training based on historical normal data, calculating the energy mean and standard deviation of each pair of modal data sources under the baseline state, and marking energy events exceeding the standard deviation range as anomalous by using the set dynamic threshold.

[0086] S24. Abnormal Co-operation Mode Extraction: When the resonant energy of a certain mode pair exceeds the set dynamic threshold, the abnormal co-operation mode extraction process will be triggered.

[0087] The abnormal collaboration pattern extraction process includes:

[0088] Tracing back the resonance source: Tracing back the resonance source along the spatiotemporal convolution kernel path, tracking the origin and propagation trajectory of energy, and drawing an energy propagation path diagram by tracing back the path;

[0089] Calculate the energy gradient and mark high-risk impact regions: During the backtracking process, the resonant energy gradient is calculated. The resonant energy gradient represents the rate and direction of energy change. Regions where the gradient changes abruptly are marked as "high-risk impact regions".

[0090] Multimodal resonance events are fused together to generate a three-dimensional anomaly co-location map, which includes the spatiotemporal coordinates, energy intensity, and energy conduction direction of each resonance event.

[0091] Physics-field coupling modeling is based on the spatiotemporal encoded vectors generated by S1 to construct a multimodal physics-field coupling tensor. ,in, The number of fractal mesh elements. For the number of modal types, For spatiotemporal feature dimensions, multimodal physical field coupling tensor It is directly composed of the spatiotemporal encoding vectors of S1 stacked together;

[0092] The construction of the energy transfer equation includes defining the intermodal resonance energy equation:

[0093] ,in, Representing mode pairs At any moment energy, The modal coupling coefficient matrix is... Indicates the index of coupled submodes or subsystems between modes. Used to represent the coupling relationship between multiple sub-patterns or subsystems. This represents the spatiotemporal convolution kernel operation. For the Laplace operator, For Hadama accumulation, For cross-modal delay time windows, spatiotemporal convolution kernels Size adaptation dynamically scales side length L (core size = ⌈3L⌉), delay time window ,in, The effective velocity for cross-modal energy transfer (such as the propagation velocity of vibration waves in underground media) is taken when underground pipeline pressure waves are coupled with surface vibration data. Automatically calculate the optimal delay window. This indicates the first mode (one of infrastructure vibration, underground pipeline pressure wave, or surface thermal radiation). This indicates the second mode (one of infrastructure vibration, underground pipeline pressure wave, or surface thermal radiation), and... Pairing refers to the interaction or coupling between two different modes. These are index pairs representing modal pairs, indicating the coupling relationship between different modes;

[0094] The dynamic resonance threshold calculation is based on training a baseline resonance model using historical normal data, according to... Set a dynamic threshold, where Modal pairs The mean and standard deviation of energy under baseline conditions;

[0095] Extraction of abnormal collaboration patterns: when At that time, execute:

[0096] a. Tracing back the resonance source along the spatiotemporal convolution kernel path to generate an anomaly propagation path map. The purpose is to trace back the source of the resonance event based on its propagation trajectory and depict the propagation path from the resonance source to the surrounding area. First, the spatiotemporal convolution kernel is used to analyze the energy transfer between different modes in the data. Through the tracing process, the location of the initial resonance source is determined, which is a relatively concentrated area in the system (such as a pipe rupture, an earthquake source, etc.). Then, these energy transfer paths are traced to generate a map that marks how energy is transferred to other areas from the resonance source, intuitively identifying the anomaly propagation mode and finding the potential range of impact expansion.

[0097] b. Calculate the resonance energy gradient The gradient abrupt change region is marked as a high-risk influence region. The gradient of the resonant energy is calculated. Specifically, this means analyzing the rate of change of energy transfer between different regions. If the rate of change of energy transfer is very fast, it means that the energy fluctuation in a certain part is abrupt, which is usually an abnormal sign. Therefore, the resonant energy gradient, that is, the rate of change of energy, is calculated to detect whether there are abnormal change regions. When regions with abrupt changes in energy change are detected, these regions are marked as "high-risk influence regions", indicating that these regions may be greatly affected and have faults or risks.

[0098] c. By fusing multimodal resonance events, a three-dimensional anomaly co-location map containing spatiotemporal coordinates, energy intensity, and propagation direction is generated. Resonance events from different modes (vibration, pressure wave, thermal radiation) are fused to form a comprehensive three-dimensional map. This map includes not only the spatiotemporal coordinates of each event (i.e., the specific location and time of occurrence), but also the energy intensity (i.e., the severity of the event) and the energy propagation direction (i.e., the direction of propagation of the anomalous event). Through this information, the map can clearly show the overall spatial distribution, intensity, and propagation path of various anomalous events.

[0099] Criteria for determining gradient mutation: ,in, The L2 norm represents the gradient and measures the magnitude of the resonant energy gradient. It is the side length of the fractal mesh unit. It is the fractal dimension;

[0100] A specific implementation example is as follows: When the drainage network pressure is abnormal, i.e., the encoded vector... With road settlement data When resonance occurs in the fractal mesh G0512 element:

[0101] 1. Calculation (threshold) );

[0102] 2. Tracing back along the spatiotemporal convolution kernel revealed that the resonance source was located in an old pipe section 300m upstream;

[0103] 3. Gradient calculations show that the largest abrupt change zone covers 6 surrounding grid cells (4.8m in diameter);

[0104] 4. The three-dimensional anomaly map triggered S4 causal reasoning, which identified the cause as "pipeline corrosion". Soil erosion The transmission chain of "roadbed collapse".

[0105] S3 specifically includes:

[0106] S31, Multimodal metabolic flow modeling: Based on the extracted abnormal collaborative patterns, a metabolic flow tensor of an urban system is constructed. The metabolic flow tensor dimensions include multiple fractal grid cells, time slices, and different types of metabolic channels, including energy, matter, and information. Different types of metabolic flows are organized and quantified in a spatiotemporal grid.

[0107] S32, Dynamic calculation of entropy yield: Based on the theory of non-equilibrium thermodynamics, the entropy chain value is calculated for each metabolic mode. The entropy chain value calculation includes metabolic power and metabolic flux. By integrating the dynamic process of the metabolic system, the entropy yield in the future time is evaluated. The entropy chain value reflects the non-equilibrium state of metabolism.

[0108] S33, Weight Gradient Field Generation: Based on the calculation results of the entropy chain value, the weight gradient field is generated by constructing a weight allocation differential equation. The weight allocation is adjusted based on the entropy diffusion coefficient and the benchmark weight regression factor. The cross-modal competition relationship is modeled through coupling parameters, and the weight distribution of each metabolic channel is adjusted to more accurately reflect the complex dynamic behavior of the urban system.

[0109] S34, Dynamic Matrix Optimization: The weight matrix is ​​optimized by entropy flow constraints, taking into account the diagnostic sensitivity and regularization terms of each mode, to obtain the optimal weight matrix of metabolic flow.

[0110] Multimodal metabolic flow modeling is based on the anomalous cooperative patterns extracted by S2, constructing a metabolic flow tensor for the urban system. The metabolic flow tensor is represented as: ,in, The number of fractal mesh elements. The number of time slices. Metabolic channel types include three types of metabolic flows: energy, matter, and information. The tangential / radial / axial components of metabolic channel types correspond as follows: energy → vibrational spectrum, matter → pressure wave, and information → thermal radiation.

[0111] The dynamic calculation of entropy yield includes modal entropy chain values ​​defined according to non-equilibrium thermodynamics:

[0112] ,in, Indicates the first The entropy of a mode (energy, matter, information) as it changes over time in a dynamic process, through metabolic power. and flux Integrate to calculate the entropy chain value. Represents the gradient. For the first Metabolic power of each modality The characteristic relaxation time is... According to fractal dimension according to Dynamic adjustment This represents the actual metabolic flux. As the equilibrium reference flux, Obtained from historical normal data during training. Indicates a tiny increment in time. From resonance energy Obtained through Fourier transform, the specific derivation is as follows:

[0113] 1. Modal energy aggregation: For each mode Aggregate it with all other modalities Resonance energy:

[0114] ,in This is the spatiotemporal integration domain, representing the spatiotemporal range of the anomaly propagation path;

[0115] 2. Time-frequency domain conversion: for Perform a short-time Fourier transform to extract the frequency domain features of the energy distribution:

[0116] ,in For the Hanning window function, The sliding step size of the time window;

[0117] 3. Metabolic power calculation: Identifying dominant frequencies in the frequency domain. Calculate its corresponding power spectral density: , The time slice length of the fractal grid is set to ensure alignment with the spatiotemporal encoding.

[0118] Construct the weight assignment differential equation to generate the weight gradient field:

[0119] ;in, Indicates the first The entropy of a mode as it changes over time during a dynamic process. The entropy diffusion coefficient is... As the benchmark weighted regression factor, These are cross-modal competition parameters, representing the modes. With mode The competitive relationship between them affects the weight update. From the modal coupling coefficient pass Dynamic association, For the first The weights of each modality For the first The baseline weights for each modality represent the reference weights for each modality. The Laplace operator represents the second-order gradient operator, used to describe the curvature of a function in space, representing the spatial diffusion of entropy chain values ​​and reflecting the rate of entropy diffusion.

[0120] In dynamic matrix optimization, the optimal weight matrix is ​​solved using the optimal transport model under entropy flow constraints. :

[0121] ,in, For each modality diagnostic sensitivity vector, For the total variational regularization term The sparsity coefficient is . This is the optimal weight matrix.

[0122] Specific implementation examples are as follows:

[0123] When heavy rain causes the underground drainage system to become overloaded, i.e., the metabolic flux of matter... During a surge:

[0124] 1. Calculate the modal dechaining values ​​of the material. (Base value 1.3);

[0125] 2. Obtained through the weighted differential equation ;

[0126] 3. The optimal transport model shifts the weights of the drainage system from... Upgraded to The causal graph of S4 is used to analyze the key nodes of the pipeline network.

[0127] S4 specifically includes:

[0128] S41, Heterogeneous Data Fusion: The three-dimensional anomaly co-location map is combined with the generated optimal weight matrix. The three-dimensional anomaly co-location map includes information about spatiotemporal coordinates, energy intensity, and conduction direction. The dynamic weight matrix assigns a weight value to each element. The data is then fused into an enhanced causal feature tensor through tensor operations. Specifically, the spatiotemporal coordinates and weight values ​​are combined through tensor product operations to form a spatiotemporal causal map.

[0129] S42, Spatiotemporal Causality Discovery: Constructing a dual-constraint causal structure learning model, which constrains the inference of causal relationships through two constraints: temporal delay and spatial constraints;

[0130] S43, Spatiotemporal Causal Graph Optimization: The optimization phase employs Bayesian structure learning and reinforcement pruning strategies. Bayesian structure learning uses the Markov chain Monte Carlo method to sample the posterior probability of causal edges and adaptively adjust the sampling step size. The sampling results are used to filter causal edges with a posterior probability higher than 0.89. The reinforcement pruning strategy includes: after filtering causal edges, performing subgraph isomorphism detection on redundant causal chains. By comparing the causal chain structure, pseudo-causal relationships that do not match the abnormal propagation path graph are removed, and false or irrelevant causal relationships are eliminated to ensure that the optimized spatiotemporal causal graph contains only valid and reliable causal inference paths.

[0131] S44, Dynamic Graph Generation: The optimized spatiotemporal causal graph includes multiple attributes, including:

[0132] Node attributes: the spatiotemporal code corresponding to each grid cell and the modal weight entropy of the node. It also includes the node's metabolic power, which represents the energy consumed by the node during causal propagation.

[0133] Edge properties: Edge properties include causal strength (reflecting the strength of the causal relationship), time delay (describing the time delay in the occurrence of the causal relationship), and energy conduction efficiency (used to evaluate the efficiency of information or energy propagation between different nodes).

[0134] Delay constraints are used to capture causal relationships between time series. A cross-modal delay time window is adopted. By calculating the variance change of the causal graph within different time windows, it is determined whether the change of a certain node is affected by other nodes. Based on the classic Granger causality detection principle, a delay parameter is introduced to enable causal inference to adaptively handle causal relationships under different time intervals.

[0135] Spatial constraints are introduced in the causal reasoning process. By using the network topology of fractal grids to define the adjacency matrix, it is ensured that causal edges are generated only between adjacent grid cells, thus limiting the propagation range of causal relationships.

[0136] The three-dimensional anomaly co-location map is represented as follows: (Spatiotemporal coordinates, energy intensity, conduction direction), and combine them with the dynamic weight matrix. Perform tensor fusion to construct an enhanced causal feature tensor:

[0137] ,in, For enhanced causal feature tensors, Represents the tensor product. For diagonalization operation, As a gradient enhancement factor, To smooth out constants and prevent division by zero errors in logarithmic operations, This indicates element-wise multiplication. The gradient of the three-dimensional anomaly co-plot represents its rate of change;

[0138] In spatiotemporal causal discovery, a causal structure learning model with dual constraints is established as follows:

[0139] a. Delay Constraints: Based on Cross-Modal Delay Time Window Construct a time-delay Granger causal detector:

[0140] , Indicates node-based The variance calculated from historical data. Indicates node-based Historical data and nodes The variance calculated from the lagged data;

[0141] b. Spatial Constraints: Utilizing the network topology of fractal meshes, define a spatial adjacency matrix. 0 indicates non-adjacent, 1 indicates adjacent, restricting causal edges to be generated only between adjacent grid cells;

[0142] Optimize the causal graph by employing Bayesian structure learning and reinforcement pruning strategies:

[0143] Using NUTS (No-U-TurnSampler) to measure the posterior probability of causal edges Perform MCMC sampling and retain For strong causal edges, the NUTS Markov Chain Monte Carlo (MCMC) method adaptively adjusts the step size during sampling, avoiding the "U-turn" (chain looping) in traditional MCMC methods. NUTS samples the posterior probability space of the causal edge, first setting an initial state and selecting a candidate causal edge from the causal graph. That is, from the node To the node The causal relationship is determined using the NUTS algorithm from the causal edges. posterior probability distribution During MCMC sampling, NUTS automatically selects an appropriate step size and explores the posterior space based on the current samples. After multiple samplings, the NUTS algorithm adaptively adjusts the sampling process to ensure convergence to the posterior distribution of the causal edges. Finally, it retains all posterior probabilities. The causal edges are considered to have significant causal relationships, which can effectively filter out those edges with high posterior probability (i.e. strong causal relationships) and optimize the causal graph.

[0144] Subgraph isomorphism detection is performed on redundant causal chains to remove pseudo-causal relationships that do not match the abnormal propagation path graph. Subgraph isomorphism detection is a process used to identify identical or similar structures in a graph, especially to identify redundant causal chains. In a causal graph, redundant causal chains can generate pseudo-causal relationships that cannot be matched in the abnormal propagation path graph. Therefore, it is necessary to extract redundant causal chains. By analyzing the redundant causal chains in the causal graph (i.e., those paths that are repeated multiple times or have the same causal structure), potential pseudo-causal relationships can be identified. Subgraph isomorphism detection is then performed on the redundant causal chains to check whether they match the paths in the abnormal propagation path graph. By comparing the causal structure of the redundant causal chains with that in the abnormal propagation path graph, subgraph isomorphism detection identifies those causal chains that are inconsistent with the abnormal path graph and removes the redundant causal chains that do not match the abnormal propagation path graph, retaining those paths that can effectively reflect the actual causal relationships.

[0145] The optimized causal structure is mapped onto a fractal spatiotemporal grid to generate a causal map containing the following elements:

[0146] Node attributes: Spatiotemporal encoding of grid cells, modal weight entropy Metabolic power ;

[0147] Edge attribute: Causality strength Delay Energy conduction efficiency , Representing nodes respectively and nodes The weight matrix, Represents a node and nodes The efficiency of energy transfer between them;

[0148] Risk Heatmap: Integrates gradient mutation regions and weight gradient fields to generate a risk propagation probability cloud map.

[0149] Specific implementation examples are as follows: When a heating pipeline leaks... When a road collapse occurs:

[0150] 1. Tensor Fusion Computation ;

[0151] 2. Delay Detection (interval value) ), triggering the generation of causal edges;

[0152] 3. NUTS sampling confirmation Preserve the causal relationship;

[0153] 4. The final graph shows:

[0154] Leakage source: fractal grid G0709 (coordinates accurate to 0.6m);

[0155] Main propagation path: G0709 G0710 G0801 (Conduction Efficiency) );

[0156] Risk thermal layer coverage radius: 8.4m (matching the actual subsidence area).

[0157] Spatiotemporal causal graphs can identify the origin of faults in a system and track the path of fault propagation. This allows for the rapid location of the source of the problem and determination of its extent when anomalies occur in urban infrastructure or systems, thereby improving the efficiency and accuracy of fault diagnosis.

[0158] By analyzing the propagation paths in the causal graph, potential high-risk areas and system weaknesses can be identified. Based on the structural characteristics of the graph, potential future failures or abnormal propagation can be predicted, early warnings can be issued, and city managers can take preventive measures to reduce the impact of emergencies on city operations.

[0159] like Figure 2 As shown, the urban health check precision analysis system based on AI multimodal data collaboration is used to implement the aforementioned urban health check precision analysis method based on AI multimodal data collaboration, and includes the following modules:

[0160] Dynamic fractal spatiotemporal coding module: Based on urban multimodal data sources, an adaptive fractal spatiotemporal grid is constructed, and the multimodal data sources are mapped into spatiotemporal coding vectors with fractal dimensions;

[0161] Multimodal Co-resonance Detection Module: By establishing a physical field-driven co-resonance network, the module calculates the resonance intensity between different modal data using the energy transfer equation between multimodal data sources, and extracts abnormal co-resonance modes that exceed the normal resonance threshold.

[0162] Metabolic entropy chain dynamic weighting module: By monitoring and analyzing the real-time metabolic rate of the urban system, the corresponding modal entropy chain value is calculated, and the optimal dynamic weight matrix is ​​generated using a non-equilibrium thermodynamic model;

[0163] Causal Heterogeneous Graph Construction Module: By integrating abnormal collaborative patterns and dynamic weight matrices from multimodal data, a spatiotemporal causal graph is established using causal discovery algorithms to display the origins and propagation paths of faults in various facilities within the urban system.

[0164] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0165] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A precise analysis method for urban physical examinations based on AI multimodal data collaboration, characterized in that: Includes the following steps: S1, Dynamic Fractal Spatiotemporal Coding: Construct an adaptive fractal spatiotemporal grid to map urban multimodal data sources into spatiotemporal coding vectors with fractal dimensions, generating a unified cross-modal representation. Urban multimodal data sources include urban infrastructure vibration spectrum, underground pipe network pressure wave, and surface thermal radiation data. S2, Multimodal Co-resonance Detection: Establish a physical field-driven co-resonance network, calculate the data resonance intensity through the intermodal energy transfer equation, and extract abnormal co-resonance modes that exceed the normal resonance threshold; S3, Dynamic Weighting of Metabolic Entropy Chain: Modal entropy chain values ​​are calculated based on the real-time metabolic rate of the urban system, and the optimal weighting matrix is ​​generated using a non-equilibrium thermodynamic model; specifically including: S31, Multimodal metabolic flow modeling: Based on the extracted abnormal collaborative patterns, a metabolic flow tensor of an urban system is constructed. The metabolic flow tensor dimensions include multiple fractal grid cells, time slices, and different types of metabolic channels, including energy, matter, and information. Different types of metabolic flows are organized and quantified in a spatiotemporal grid. S32, Dynamic calculation of entropy yield: Based on the theory of non-equilibrium thermodynamics, the entropy chain value is calculated for each metabolic mode. The entropy chain value calculation includes metabolic power and metabolic flux. By integrating the dynamic process of the metabolic system, the entropy yield in the future time is evaluated. The entropy chain value reflects the non-equilibrium state of metabolism. S33, Weight gradient field generation: Based on the calculation results of the entropy chain value, the weight gradient field is generated by constructing a weight allocation differential equation. The weight allocation is adjusted based on the entropy diffusion coefficient and the benchmark weight regression factor. The cross-modal competition relationship is modeled through coupling parameters to adjust the weight distribution of each metabolic channel. S34, Dynamic Matrix Optimization: The weight matrix is ​​optimized by entropy flow constraints, taking into account the diagnostic sensitivity and regularization terms of each mode, to obtain the optimal weight matrix of metabolic flow. S4. Construction of Causal Heterogeneous Graph: Integrating abnormal collaborative patterns and dynamic weight matrices, a spatiotemporal causal graph that displays the origin and propagation path of faults is generated using a causal discovery algorithm.

2. The method for precise analysis of urban physical examinations based on AI multimodal data collaboration as described in claim 1, characterized in that, S1 specifically includes: S11, Dynamic Calculation of Fractal Dimension: Based on the spatiotemporal density distribution of multimodal data sources, the fractal dimension is calculated in real time to describe the self-similarity of data in space and time. The fractal dimension reflects the complexity and structure of the data point distribution. The distribution of the grid is dynamically adjusted by analyzing the density and spatial resolution of the data. The density of the data points determines the calculation of the fractal dimension. The minimum spatial resolution is used to determine the accuracy and scale of the grid, allowing the model to adapt to different data sources and handle data point densities at different scales. S12, Fractal Mesh Generation: When generating fractal meshes, the urban geographic coordinate system is used as the basis to construct a fractal mesh structure with self-similar properties. The size of the mesh cells is adjusted according to the fractal dimension, the mesh side length is scaled according to the self-similar characteristics of different data sources, and the network structure is divided into meshes according to the characteristics of different regions. S13, Multimodal Data Mapping: In the data mapping phase, different processing is applied to different types of urban data sources to adapt to their unique physical characteristics, specifically including: For infrastructure vibration spectrum data, the spatiotemporal distribution characteristics of vibration are characterized by extracting the energy value of its main frequency band. A time decay weighting function is applied to enhance the timeliness of recent data. The data weighted by the time decay weighting function is mapped to the tangential component of the grid to reflect the spatial diffusion characteristics of vibration. For underground pipe network pressure wave data, the spatiotemporal convolution value of the pressure gradient is calculated to obtain the variation pattern of the pressure wave in time and space. The Gaussian smoothing method is used to remove noise in the data to ensure the smoothness of the pressure data. The processed pressure data is mapped to the radial component of the grid to reflect the transmission characteristics of underground pressure waves. For surface thermal radiation data, high-frequency anomalous temperature changes are extracted by wavelet packet decomposition. The anomalous temperature changes are supplemented by spatial interpolation to obtain anomalous temperature components. The anomalous temperature components are then mapped to the axial components of the grid to capture the transmission characteristics of thermal radiation. S14, Cross-modal fusion coding: The tangential component of vibration data, the radial component of pressure wave data, and the axial component of thermal radiation data are fused according to rules. This includes combining the components into a unified spatiotemporal coding vector through nonlinear fusion methods. During fusion, the relative importance between different modal data sources is considered, and the contribution of each mode is dynamically adjusted through a weight matrix.

3. The method for precise analysis of urban physical examinations based on AI multimodal data collaboration as described in claim 1, characterized in that, S2 specifically includes: S21, Physical Field Coupling Modeling: Using the spatiotemporal encoded vector generated by S1, it is transformed into a multimodal physical field coupling tensor. The construction of the multimodal physical field coupling tensor is based on a three-dimensional structure, where each dimension represents different fractal mesh units, modal types, and spatiotemporal features. Information from different modal data sources is integrated into a unified physical field representation, including mapping the spatiotemporal features extracted from multiple data sources into a high-dimensional tensor structure. S22, Energy Transfer Equation Construction: Construct an equation describing the resonant energy transfer between modes. This equation includes the coupling coefficient between modes, spatiotemporal convolution operation, and energy transfer characteristics. The energy transfer between multimodal data sources depends on spatiotemporal characteristics, spatial coupling relationship, energy propagation speed, and the influence of physical properties. It also introduces spatial smoothing under the action of the Laplacian operator and the nonlinear characteristics of the Hadamard product to describe the propagation and interaction of energy between multimodal data sources in spatiotemporal space. S23, Dynamic Resonance Threshold Calculation: Set a dynamic threshold to determine which resonance events belong to anomalous multimodal data sources, including training based on historical normal data, calculating the energy mean and standard deviation of each pair of modal data sources under the baseline state, and marking energy events exceeding the standard deviation range as anomalous by using the set dynamic threshold. S24. Abnormal Co-operation Mode Extraction: When the resonant energy of a certain mode pair exceeds the set dynamic threshold, the abnormal co-operation mode extraction process will be triggered.

4. The method for precise analysis of urban physical examinations based on AI multimodal data collaboration according to claim 3, characterized in that, The abnormal collaboration pattern extraction process includes: Tracing back the resonance source: Tracing back the resonance source along the spatiotemporal convolution kernel path, tracking the origin and propagation trajectory of energy, and drawing an energy propagation path diagram by tracing back the path; Calculate the energy gradient and mark high-risk influence regions: During the backtracking process, the resonant energy gradient is calculated. The resonant energy gradient represents the rate and direction of energy change. Regions where the gradient changes abruptly are marked as "high-risk influence regions". Multimodal resonance events are fused together to generate a three-dimensional anomaly co-location map, which includes the spatiotemporal coordinates, energy intensity, and energy conduction direction of each resonance event.

5. The method for precise analysis of urban physical examinations based on AI multimodal data collaboration according to claim 1, characterized in that, The metabolic flux tensor is represented as: ,in, The number of fractal mesh elements. The number of time slices. Metabolic channel types include three types of metabolic flows: energy, matter, and information. The tangential / radial / axial components of metabolic channel types correspond as follows: energy → vibrational spectrum, matter → pressure wave, and information → thermal radiation.

6. The method for precise analysis of urban physical examinations based on AI multimodal data collaboration according to claim 1, characterized in that, The dynamic calculation of entropy yield includes modal entropy chain values ​​defined according to non-equilibrium thermodynamics: ,in, Indicates the first The entropy generated by the mode during dynamic processes over time is calculated by integrating metabolic power and metabolic flux to obtain the entropy chain value. Represents the gradient. For the first Metabolic power of each modality For characteristic relaxation time, This represents the actual metabolic flux. As the equilibrium reference flux, Indicates a tiny increment in time. From resonance energy Obtained by Fourier transform.

7. The method for precise analysis of urban physical examinations based on AI multimodal data collaboration according to claim 4, characterized in that, S4 specifically includes: S41, Heterogeneous Data Fusion: The three-dimensional anomaly co-location map is combined with the generated optimal weight matrix. The three-dimensional anomaly co-location map includes information about spatiotemporal coordinates, energy intensity, and conduction direction. The dynamic weight matrix assigns a weight value to each element. The data is then fused into an enhanced causal feature tensor through tensor operations. Specifically, the spatiotemporal coordinates and weight values ​​are combined through tensor product operations to form a spatiotemporal causal map. S42, Spatiotemporal Causal Discovery: Construct a causal structure learning model with dual constraints, using temporal and spatial constraints to constrain the inference of causal relationships; S43, Spatiotemporal Causal Graph Optimization: The optimization phase employs Bayesian structure learning and reinforcement pruning strategies. The Bayesian structure learning uses the Markov chain Monte Carlo method to sample the posterior probability of causal edges and adaptively adjust the sampling step size. The sampling results are used to filter causal edges with a posterior probability higher than 0.

89. The reinforcement pruning strategy includes: after filtering causal edges, performing subgraph isomorphism detection on redundant causal chains, and removing pseudo-causal relationships that do not match the abnormal propagation path graph by comparing the causal chain structure. S44, Dynamic Graph Generation: The optimized spatiotemporal causal graph includes multiple attributes, including: Node attributes: the spatiotemporal code corresponding to each grid cell and the modal weight entropy of the node. It also includes the node's metabolic power, which represents the energy consumed by the node during causal propagation. Edge properties: Edge properties include causal strength, time delay, and energy conduction efficiency.

8. The method for precise analysis of urban physical examinations based on AI multimodal data collaboration as described in claim 7, characterized in that, The time delay constraint is used to capture the causal relationship between time series. It adopts a cross-modal delay time window and determines whether the change of a certain node is affected by other nodes by calculating the variance change of the causal graph within different time windows. The spatial constraints are introduced during the causal reasoning process. The adjacency matrix is ​​defined using the network topology of the fractal grid to ensure that causal edges are generated only between adjacent grid cells, thus limiting the propagation range of causal relationships.

9. A precise urban health checkup analysis system based on AI multimodal data collaboration, used to implement the precise urban health checkup analysis method based on AI multimodal data collaboration as described in any one of claims 1-8, characterized in that, Includes the following modules: Dynamic fractal spatiotemporal coding module: Based on urban multimodal data sources, an adaptive fractal spatiotemporal grid is constructed, and the multimodal data sources are mapped into spatiotemporal coding vectors with fractal dimensions; Multimodal Co-resonance Detection Module: By establishing a physical field-driven co-resonance network, the module calculates the resonance intensity between different modal data using the energy transfer equation between multimodal data sources, and extracts abnormal co-resonance modes that exceed the normal resonance threshold. The metabolic entropy chain dynamic weighting module: By monitoring and analyzing the real-time metabolic rate of the urban system, it calculates the corresponding modal entropy chain values ​​and uses a non-equilibrium thermodynamic model to generate the optimal dynamic weighting matrix; specifically including: Multimodal metabolic flow modeling: Based on the extracted anomalous cooperative patterns, a metabolic flow tensor of an urban system is constructed. The metabolic flow tensor dimensions include multiple fractal grid cells, time slices, and different types of metabolic channels, including energy, matter, and information. Different types of metabolic flows are organized and quantified in a spatiotemporal grid. Entropy yield dynamic calculation: Based on the theory of non-equilibrium thermodynamics, the entropy chain value is calculated for each metabolic mode. The entropy chain value calculation includes metabolic power and metabolic flux. By integrating the dynamic process of the metabolic system, the entropy yield in the future time is evaluated. The entropy chain value reflects the non-equilibrium state of metabolism. Weight gradient field generation: Based on the calculation results of entropy chain value, the weight gradient field is generated by constructing a weight allocation differential equation. The weight allocation is adjusted based on the entropy diffusion coefficient and the benchmark weight regression factor. The cross-modal competition relationship is modeled through coupling parameters to adjust the weight distribution of each metabolic channel. Dynamic matrix optimization: The weight matrix is ​​optimized by entropy flow constraints, taking into account the diagnostic sensitivity and regularization terms of each mode, to obtain the optimal weight matrix of metabolic flow. Causal Heterogeneous Graph Construction Module: By integrating abnormal collaborative patterns and dynamic weight matrices from multimodal data, a spatiotemporal causal graph is established using causal discovery algorithms to display the origins and propagation paths of faults in various facilities within the urban system.

Images