A physical simulation local updating method and device of a city information model

Abstract

The application discloses a physical simulation local updating method and device of a city information model, relates to the technical field of city information models and digital twin simulation, and comprises the following steps: acquiring a reference state, including a reference geometry, a simulation model and a reference physical field; generating a geometry disturbance fingerprint in response to a geometry editing operation, and determining a seed node based on the geometry disturbance fingerprint; determining a target physical field and a pre-constructed physical propagation influence graph of the geometry disturbance fingerprint according to a physical attribute change type of the geometry disturbance fingerprint; propagating an influence score along the graph edges of the physical propagation influence graph from the seed node as a starting point, and including simulation nodes with an influence score higher than a threshold value in an initial disturbance domain; performing local correction reasoning based on the initial disturbance domain and the reference physical field to obtain a local physical field correction amount, and superimposing the local physical field correction amount on the reference physical field to generate an updated physical field. Therefore, the calculation overhead of re-simulation of the whole domain after each local geometry editing can be avoided, and efficient updating can be realized in a local range actually affected.

Description

Technical Field

[0001] This application relates to the field of urban information modeling and digital twin simulation technology, and in particular to a method and apparatus for local updating of physical simulation of urban information models. Background Technology

[0002] In urban planning and digital twin platforms, designers often only perform geometric edits on local buildings or the ground surface (such as adjusting building height, location, or orientation), but need to obtain an updated global physics field to evaluate the rationality of the design. Traditional numerical simulation methods are computationally extremely expensive, with a single solution taking several hours. While existing surrogate models have shortened inference time, their basic paradigm still uses the modified complete urban geometry as the overall input and re-outputs the complete physics field. Even if only one building is modified, the computational scale is still comparable to global inference, making it impossible to effectively reuse the highly reliable baseline physics field obtained before editing.

[0003] Some solutions attempt to incrementally update local regions near geometric changes to reduce computational scale. A common approach is to define this local area using a fixed radius, a fixed number of grid layers, or a fixed number of graph jumps. While these methods are standardized and easy to implement, the degree of matching between the defined perturbation domain size and the actual physical impact of the geometric change often varies significantly depending on the operating conditions, and their applicability and accuracy need improvement. Furthermore, determining the local area often relies on manual experience or repeated trials, making it difficult to achieve a stable and reliable balance between computational efficiency and result accuracy.

[0004] Therefore, how to provide a method and system that can adaptively determine the effective computational domain in the sense of physical conservation based on existing reference physical fields and local geometric editing is a key research topic for those skilled in the art. Summary of the Invention

[0005] In a first aspect, this application provides a method for local updating of physical simulation of a city information model. The method includes: acquiring a baseline state, the baseline state including a baseline geometry of the city information model before editing, a simulation model, and a baseline physical field, the baseline physical field being the distribution information of physical field variables defined on simulation nodes in the simulation model; generating a geometric perturbation fingerprint in response to a geometric editing operation on the baseline geometry of the city information model, the geometric perturbation fingerprint including change information of geometric units in the baseline geometry; determining the affected target physical field based on the type of physical attribute change represented by the geometric perturbation fingerprint, the target physical field corresponding to a pre-constructed physical propagation influence map, the physical propagation influence map including each simulation node in the simulation model and graph edges connecting adjacent simulation nodes, the graph edges being associated with a characterizing the influence strength of the perturbation propagating along the graph edges. The physical propagation influence graph is constructed according to the type of physical field, based on the simulation nodes and preset propagation influence factors, which include the propagation mechanisms of different types of physical fields. Based on the geometric perturbation fingerprint, the nodes in the simulation model that undergo geometric changes are identified, and these nodes are used as seed nodes. Starting from the seed nodes, the propagation influence scores of the simulation nodes are calculated along the edges of the physical propagation influence graph of the target physical field based on the corresponding propagation edge weights. Simulation nodes with propagation influence scores exceeding a preset threshold are included in the initial perturbation domain. Based on the initial perturbation domain and the reference physical field, local correction inference is performed to determine the local physical field correction amount of each simulation node in the initial perturbation domain relative to the reference physical field. Based on the local physical field correction amount and the reference physical field, an updated reference physical field is generated.

[0006] The physical simulation local update method of the urban information model provided in this application has the following advantages: First, by reusing the baseline physical field before editing, the physical field correction is predicted only in the locally affected perturbation domain, effectively avoiding the global calculation of resimulating the entire urban area after each geometric edit.

[0007] On the other hand, this application generates an initial perturbation domain through a propagation mechanism of physical propagation effect maps and propagation effect scores. This ensures that the shape and extent of the perturbation domain match the actual physical effects of geometric changes, rather than relying on pre-defined ranges lacking physical basis, such as fixed radii, fixed mesh layers, or fixed graph jumps. The propagation edge weights are determined based on various factors, including the flow direction of the target physical field in the reference physical field, the spatial distance between simulation nodes, and building occlusion relationships. This allows the same geometric edit to generate perturbation domains of different shapes under different wind directions, terrains, or building layouts, avoiding the problem of fixed ranges being too large or too small under different working conditions. A stable and reliable balance is achieved between computational accuracy and computational efficiency.

[0008] Furthermore, this application determines the affected target physical field by identifying the sub-fingerprint types contained in the geometric perturbation fingerprint, thus achieving automatic mapping between geometric editing and physical field types. Different sub-fingerprints correspond to different physical propagation mechanisms: changes in geometric occupancy affect the wind and thermal fields, changes in material affect the thermal field, changes in permeability affect the urban flooding field, and changes in channel connectivity affect both the wind and urban flooding fields. This allows this application to automatically adapt to various urban physical simulation scenarios, such as wind, thermal, and urban flooding physical fields, without requiring users to manually specify the target physical field type.

[0009] Furthermore, this application directly outputs the correction value through local correction inference instead of re-outputting the absolute physical field. This allows the surrogate model to learn only the incremental impact of local geometric changes on the physical field, while the theoretical correction value for unaffected regions is close to zero. Compared to surrogate models that directly predict the absolute physical field, this application can retain high-frequency details in unchanged regions of the baseline physical field, while reducing the model's learning span and inference difficulty. This is beneficial for improving the prediction accuracy and generalization ability of the surrogate model in local correction scenarios.

[0010] Secondly, this application also provides a physical simulation local update apparatus for a city information model, including a unit for performing any of the physical simulation local update methods for a city information model in the first aspect.

[0011] Thirdly, this application also provides a computer storage medium that can store multiple instructions, which are adapted to be loaded by a processor and executed by the processor to perform a physical simulation local update method for any of the urban information models in the first aspect.

[0012] Fourthly, embodiments of this application also provide a computer program product containing instructions, which, when run on an electronic device, causes the electronic device to execute any one of the physical simulation local update methods for urban information models in the first aspect.

[0013] Fifthly, embodiments of this application also provide a chip module, including a transceiver component and a chip, wherein the chip is used to execute any of the physical simulation local update methods for urban information models in the first aspect.

[0014] In a sixth aspect, embodiments of this application also provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the method described in any one of the first aspects is executed when the processor executes the program.

[0015] It is understood that the physical simulation partial update device for the urban information model, computer storage medium, computer program, computer program product, chip system, and electronic device provided above are all used to execute the method shown in any implementation of the first aspect of the embodiments of this application. Therefore, the beneficial effects that can be achieved can be referred to the beneficial effects in the corresponding methods, and will not be repeated here. Attached Figure Description

[0016] Figure 1 This is a schematic flowchart of a physical simulation local update method for a city information model provided in an embodiment of this application; Figure 2 This is a flowchart illustrating another method for locally updating a physical simulation of a city information model, as provided in an embodiment of this application. Figure 3 This is a schematic flowchart of a method for detecting the boundary closure between the inner and outer narrow bands of a boundary narrow band in an initial perturbation domain, provided by an embodiment of this application. Figure 4 This is a flowchart illustrating another method for locally updating a physical simulation of a city information model, as provided in an embodiment of this application. Figure 5 This is a schematic diagram of a physical simulation local update device for a city information model provided in an embodiment of this application. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described below in conjunction with the accompanying drawings.

[0018] It should be noted that this application embodiment uses an electronic device as an example to illustrate the execution subject of the physical simulation partial update method for the urban information model provided in this application. This electronic device can also be understood as the physical simulation partial update device for the urban information model shown in this application embodiment. In this application embodiment, the electronic device can be a microprocessor or computer for executing program code, etc. Any electronic device that can be used to execute the method provided in this application embodiment falls within the protection scope of this application embodiment, and this application does not impose any limitations. For example, the electronic device can be a desktop computer, a laptop, a mobile terminal, a 32-bit microprocessor, or a 64-bit microprocessor, etc., and this application embodiment does not limit this.

[0019] Example 1:

[0020] Please see Figure 1 , Figure 1 This is a flowchart illustrating a method for local updating of a city information model based on physical simulation, as provided in an embodiment of this application. Figure 1 As shown, the physical simulation local update method for the city information model includes the following steps: S101, the electronic device acquires a reference state, which includes reference geometry, simulation model, and reference physical field.

[0021] The baseline geometry includes the city's 3D model data before editing, such as geometric information like building outlines, heights, roof shapes, terrain elevations, road boundaries, and vegetation distribution.

[0022] The simulation model includes simulation nodes obtained from the discretization of the baseline geometry and the connections between these nodes (i.e., mesh topology or graph topology). Simulation nodes represent discretized spatial units, which can be at least one of the following: mesh elements (such as the center point or vertex of a 3D volume mesh), surface patches (such as patches on a building facade), building blocks (such as the overall representation of a single building), or adaptive sampling points (such as points adaptively distributed according to the physical field gradient). The connections between simulation nodes are spatial adjacency relationships, indicating that two simulation nodes are adjacent in 3D space and that there is a physical path connecting the two nodes.

[0023] The baseline physics field is the distribution information of physical field variables defined on the simulation nodes in the simulation model. Examples include one or more of the velocity, pressure, and turbulent kinetic energy fields in a windy environment; one or more of the temperature and radiation flux fields in a thermal environment; and one or more of the water depth and flow velocity fields in an urban flooding environment. The baseline physics field can be obtained by a high-precision numerical solver (such as a solver based on the finite volume method or finite element method) or by a validated surrogate model. For applications with high safety requirements, high-precision numerical results are preferred as the initial baseline.

[0024] In some possible implementations, the aforementioned baseline state may further include building topographic surface environment data. This building topographic surface environment data is mapped to simulation nodes through a mapping relationship between geometric elements and simulation nodes. The building topographic surface environment data includes at least one of the following: building layout, topographic elevation, surface roughness, underlying surface material, and drainage network connectivity. This data is used for subsequent determination of propagation edge weights.

[0025] In some possible implementations, electronic devices can also assign a uniform state version number (e.g., V0) to the base state, so that each subsequent edit can be clearly traced back to the base state it depends on.

[0026] S102, the electronic device generates a geometric perturbation fingerprint in response to a geometric editing operation on the baseline geometry of the city information model.

[0027] Specifically, the electronic device responds to a geometric editing operation on the baseline geometry of the city information model, acquires the edited geometry, and generates a geometric perturbation fingerprint based on the difference between the baseline geometry and the edited geometry.

[0028] Geometric editing operations include, but are not limited to, one or more of the following: creating, deleting, translating, rotating, adjusting height, creating partial openings, changing roof shapes, raising or lowering terrain, changing road elevations, replacing underlying surface materials, and changing drainage facility connections. Editing operations can be entered by the user through a 3D editor or imported from external programs.

[0029] Electronic devices map the geometric models before and after editing to a unified spatial reference frame, and identify the changed geometric units and their change information through methods such as voxel difference, mesh Boolean difference, surface correspondence, or component property difference. Geometric perturbation fingerprints include change information of geometric units in the baseline geometry.

[0030] For example, a geometrically perturbed fingerprint may include at least one of the following sub-fingerprints: Geometric occupancy fingerprint: used to describe changes in the volume of a geometric unit, such as an increase in building height leading to additional volume at the top, or a decrease in volume due to building demolition.

[0031] Boundary morphology sub-fingerprint: used to describe changes in the surface position and / or normal direction of geometric units, such as changes in the surface position caused by building rotation, or changes in the normal direction caused by changing the roof from a flat roof to a pointed roof.

[0032] Sub-fingerprint of physical properties: used to describe at least one change in the material, roughness, and permeability of a geometric unit, such as changing a concrete wall to a glass curtain wall, or changing grass to paved ground.

[0033] Topological sub-fingerprint: used to describe at least one change in the connection relationship between geometric units, such as channels, roads, drainage pipelines, and open spaces, such as the construction of a new road connecting two previously isolated areas, or the rerouting of drainage pipelines.

[0034] Each perturbation unit records at least the change location, change type, change magnitude, geometric principal direction, surface normal change, physical property difference, and associated topology. For the same edit event, the electronic device can simultaneously generate multiple sub-fingerprints, and each sub-fingerprint collectively determines the seed node and initial direction for subsequent influence propagation.

[0035] S103, the electronic device determines the affected target physical field based on the type of physical property change characterized by the geometric perturbation fingerprint.

[0036] In this application embodiment, the types of physical fields include wind physical fields, thermal physical fields, and urban flooding physical fields. Different types of geometric changes affect different physical fields. The electronic device determines the affected target physical field based on the sub-fingerprint types contained in the geometric perturbation fingerprint, specifically: 1) In the case where the geometric perturbation fingerprint includes the geometric occupancy sub-fingerprint, the target physical field includes the wind physical field and the thermo physical field. This is because the increase or decrease of the entity volume will change the windward area and wake characteristics of the building, thus affecting the wind physical field, and at the same time, it will change the shadow range and radiation shading, thus affecting the thermo physical field.

[0037] 2) When the geometric perturbation fingerprint includes the boundary morphology sub-fingerprint, the target physical field includes both wind and thermophysical fields. This is because changes in surface position and normal will alter the airflow separation point and flow characteristics (affecting the wind physical field), while also changing the shadow projection position and the solar radiation incident angle (affecting the thermophysical field).

[0038] 3) When the geometric perturbation fingerprint includes a physical property sub-fingerprint, the target physical field includes at least one of the following: a thermophysical field, a wind physical field, and an urban flooding physical field. Among them, the material variation in the physical property sub-fingerprint corresponds to the thermophysical field (changes in material thermal properties affect heat exchange), the roughness variation corresponds to the wind physical field (changes in roughness affect airflow friction resistance), and the permeability variation corresponds to the urban flooding physical field (changes in permeability affect the surface runoff infiltration capacity).

[0039] 4) When the geometric perturbation fingerprint includes a topological sub-fingerprint, the target physical field includes at least one of the following: a thermophysical field, a wind physical field, and an urban flooding physical field. Among them, changes in the channel or open space connectivity in the topological sub-fingerprint correspond to the thermophysical field and the wind physical field (changes in channel morphology affect wind and heat advection paths and ventilation corridors), and changes in drainage pipeline connectivity correspond to the urban flooding physical field (changes in drainage connectivity affect confluence paths).

[0040] In this embodiment, the target physical field corresponds to a pre-constructed physical propagation influence map. This physical propagation influence map is constructed according to the type of physical field, based on each simulation node in the simulation model, and preset propagation influence factors. The preset propagation influence factors include the propagation mechanisms of different types of physical fields. For example, the propagation mechanism of wind physical fields is based on airflow transport and wake effects (disturbances mainly propagate along the prevailing wind direction and local velocity direction), the propagation mechanism of thermal physical fields is based on radiative heat transfer and convective heat transfer (disturbances propagate along the visible radiation path and wind-heat advection direction), and the propagation mechanism of urban flooding physical fields is based on surface runoff and drainage network transmission (disturbances propagate along the terrain slope and drainage connectivity direction).

[0041] In this embodiment, the physical propagation impact map is a persistently built and globally reused asset based on physical field type. After determining the target physical field, the electronic device directly obtains the pre-built physical propagation impact map corresponding to that target physical field. This physical propagation impact map includes each simulation node in the simulation model and graph edges connecting adjacent simulation nodes. Each graph edge is associated with a propagation edge weight, which characterizes the strength of the impact of the disturbance propagating along the graph edges.

[0042] For example, the propagation edge weights are determined based on one or more of the following influencing factors: spatial distance between simulation nodes, flow direction of the target physical field in the reference physical field, terrain slope between simulation nodes, building occlusion relationship between simulation nodes, and channel connectivity between simulation nodes.

[0043] The propagation weights can be a weighted combination or product of various influencing factors. Among them, the spatial distance factor between simulation nodes can be used as the basic weight. Based on the principle that disturbances attenuate with increasing spatial distance and that propagation ability is stronger the closer the distance, the basic weights can be calculated based on exponential or inverse decay functions, making the weights larger for closer nodes and smaller for farther nodes. Physical influencing factors include one or more of the following: the flow direction factor of the target physical field in the baseline physical field, the building shading factor, the terrain aspect factor, and the channel connectivity factor.

[0044] The weight of the flow direction factor is based on the principle that the propagation ability of disturbances along the dominant propagation direction of physical quantities is stronger than that in the opposite direction, and the weight of the edge weight in the direction perpendicular to the dominant flow direction is centered. It can be calculated using the direction cosine function. The dot product operation is performed between the direction vectors of the nodes and the direction vector of the dominant flow direction of physical quantities at the nodes. The larger the dot product value, the more consistent the direction is with the dominant flow direction, and the larger the edge weight is; when the dot product value is zero, the edge weight is centered; when the dot product value is negative, the edge weight decreases. Regarding the dominant propagation direction, it is the local wind speed direction in a wind field, the water flow direction in urban flooding, and the heat flow direction in a thermal field.

[0045] The weight of the building shading factor is determined by a binary judgment function or transmission coefficient based on the principle that physical buildings block the propagation path of disturbances. If the connection between nodes is completely blocked by the building entity, the edge weight is reduced to zero; if it is not blocked, the original weight is maintained. For partial shading such as open floors, windows or building gaps, the transmission coefficient is set based on the porosity or opening area ratio of the shading object, and the value is between zero and one.

[0046] The weights of the terrain aspect factor are determined using an elevation difference normalization function, based on the principle that the propagation ability of disturbances along the downward direction of the terrain is stronger than that along the upward direction of the terrain. The larger the elevation difference between nodes and the closer it is to the downward direction, the larger the edge weight; conversely, the edge weight in the upward direction is reduced.

[0047] The channel connectivity factor weight is determined using a connectivity indicator function based on the principle that connectivity paths such as roads, channels, drainage pipelines, or open spaces enhance the directional propagation capability of disturbances. If such connectivity paths exist between nodes, the edge weights receive an additional bonus; otherwise, no bonus is applied.

[0048] S104, the electronic device uses geometric perturbation fingerprints to identify the nodes in the simulation model that have undergone geometric changes, and uses these nodes as seed nodes.

[0049] Specifically, the electronic device queries the geometric units corresponding to each simulation node in the simulation model based on the spatial location contained in the change information of each geometric unit in the geometric perturbation fingerprint, determines the simulation node corresponding to the changed geometric unit as the changed node, and uses these changed nodes as seed nodes.

[0050] S105, the electronic device starts from the seed node and calculates the propagation influence score of the simulation node along the graph edge of the physical propagation influence graph of the target physical field based on the corresponding propagation edge weight. Simulation nodes with propagation influence scores exceeding a preset threshold are included in the initial disturbance domain.

[0051] Specifically, the electronic device assigns initial influence scores to various child nodes (the seed node has the highest initial influence score). Starting from each child node, the propagation influence score is calculated layer by layer outward along the edges of the physical propagation influence graph, based on the initial influence score and the edge weights. During disturbance propagation, a simulation node may receive influence scores from multiple propagation paths simultaneously. The electronic device accumulates the influence scores from different propagation paths as the cumulative propagation influence score for that simulation node. The influence score decays with the size of the propagation edge weight; the larger the propagation edge weight (i.e., the stronger the disturbance propagation capability in that direction), the slower the influence score decays; the smaller the propagation edge weight, the faster the influence score decays. When the cumulative influence score of a simulation node exceeds a preset initial inclusion threshold, that simulation node is included in the initial disturbance domain.

[0052] In the embodiments of this application, the propagation edge weights are dynamically determined based on the propagation mechanism of the physical field and the flow direction information in the reference physical field. The initial perturbation domain exhibits an asymmetric, non-convex, and direction-dependent shape, rather than a sphere with a fixed radius or a subgraph with a fixed number of jumps.

[0053] In some possible implementations, the electronic device also includes the simulated nodes within a preset neighborhood of the seed node in the initial perturbation domain as a minimum safe neighborhood protection, ensuring that nodes near geometric changes are not missed due to threshold determination.

[0054] S106, the electronic device performs local correction inference based on the initial perturbation domain and the reference physical field to determine the local physical field correction amount of each simulation node in the initial perturbation domain relative to the reference physical field.

[0055] In some possible implementations, when performing local correction inference, the electronic device first divides the region on the simulation model corresponding to the initial perturbation domain into a core region, a computation region, and a boundary narrow band. The core region contains the simulation nodes that undergo geometric changes; the computation region encompasses the core region and extends outward to the boundary of the initial perturbation domain; the boundary narrow band includes an inner narrow band and an outer narrow band. The inner narrow band includes several layers of simulation nodes inside the edge of the computation region, and the outer narrow band includes several layers of simulation nodes outside the edge of the computation region.

[0056] The electronic device performs local correction inference on a computational region within an initial perturbation domain, based on a reference physics field. The inputs to the local correction inference can include new geometric data within the perturbation domain, the perturbation fingerprint, the reference physics field, the reference boundary flux, external boundary conditions, and physical states inherited from the reference physics field at the boundary of the perturbation domain. Local correction inference can be performed using a surrogate model, which can be a graph neural network, a geometric neural operator, a Fourier neural operator, a deep operator network, a convolutional network, a reduced-order basis model, or a combination thereof.

[0057] In the real-time stream of this application, the surrogate model for local correction inference outputs the physical field correction (i.e., the difference field) of each simulation node within the computational domain relative to the reference physical field, rather than re-outputting the absolute physical field of the entire region. Since the prediction target is the correction amount, the theoretical correction value of the unperturbed region is close to zero, which allows the model to retain the high-frequency details that have not changed in the reference physical field.

[0058] During the corrective inference process, the electronic device applies different corrective constraints to simulation nodes in different regions of the computation area: For simulation nodes within the core region of the computational domain, the maximum correction value for the physics field is the first magnitude. The core region is where geometric changes occur directly, and the physics field changes are most significant, allowing for the largest possible correction value.

[0059] For simulation nodes (i.e., the intermediate region) located outside the core region but within the inner boundary of the inner narrow band (the inner boundary of the inner narrow band refers to the boundary of the inner narrow band closer to the core region), the maximum correction value of the physics field is the second amplitude, which is smaller than the first amplitude. The intermediate region is a transitional region affected by disturbance propagation, and the correction value gradually decreases as it moves away from the core region.

[0060] For simulation nodes within the inner narrow band of the computational domain, the maximum correction value for the physics field is the third amplitude, which is smaller than the second amplitude. The inner narrow band is the outermost layer of the computational domain, and the correction value here should have converged to near zero to achieve a natural connection with the external reference physics field.

[0061] S107, the electronic device generates an updated reference physical field based on the local physical field correction and the reference physical field.

[0062] Specifically, the electronic device first superimposes the local physics field correction onto the physical quantities of the corresponding simulation nodes in the reference physics field to obtain the updated physical quantity values ​​of each simulation node within the initial perturbation domain. For simulation nodes located outside the initial perturbation domain in the simulation model, their physical quantity values ​​are directly adopted from the corresponding values ​​in the reference physics field. Then, the electronic device merges the updated physical quantity values ​​within the initial perturbation domain with the reused reference physical quantity values ​​outside the initial perturbation domain to form a complete updated physics field. This updated physics field, together with the edited geometry, constitutes a new reference state, used for incremental updates in subsequent geometry editing operations.

[0063] The physical simulation local update method for the urban information model provided in this application directly reuses the baseline physical field and intermediate states before editing, predicting corrections only within the actually affected perturbation domain, which can significantly reduce the computational scale after local building adjustments. Simultaneously, the perturbation domain is jointly generated by the geometric perturbation fingerprint and the physical propagation influence map, and adaptively expands based on the boundary mismatch direction, avoiding the problem of the range being too small or too large due to fixed radii or fixed hop numbers. By detecting state continuity, normal flux, equation residuals, and perturbation attenuation before allowing stitching, and performing conservation corrections in narrowband, non-physical jumps and spurious sources and sinks caused by simple interpolation are reduced.

[0064] Furthermore, versioned state caching allows for the reuse of the most recent closure result, supporting incremental accumulation during multiple rounds of scheme adjustments. The core closure self-expansion mechanism can be combined with graph neural networks, neural operators, reduced-order models, or local numerical solvers, making it widely applicable.

[0065] In some other possible implementations, after determining the local physical field correction (after S106) and before generating the updated reference physical field (before S107), the electronic device also performs boundary closure detection, and if a boundary non-closure is detected, expands the perturbation domain, specifically, such as... Figure 2 As shown, the method also includes: S108, the electronic device performs closure detection on the boundary between the inner and outer narrow bands of the boundary narrow band in the initial disturbance domain.

[0066] In this embodiment, closure detection is a crucial step in determining whether the current perturbation domain has covered the main physical effects of the current geometric edit. Please refer to [link / reference]. Figure 3 , Figure 3 This is a schematic diagram illustrating the boundary closure detection and directional expansion process provided in an embodiment of this application. Figure 3 As shown, the closure detection includes the following sub-steps: S1081, the boundary between the inner and outer narrow bands of the boundary narrow band in the initial perturbation domain is taken as the closed detection boundary.

[0067] The closure detection boundary is the interface between the computational domain and the outer narrow band. Closure detection is performed at this boundary, with the inner narrow band (representing the convergence state on the local update field side) and the outer narrow band (representing the state tending towards the reference field side) on the outer side. The closure is determined by comparing the physical states on both sides.

[0068] S1082, divide the closed detection boundary into different boundary segments according to the normal direction and / or the disturbance propagation direction.

[0069] In one optional implementation, the boundary segment division of the closed detection boundary is based on at least one of the following: the external normal direction, the disturbance propagation direction, and the turning point of the terrain or building geometry. There is a hierarchy of priority among these division criteria: the normal direction serves as the basic division criterion, used to determine the basic outline segmentation of the boundary segment; the disturbance propagation direction serves as the correction criterion, used to merge or subdivide the boundary segment based on the physical propagation characteristics on the basis of the normal segmentation; and the turning point of the terrain or building geometry serves as the mandatory boundary, used to ensure that key geometric features do not cross the boundary segment.

[0070] Specifically, the initial segmentation is based on the outer normal direction of the boundary. For each point on the closed detection boundary, its outer normal direction points outward from the initial perturbation domain. When moving along the boundary, if the normal direction changes beyond a preset angle threshold (e.g., 30 or 45 degrees), the boundary is divided into different boundary segments at that location. For example, the four sides of a rectangular perturbation domain correspond to four normal directions (east, south, west, and north), each constituting an independent boundary segment; for irregular perturbation domains with concave and convex contours, each continuous boundary segment with similar normal directions is divided into a boundary segment.

[0071] Secondly, based on the normal direction division, boundary segments are merged or further subdivided according to the disturbance propagation direction. The disturbance propagation direction refers to the dominant direction of physical quantity propagation from the seed node to the boundary (such as the dominant wind direction in a wind field, or the slope aspect in urban flooding). When adjacent normal boundary segments have similar closure characteristics along the same disturbance propagation direction, they can be merged into the same boundary segment; when the disturbance propagation directions of different areas within the same normal boundary segment differ significantly (for example, one part of the boundary segment is upwind of a building passage, and another part is downwind of the passage exit), the boundary segment is further subdivided into multiple sub-segments along the turning point of the propagation direction. For example, in wind environment simulation, although the upwind and downwind boundaries have opposite normal directions, their propagation characteristics differ significantly, so they are divided into different boundary segments; at each building passage exit in an urban canyon, the propagation of disturbances along the passage direction is locally independent, so each passage exit constitutes a separate boundary segment.

[0072] Finally, mandatory boundary segment division points are set based on the turning points of terrain or building geometry. When there are abrupt changes in terrain elevation (such as steep slopes or retaining walls), significant corners or notches in the building outline, or drastic changes in channel width, a boundary segment is still forcibly divided at that location, even if the change in the normal direction at that location does not exceed a preset angle threshold. For example, if there are abrupt changes in airflow separation and wake morphology at a building corner, unifying both sides of the corner into a single boundary segment may cause drastic local fluctuations in the closure detection index; similarly, changes in channel width at the canyon entrance and exit lead to changes in flow state, and mandatory segmentation should be enforced at the location of the width change. When there is a conflict between the normal direction division and the mandatory boundary, the mandatory boundary takes priority. If a mandatory boundary is located inside a normal segment, then the mandatory boundary serves as the dividing line, splitting the normal boundary segment into two independent boundary segments.

[0073] S1083, for each boundary segment, sample the inner and outer sides of the corresponding boundary narrow band to obtain the first state distribution data and the second state distribution data.

[0074] The first state distribution data comes from the local update field side, and the second state distribution data comes from the reference physical field side.

[0075] S1084, based on the first state distribution data and the second state distribution data, evaluates the state continuity index, normal flux index, control equation residual index, and disturbance attenuation index of the boundary segment.

[0076] Among them, the state continuity index is used to determine the difference in physical state quantities on both sides of the boundary, including but not limited to comparing whether the velocity, pressure, temperature, concentration, water depth or other state quantities of the local update field and the external reference field are continuous on both sides of the boundary.

[0077] The normal flux index is used to compare whether the normal flux across a boundary surface is balanced. Normal flux refers to the rate of transport of a physical quantity along the normal direction of the boundary surface; its direction is determined by the projection of the velocity vector onto the normal direction of the boundary. The flux is positive when a physical quantity flows out of the boundary along the normal direction, and negative when it flows in. The normal flux difference is the difference between the outflow flux and the inflow flux at the boundary surface. In a physically conserved system, if the boundary is closed, the net flux across the boundary should be zero, meaning the total outflow of physical quantities equals the total inflow, and there is no net exchange of matter or energy between the two sides of the boundary. Specifically, the flux forms assessed by the normal flux index differ depending on the type of physical field: in wind physical fields, mass flux (density multiplied by the normal component of velocity) and momentum flux are assessed; in thermal physical fields, heat flux (including conduction, radiation, and convection) is assessed; in urban flooding physical fields, water flux (water depth multiplied by the normal component of flow velocity) is assessed; and in pollutant diffusion physical fields, pollutant mass flux (pollutant concentration multiplied by the normal component of velocity) is assessed. In closure detection, if the normal flux difference at a certain boundary segment exceeds a preset threshold, it indicates that there is still net physical quantity flowing out or in at that boundary segment, and the boundary is not yet closed, requiring expansion of the disturbance domain along the corresponding direction.

[0078] The residual index of the governing equations is used to determine whether the physical field within the boundary narrow band satisfies the corresponding governing equations. Different types of physical fields correspond to different governing equations. The governing equations are mathematical equations describing the distribution and evolution of physical fields in space. After substituting the physical quantities (velocity, temperature, concentration, water depth, etc.) of each simulation node within the boundary narrow band into the corresponding governing equations, the difference between the left and right sides of the equations is the governing equation residual. The value of this residual reflects the degree to which the physical field deviates from the governing equations: theoretically, if the physical field completely satisfies the conservation laws, the residual should be zero; in discrete numerical simulations, due to mesh approximation and numerical errors, the residual is close to but not equal to zero. Specifically, the wind physical field and the waterlogging physical field correspond to the continuity equation (mass conservation equation), the thermophysical field corresponds to the energy equation (first law of thermodynamics), the pollutant diffusion physical field corresponds to the transport equation (convection-diffusion equation), and the waterlogging physical field also corresponds to the water balance equation (shallow water equation).

[0079] The normal flux index and the governing equation residual index are related through the Gaussian divergence theorem. The volume integral of the governing equation residual within a region equals the normal flux difference at the boundary of that region. That is, in the continuous case, they are equivalent: the degree to which conservation laws are not satisfied within the region (the volume integral of the governing equation residual) determines the net flux across the boundary of that region (the normal flux difference). However, in discrete numerical simulations, due to factors such as mesh approximation and interpolation errors, the normal flux difference and the governing equation residual are calculated separately and are not completely equal. Therefore, both indices need to be checked simultaneously to ensure both internal physical correctness and boundary continuity. If the normal flux difference meets the tolerance but the governing equation residual is large, it indicates that although the inflow and outflow are balanced at the boundary, there are spurious source and sink terms within the narrow band, and the correction result does not satisfy the physical conservation laws within the narrow band. If the governing equation residual meets the tolerance but the normal flux difference does not, it indicates that the physical laws are satisfied within the narrow band, but there is still flux imbalance at the boundary, with external physical quantities still continuously flowing in or out, and the boundary has not yet truly closed. During narrowband correction, both metrics need to be monitored simultaneously to ensure that they converge to within the tolerance range.

[0080] The disturbance attenuation index is used to determine whether the magnitude of the physical field correction at each simulation node on the boundary segment has decreased to an acceptable range relative to the maximum correction magnitude at simulation nodes within the core area. This index reflects the degree of attenuation of the disturbance propagating outward from the core area to the boundary: if the disturbance has sufficiently attenuated at the boundary, the correction magnitude at the boundary nodes will be much smaller than the maximum correction magnitude in the core area, indicating that the boundary has moved away from the area affected by geometric changes, and the boundary segment can be closed; conversely, if the correction magnitude at the boundary is still relatively large, it indicates that the disturbance has not sufficiently attenuated, the boundary segment should not be closed, and the disturbance domain needs to be further expanded along the direction of the slowest decrease in correction magnitude. By setting an acceptable range threshold (e.g., the boundary correction magnitude is less than 5% of the maximum correction magnitude in the core area), this index can be quantified into an automatically determined condition.

[0081] S1085 If the state continuity index, normal flux index, control equation residual index, and disturbance attenuation index all satisfy their respective preset threshold conditions, and these indices remain stable in multiple consecutive local correction inference iterations, then the boundary segment is determined to be closed.

[0082] S1086 If there are unclosed boundary segments, the initial perturbation domain is extended along the unclosed direction, and the local correction inference and boundary closure detection are re-executed on the extended initial perturbation domain until all boundary segments are closed.

[0083] Specifically, during expansion, the dominant mismatch type of the unclosed boundary segment is first determined. The dominant mismatch type includes at least one of the following: state variable jump-dominant, flux imbalance-dominant, governing equation residual-dominant, and insufficient perturbation attenuation-dominant. The expansion direction is determined based on the dominant mismatch type: if it is state variable jump-dominant, expansion proceeds along the state gradient direction; if it is flux imbalance-dominant, expansion proceeds along the composite direction of the boundary normal and the dominant propagation direction of the physical quantity; if it is governing equation residual-dominant, expansion proceeds along the high residual connected region; and if it is insufficient perturbation attenuation-dominant, expansion proceeds along the direction where the correction decreases the slowest.

[0084] When re-performing local correction inference on the expanded initial perturbation domain, the electronic device only re-performs local correction inference on the newly added simulation nodes in the expansion direction, while retaining the physical field correction amount corresponding to the original initial perturbation domain, thus avoiding redundant calculations. Therefore, the expansion process forms a closed loop of "detecting unclosed boundaries -> adding nodes in a directional manner -> local incremental inference -> re-detection".

[0085] Optionally, the electronic device is also configured with maximum perturbation domain proportion, maximum expansion distance, maximum number of iterations, and maximum uncertainty threshold as backoff trigger conditions. When the expansion still fails to close the domain, or when the external boundary conditions themselves undergo a global change, the electronic device selects local high-precision numerical solution, expanded region high-precision solution, or global re-solution based on the task accuracy level. The backoff results can be used to supplement the proxy training samples, thereby improving the local prediction range and accuracy of subsequent similar perturbations.

[0086] In another optional implementation, after performing local correction inference on the computational region of the initial perturbation domain, the electronic device can also determine whether boundary closure detection needs to be performed based on the confidence level obtained from the local correction inference: if the confidence level is higher than the high confidence threshold, it indicates that the surrogate model's prediction of the current perturbation domain has high credibility, and the updated physical field can be directly synthesized; if the confidence level is lower than the low confidence threshold, boundary closure detection and necessary perturbation domain expansion are triggered to ensure that the correction result satisfies the physical conservation constraints at the boundary. Through this method, the electronic device can flexibly select the execution path according to different situations while ensuring the credibility of the result, achieving a balance between computational accuracy and computational efficiency.

[0087] The above-mentioned S107 specifically includes the following S109-S110: S109, after determining the closure of the boundary segment, the electronic device superimposes the local physical field correction amount onto the physical quantity of the corresponding simulation node in the reference physical field to determine the local updated physical field corresponding to the calculation area in the initial perturbation domain, and performs conservation correction in the boundary narrow band to obtain the boundary physical field after conservation correction corresponding to the boundary narrow band.

[0088] Specifically, the conservation corrections performed by the electronic device within the narrow boundary band include: 1) Based on the locally updated physical field, the locally updated physical field value at the inner boundary of the inner narrow band is used as the inner boundary constraint value; and the reference physical field value at the outer boundary of the outer narrow band is used as the outer boundary constraint value.

[0089] 2) Based on the distances from each simulation node in the boundary narrow band to the inner and outer boundaries, weighted interpolation is performed on the inner and outer boundary constraint values ​​to generate the initial transition field corresponding to the boundary narrow band. That is, when closer to the inner boundary, the transition field is biased towards the local updated physics field; when closer to the outer boundary, the transition field is biased towards the reference physics field.

[0090] 3) Obtain the normal flux difference at the outer boundary of the narrow boundary band in the initial transition field, as well as the residuals of the governing equations within the initial transition field, and iteratively correct the initial transition field. In each iteration, adjust the physical field values ​​of the simulation nodes within the narrow boundary band to reduce the normal flux difference and decrease the residuals of the governing equations, until both the normal flux difference and the residuals of the governing equations satisfy the tolerance, thus obtaining the boundary physical field after conservation correction.

[0091] S110, the electronic device synthesizes and updates the physical field based on the locally updated physical field, the boundary physical field after conservation correction, and the reference physical field.

[0092] In this embodiment of the application, the region where the simulation node is located outside the initial perturbation domain in the simulation model is the external reuse area.

[0093] Specifically, for simulation nodes in the computational domain other than the inner narrowband, their physical field values ​​are obtained from the locally updated physical field; for simulation nodes in the boundary narrowband, their physical field values ​​are obtained from the boundary physical field after conservation correction; and for simulation nodes in the outer reuse area, their physical field values ​​are obtained from the reference physical field. The updated physical field is thus synthesized.

[0094] In other possible implementations, the electronic device assigns a new state version number (e.g., V1) to the updated state and records metadata such as the parent version number, edit events, perturbation domain trajectory, closure index, and cache update range. The electronic device can continue to use the original cache for spatial partitions not covered by the initial perturbation domain, while spatial partitions covered by the perturbation domain store the updated physics and intermediate embeddings. Alternatively, it can directly store the association between the updated physics, the edited geometry, and the edited and updated simulation model (the edited and updated simulation model refers to the simulation model obtained after updating simulation nodes based on the edited geometry and the updated physics). When continuous editing occurs, the electronic device uses the most recently closed version as the new baseline, avoiding recalculation from the initial scene.

[0095] Electronic devices can also update the physical propagation influence diagram corresponding to the target physical field based on the updated physical field and the updated simulation model (e.g., update the edge weights or node states in the propagation influence diagram), ensuring that the physical propagation influence diagram is consistent with the current baseline state during the next edit.

[0096] Example 2:

[0097] The following detailed description, using a complete embodiment of incremental wind environment simulation in an industrial park, clearly and completely illustrates the technical solution of this application. This embodiment is used to illustrate the implementation of this application and does not constitute a limitation on the number of buildings, grid scale, proxy model type, threshold range, or computing equipment. Those skilled in the art can apply the same closed self-expanding mechanism to physical fields such as thermal environments, pollutant diffusion, or urban flooding confluence without departing from the core concept of this application.

[0098] This embodiment is applied to a city industrial park planning and design platform. Please refer to [link / reference]. Figure 4 , Figure 4 This is a schematic diagram illustrating the initial perturbation domain generation and expansion provided in an embodiment of this application. (See diagram below.) Figure 4 As shown, the initial disturbance domain generation and expansion process of the industrial park includes the following steps.

[0099] S401, import the park model and establish a baseline simulation state.

[0100] (1) Importing geometric and boundary conditions. The electronic equipment reads the CityGML or IFC building model, terrain elevation, road and green space boundaries, and converts them to the project coordinate system. The building model retains component identification, bottom outline, height, roof type and surface roughness; the ground area retains elevation, roughness and accessibility attributes. The electronic equipment reads boundary conditions such as prevailing wind direction, inlet reference wind speed, atmospheric stability and surface roughness.

[0101] (2) Simulation Model Establishment. A buffer zone was set up around the perimeter of the park for the electronic equipment, and a three-dimensional mesh was established for numerical simulation. The simulation model includes simulation nodes obtained from the discretization of the baseline geometry and the connection relationships between these nodes. To facilitate incremental updates, the mesh was divided into multiple spatial blocks, and each simulation node recorded its own spatial block, adjacent nodes, building component mappings, and boundary surface normals. The pedestrian height assessment section was set approximately 1.5 meters above ground level.

[0102] (3) Calculation of the reference physical field. The electronic equipment calls the high-precision wind field solver to obtain the reference velocity, pressure and turbulence-related variables, and calculates the mass flux and local divergence residuals of each spatial block boundary. The reference results are stored through spatial index and mapped to the three-dimensional park model.

[0103] (4) State caching. The electronic device stores the baseline geometry, simulation model, baseline physics field, and building topography and surface environment data together to form the baseline state V0. At the same time, the electronic device pre-constructs the physical propagation influence map corresponding to the wind physics field according to the type of wind physics field and based on each simulation node in the simulation model and the propagation mechanism of the wind physics field, as a persistent and reusable asset.

[0104] S402 captures architectural edits and generates perturbation fingerprints.

[0105] (1) Event Acquisition. The planners adjusted the height of the target R&D building from 24 meters to 42 meters, rotated the building plan 15 degrees around the center, and added an overhanging canopy at the east entrance. The 3D editor outputs the edited component identifier, transformation parameters, old and new bounding boxes, and operation timestamp.

[0106] (2) Geometric Difference. The electronic device calculates the occupancy difference between the old and new buildings in a unified grid, and identifies newly added solid volumes, disappeared volumes, and moving surfaces. Changes in height generate newly added volumes at the top; rotation operations generate newly added and disappeared areas with different directions around the building; canopies form locally added obstacles near pedestrian height.

[0107] (3) Disturbance attribute extraction. For each change unit, the electronic device records the spatial location, change volume, change of surface normal, windward or leeward relationship relative to the prevailing wind direction, change of local blockage rate, and characteristic height. Change units located on the windward side of the building are marked as pressure disturbance seeds, change units located on the leeward side and downstream of the roof are marked as wake disturbance seeds, and the edge of the canopy is marked as near-surface high-velocity gradient disturbance seeds.

[0108] (4) Output Results. The electronic device generates the geometric perturbation fingerprint F1 corresponding to this editing event. This perturbation fingerprint contains multiple spatial perturbation units and three levels: changes in the overall building height and rotation (corresponding to the geometric occupancy sub-fingerprint and boundary morphology sub-fingerprint), changes in surface blockage (corresponding to the roughness changes in the physical property sub-fingerprint), and local changes in the entrance canopy (corresponding to the boundary morphology sub-fingerprint). Each sub-fingerprint together determines the seed node and initial direction for the propagation of subsequent influences.

[0109] S403, determine the target physical field and obtain the corresponding physical propagation effect diagram.

[0110] Based on the sub-fingerprint types contained in the geometric perturbation fingerprint, the electronic device determines the affected target physical field. In this embodiment, the geometric perturbation fingerprint includes geometric occupancy sub-fingerprints (volume increase or decrease) and boundary morphology sub-fingerprints (surface position and normal changes), so the target physical field includes wind physical field and thermophysical field.

[0111] The electronic device acquires a pre-constructed physical propagation influence map corresponding to the wind physical field. This physical propagation influence map includes each simulation node in the simulation model and graph edges connecting adjacent simulation nodes, with propagation edge weights associated with the edges. The propagation edge weights are determined based on the flow direction of the target physical field in the reference physical field (i.e., the local velocity direction in the wind field), the spatial distance between simulation nodes, and the building shading relationships between simulation nodes.

[0112] S404: Determine the seed node and propagate the influence score to generate the initial perturbation domain.

[0113] Based on the spatial locations contained in the geometric perturbation fingerprint F1, the electronic device identifies simulation nodes in the simulation model that undergo geometric changes and uses these nodes as seed nodes. For example, nodes corresponding to the increase in building height, nodes corresponding to the rotational change area, and nodes corresponding to the canopy are all seed nodes.

[0114] Starting from the seed node, the influence score is propagated outward along the edges of the physical field's propagation influence map, based on the corresponding propagation edge weights. The wake effect generated by the increase in height mainly extends downwind, while rotational changes create asymmetrical propagation on both sides of the building, and canopy disturbances form a local high-weight region near the ground around the entrance. The influence scores from different propagation paths are accumulated for each simulation node. The electronic system includes simulation nodes with accumulated influence scores higher than a preset threshold in the initial disturbance domain and adds a protective layer to the building's near-wall surface and high-gradient channels.

[0115] This creates an asymmetric initial disturbance domain D1 that extends downwind, is shorter upwind, and is laterally affected by the passageways of adjacent buildings. Simultaneously, electronic devices establish a narrow boundary band N1 consisting of two to several layers of meshes around D1.

[0116] S405, perform local correction inference within the initial perturbation domain.

[0117] The electronic device extracts the new building geometry, grid coordinates, reference velocity and pressure, F1 perturbation fingerprint, D1 boundary inheritance state, and dominant wind direction parameters within D1. A local surrogate model (such as a neural network or deep operator network) receives the above inputs and outputs velocity correction, pressure correction, and local confidence.

[0118] For existing grid blocks in D1, the electronic device reuses the intermediate embeddings cached in V0 and updates their local features with perturbation fingerprints. For grid cells added or reactivated due to building changes, the electronic device re-encodes them. The model primarily calculates the changed regions and their propagation neighborhoods, rather than re-encoding all fifty-two buildings in the park. Within D1, the electronic device adds the baseline field U0 to the predicted correction amount to obtain the first-round local update field U1-local. Larger corrections are allowed in the core area, while the correction amount near the boundary gradually decreases under the influence of training constraints and inherited boundaries.

[0119] S406 performs boundary closure detection and directional expansion.

[0120] Based on the boundary normal and spatial location, the electronic equipment divides the outer boundary of D1 into upwind, downwind, north, and south boundaries, as well as several building passageway boundary segments. An inner and outer sampling band are established for each boundary segment. The velocity and pressure differences between U1-local and U0 on both sides of the boundary are compared to calculate the mass flux imbalance across the boundary surface, the divergence residual within the narrow band is calculated, and the attenuation ratio of the boundary correction to the maximum correction in the core area is statistically analyzed.

[0121] In the first round of judgment, the velocity difference, flux difference, and correction amount of the upwind boundary and some lateral boundaries were all below the threshold and were marked as closed; the correction amount of the downwind boundary was still large, and the passage boundary between two adjacent buildings showed a continuous flux imbalance and was marked as not closed.

[0122] The electronic device does not expand D1 as a whole, but instead adds several grid blocks along the downwind wake path and a lateral grid layer along the channel exit direction, forming D2. The original upwind and closed side boundaries remain unchanged. The initial state of the newly added region in D2 is taken from U0, and the region near D1 inherits the results of the previous local update. The electronic device only performs incremental message passing on the newly added nodes in D2 and the neighborhood of the original downwind boundary, updating the local correction field. After recalculating the closure index, there is still one local boundary segment on the downwind boundary that is not closed. The electronic device continues to expand along the high residual direction of this segment until all boundary segments meet the threshold. To avoid random errors, the electronic device requires all boundary segments to remain closed in two consecutive correction iterations. After the condition is met, the final perturbation domain is denoted as Dfinal, and the expansion trajectory from D1 to Dfinal is recorded.

[0123] S407 performs narrowband conservation correction and synthesizes the global results.

[0124] Within the narrow band of the Dfinal boundary, the electronic device generates smooth weights based on the distances to the inner and outer boundaries, forming an initial continuous transition between the local update field and U0. With the goal of minimizing the net mass flux difference in the boundary segments, small-scale iterations are performed on the velocity normal component and pressure correction within the narrow band to ensure that the inflow and outflow fluxes in each boundary segment meet the tolerance, while simultaneously preventing the correction from excessively propagating into the core region.

[0125] After correction, the narrowband divergence residuals and state transitions are recalculated. If the verification passes, global synthesis is allowed. The Dfinal core region uses the locally updated results, the narrowband uses the corrected results, and the outer region uses U0, forming a new global wind field U1. The electronic equipment simultaneously calculates the difference field between U1 and U0 to display the actual affected area of ​​this building edit.

[0126] S408 performs credibility checks, version updates, and 3D display.

[0127] The electronic device uses a combination of local proxy confidence, number of closure expansion rounds, the proportion of the final perturbation domain to the global domain, and narrowband correction amplitude to determine the result's reliability level. If the local confidence is low but the perturbation domain is still small, the electronic device calls the local high-precision solver within Dfinal; if the final perturbation domain continues to expand and approaches the global domain, it switches to global domain resolving. The electronic device writes the new geometry, U1, Dfinal, closure index, and local cache into the new version V1. Spatial blocks not covered by Dfinal continue to reference the V0 cache, while covered spatial blocks save the updated physics and intermediate embeddings.

[0128] In 3D visualization, the platform overlays the initial disturbance domain D1, the final disturbance domain Dfinal, and the expansion directions of each round. Closed boundary segments, previously open boundary segments, and triggered retreat areas are identified using different line types. Planners can view areas where pedestrian height wind speed increases, decreases, or exceeds planning thresholds, and check whether changes in a particular area are caused by building height, rotation, or local disturbances of canopies.

[0129] If planners move adjacent buildings again, the system generates a new disturbance fingerprint F2 based on V1, reuses the update cache of the overlapping area, and realizes continuous incremental simulation under multiple rounds of urban design editing.

[0130] It should be understood that not all steps involved in the above embodiments need to be executed. For example, the closure detection step in S107 and the subsequent directional expansion step are only required when the boundary of the initial perturbation domain does not meet the closure condition; if the boundary is already closed, the directional expansion step does not need to be executed. Those skilled in the art can adjust the execution order of the steps according to actual needs, or merge or split some steps for execution, and these adjustments do not depart from the protection scope of this application.

[0131] Please refer to Figure 5 This document provides a schematic diagram of the structure of a physical simulation local update device for a city information model, as described in an embodiment of this application. Figure 5 As shown, the physical simulation local update device for the city information model in this embodiment of the application may include: The baseline state acquisition module 501 is used to acquire the baseline state, which includes the baseline geometry, simulation model, and baseline physical field of the city information model before editing. The simulation model includes simulation nodes obtained by discretizing the baseline geometry, and the baseline physical field is the distribution information of physical field variables defined on the simulation nodes in the simulation model. The perturbation fingerprint generation module 502 is used to respond to a geometric editing operation on the reference geometry of the city information model, obtain the edited geometry, and generate a geometric perturbation fingerprint based on the difference between the reference geometry and the edited geometry. The geometric perturbation fingerprint includes change information of geometric units in the reference geometry. The target physics field determination module 503 is used to determine the affected target physics field based on the type of physical property change characterized by the geometric perturbation fingerprint. The target physics field corresponds to a pre-constructed physical propagation influence graph. The physical propagation influence graph includes each simulation node in the simulation model and graph edges connecting adjacent simulation nodes. The graph edges are associated with propagation edge weights used to characterize the strength of the influence of the perturbation propagating along the graph edges. The seed node determination module 504 is used to determine the change nodes of the simulation model that have undergone geometric changes based on the geometric perturbation fingerprint, and to use the change nodes as seed nodes; The initial perturbation domain generation module 505 is used to calculate the propagation influence score of the simulation node based on the corresponding propagation edge weight along the graph edge of the physical propagation influence graph of the target physical field, starting from the seed node, and to include the simulation node whose propagation influence score exceeds a preset threshold into the initial perturbation domain. The correction inference module 506 is used to perform local correction inference based on the initial perturbation domain and the reference physical field to determine the local physical field correction amount of each simulation node in the initial perturbation domain relative to the reference physical field. The result generation module 507 is used to generate an updated physical field based on the local physical field correction and the reference physical field.

[0132] In the embodiments of this application, any of the implementation methods mentioned in the method embodiments are also applicable to the physical simulation local update device for the urban information model provided in this application. For specific execution steps, please refer to the description of the foregoing method embodiments, which will not be detailed here.

[0133] This application also provides a computer storage medium that can store multiple instructions. These instructions are adapted to be loaded and executed by a processor using the physical simulation local update method for the city information model provided in this application. For details of the execution process, please refer to the specific description of the method embodiments shown above, which will not be elaborated here.

[0134] This application also provides a computer program product containing instructions that, when run on an electronic device, cause the electronic device to execute the method steps of the method embodiments shown above.

[0135] This application also provides a chip module, including a transceiver component and a chip, wherein the chip is used to execute the method steps of the above-described method embodiments.

[0136] It is understood that the physical simulation partial update system, the physical simulation partial update device, the computer storage medium, the computer program, the computer program product, and the chip provided above are all used to execute the method shown in any implementation of the corresponding aspect of the embodiments of this application. Therefore, the beneficial effects that can be achieved can be referred to the beneficial effects in the corresponding method, and will not be detailed here.

[0137] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it includes the processes of the embodiments of the above methods.

[0138] The term "at least one" in this application refers to one or more items. "More than one item" means two or more items. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship. Furthermore, it should be understood that although the terms "first," "second," etc., may be used to describe objects in this application, these objects should not be limited to these terms. These terms are only used to distinguish the objects from each other.

[0139] The terms “including” and “having” mentioned above, and any variations thereof, are intended to cover non-exclusive inclusion.

[0140] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for local updating of physical simulation of a city information model, characterized in that, The method includes: Obtain a baseline state, which includes the baseline geometry, simulation model, and baseline physical field of the city information model before editing. The baseline physical field is the distribution information of physical field variables defined on the simulation nodes in the simulation model. In response to a geometric editing operation on the baseline geometry of the city information model, a geometric perturbation fingerprint is generated, the geometric perturbation fingerprint including change information of geometric units in the baseline geometry; Based on the physical property change type characterized by the geometric perturbation fingerprint, the affected target physical field is determined. The target physical field corresponds to a pre-constructed physical propagation influence map. The physical propagation influence map includes each simulation node in the simulation model and graph edges connecting adjacent simulation nodes. The graph edges are associated with propagation edge weights used to characterize the strength of the influence of the perturbation propagating along the graph edges. The physical propagation influence map is constructed according to the type of physical field, based on the simulation nodes and preset propagation influence factors. The preset propagation influence factors include the propagation mechanism of different types of physical fields. Based on the geometric perturbation fingerprint, the nodes in the simulation model that have undergone geometric changes are identified, and these nodes are used as seed nodes. Starting from the seed node, along the graph edges of the physical propagation influence graph of the target physical field, calculate the propagation influence score of the simulation node based on the corresponding propagation edge weight, and include simulation nodes whose propagation influence scores exceed a preset threshold into the initial perturbation domain; Based on the initial perturbation domain and the reference physical field, local correction inference is performed to determine the local physical field correction amount of each simulation node in the initial perturbation domain relative to the reference physical field, and an updated reference physical field is generated based on the local physical field correction amount and the reference physical field.

2. The method as described in claim 1, characterized in that, The baseline state also includes building topography and surface environment data. This data is mapped to the simulation nodes through a mapping relationship between geometric elements and simulation nodes. The building topography and surface environment data includes at least one of the following: building layout, topographic elevation, surface roughness, underlying surface material, and drainage network connectivity. The preset propagation influencing factors also include the geometric spatial relationships between nodes and building topography and surface environment data. The propagation edge weights are determined based on one or more of the following influence factors: The reference physical field includes the flow direction of the target physical field, the spatial distance between simulation nodes, the terrain slope between simulation nodes, the building occlusion relationship between simulation nodes, and the channel connectivity between simulation nodes.

3. The method as described in claim 1 or 2, characterized in that, The types of physical fields include wind physical fields, thermal physical fields, and waterlogging physical fields; The geometric perturbation fingerprint includes at least one of the following sub-fingerprints: Geometric occupancy fingerprints are used to describe changes in the volume of geometric units. Boundary morphology sub-fingerprints are used to describe the surface position changes and / or normal direction changes of geometric units; Sub-fingerprint of physical properties is used to describe changes in at least one of the following: material, roughness, and permeability of a geometric unit; Topological sub-fingerprints are used to describe at least one variation in the connections between geometric units, such as channels, roads, drainage pipes, and open space connections. The determination of the affected target physical field based on the type of physical property change characterized by the geometric perturbation fingerprint includes: In the case where the geometric perturbation fingerprint includes the geometric occupant fingerprint, the target physical field includes the wind physical field and the thermophysical field; In the case where the geometric perturbation fingerprint includes the boundary morphology sub-fingerprint, the target physical field includes the wind physical field and the thermophysical field; When the geometric perturbation fingerprint includes the physical property sub-fingerprint, the target physical field includes at least one of the thermophysical field, the wind physical field, and the waterlogging physical field, wherein the material variation in the physical property sub-fingerprint corresponds to the thermophysical field, the roughness variation corresponds to the wind physical field, and the permeability variation corresponds to the waterlogging physical field; When the geometric perturbation fingerprint includes the topological sub-fingerprint, the target physical field includes at least one of the thermophysical field, the wind physical field, and the waterlogging physical field, wherein changes in the channel or open space connectivity in the topological sub-fingerprint correspond to the thermophysical field and the wind physical field, and changes in the drainage pipeline connectivity correspond to the waterlogging physical field.

4. The method as described in claim 1 or 2, characterized in that, The step of performing local correction inference based on the initial perturbation domain and the reference physical field includes: The simulation model is divided into a core region, a computation region, and a boundary narrow band. The core region contains simulation nodes that undergo geometric changes. The computation region covers the core region and extends outward to the boundary of the initial perturbation domain. The boundary narrow band includes an inner narrow band and an outer narrow band. The inner narrow band includes several layers of simulation nodes inside the edge of the computation region, and the outer narrow band includes several layers of simulation nodes outside the edge of the computation region. Based on the reference physical field, local correction inference is performed on the computational region in the initial perturbation domain; Among them, the maximum correction amount of the physical field correction amount for the simulation node in the core area of ​​the computing area is the first amplitude; The maximum correction value for the physical field correction of the simulation node located outside the core area and within the inner boundary of the inner narrow band within the computing area is the second amplitude, which is smaller than the first amplitude. The maximum correction value for the physical field correction of the simulation node within the narrow band of the computation region is the third amplitude, which is less than the second amplitude.

5. The method as described in claim 4, characterized in that, After determining the local physics field correction amount of each simulation node within the initial perturbation domain relative to the reference physics field, and before generating the updated reference physics field based on the local physics field correction amount and the reference physics field, the method further includes: The boundary between the inner and outer narrow bands of the boundary narrow band in the initial perturbation domain is taken as the closure detection boundary; The closed detection boundary is divided into different boundary segments according to the normal direction and / or the disturbance propagation direction; For each of the boundary segments, samples are taken from the inner and outer sides of the boundary narrow band corresponding to the boundary segment to obtain first state distribution data and second state distribution data. Based on the first state distribution data and the second state distribution data, the state continuity index, normal flux index, control equation residual index, and disturbance attenuation index of the boundary segment are evaluated. The state continuity index is used to determine the difference in physical state quantities on both sides of the boundary. The normal flux index is used to compare whether the normal flux passing through the boundary surface is balanced. The control equation residual index is used to determine whether the physical field within the narrow band of the boundary satisfies the control equation. The disturbance attenuation index is used to determine whether the magnitude of the physical field correction of each simulation node on the boundary segment has decreased to an allowable range relative to the maximum correction in the core area. If the state continuity index, the normal flux index, the control equation residual index, and the disturbance attenuation index all satisfy their respective preset threshold conditions, and the indexes remain stable in multiple consecutive local correction inference iterations, then the boundary segment is determined to be closed. If there are unclosed boundary segments, the initial perturbation domain is extended along the unclosed direction, and the local correction inference and boundary closure detection are re-executed on the extended initial perturbation domain until all boundary segments are closed; The step of generating an updated physical field based on the local physical field correction and the reference physical field includes: If all boundary segments are closed, an updated physical field is generated based on the local physical field correction and the reference physical field.

6. The method as described in claim 5, characterized in that, The expansion of the initial perturbation domain along the unclosed direction includes: Determine the dominant mismatch type of the unclosed boundary segment, wherein the dominant mismatch type includes at least one of the following: state variable jump dominant type, flux imbalance dominant type, governing equation residual dominant type, and disturbance attenuation insufficient dominant type. The expansion direction is determined based on the dominant mismatch type. If it is a state variable jump type, it expands along the state gradient direction; if it is a flux imbalance type, it expands along the synthesis direction of the boundary normal and the dominant propagation direction of the physical quantity; if it is a control equation residual type, it expands along the high residual connected region; if it is a disturbance attenuation insufficient type, it expands along the direction where the correction decreases the slowest. The re-execution of local correction inference and boundary closure detection on the expanded initial perturbation domain includes: The local correction inference is re-executed for the newly added simulation nodes in the expansion direction, and the physical field correction amount corresponding to the original initial perturbation domain is retained to obtain the expanded local physical field correction amount.

7. The method as described in claim 5 or 6, characterized in that, The simulation node region located outside the initial perturbation domain in the simulation model is the external reuse region. After determining the boundary segment closure and before generating the updated physical field based on the local physical field correction and the reference physical field, the method further includes: Based on the local physical field correction, it is superimposed on the physical quantities of the corresponding simulation nodes in the reference physical field to determine the local updated physical field in the initial perturbation domain corresponding to the computation area; Perform a conservation correction within the boundary narrow band, the conservation correction comprising: Based on the locally updated physical field, the locally updated physical field value at the inner boundary of the inner narrow band is used as the inner boundary constraint value; and the reference physical field value at the outer boundary of the outer narrow band is used as the outer boundary constraint value. Based on the distances from each simulation node in the boundary narrow band to the inner and outer boundaries, the inner boundary constraint values ​​and the outer boundary constraint values ​​are weighted and interpolated to generate the initial transition field corresponding to the boundary narrow band. The normal flux difference at the outer boundary of the boundary narrow band in the initial transition field and the residual of the control equation in the initial transition field are obtained. The initial transition field is iteratively corrected. In each iteration, the physical field values ​​of the simulation nodes in the boundary narrow band are adjusted so that the normal flux difference decreases and the residual of the control equation decreases until both the normal flux difference and the residual of the control equation satisfy the tolerance, thus obtaining the boundary physical field after conservation correction. The step of generating an updated physical field based on the local physical field correction and the reference physical field includes: The updated reference physical field is synthesized by using the values ​​in the locally updated physical field for the physical field values ​​of the simulation nodes in the computational region other than the inner narrow band, the values ​​in the boundary physical field after conservation correction for the physical field values ​​of the simulation nodes in the boundary narrow band, and the values ​​in the reference physical field for the physical field values ​​of the simulation nodes in the outer multiplexing region.

8. A physical simulation local update device for a city information model, characterized in that, Includes a unit for performing the method as described in any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a computer program, which, when executed, performs the method according to any one of claims 1 to 7.

10. An electronic device, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method of any one of claims 1 to 7 is performed.

Images