A multi-level parallel world simulation model dynamic loading method and system

Abstract

The application provides a kind of multilevel parallel world simulation model dynamic loading method and system, obtain the multiple precision fluid dynamics data set containing fluid motion state information of different fine degree;Scan the area where the fluid motion change rate in multiple precision fluid dynamics data set exceeds the preset change rate threshold to generate optical marking area;Convert optical marking area into visible light pattern and project into the preset display plane, generate visual focus area;Real-time optimization low fine degree fluid motion state information generates detail enhancement data;Determine the data source of the preset region to be loaded in combination with visual focus area and detail enhancement data;According to the data source, select the high detail data set in multiple precision fluid dynamics data set or detail enhancement data to complete the real-time loading of multilevel fluid simulation model using the selection result.The application realizes the high-fidelity real-time loading of multilevel fluid simulation model, improves the loading efficiency and the authenticity of visual effect.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of model dynamic loading, and particularly relates to a multi-level parallel world simulation model dynamic loading method and system. BACKGROUND

[0002] With the development of virtual reality and augmented reality technologies, there is an increasing demand for real-time simulation visual presentations that include high-fidelity physical details. These technologies are widely used in fields such as film production, game development, education and training, and engineering simulation, requiring the provision of realistic visual effects while maintaining the instant responsiveness of interactions, especially the accuracy and real-time performance of fluid dynamics data in the natural landscape reproduction of complex scenes. To meet these demands, not only efficient data processing methods are needed, but also innovative display technologies are needed to enhance user experience.

[0003] The current mainstream solution is a physical simulation engine based on high-performance graphics processors to achieve the above goals. By optimizing algorithms for parallel processing of multi-precision fluid dynamics data on high-performance graphics processors, the rendering efficiency is improved. Combined with deep learning models to predict fluid motion, the computational resource consumption is reduced, and accurate performance is ensured in dynamically changing areas. This method greatly improves the realism and interactive experience of simulation, allowing users to explore and learn in a near-real environment. The existing solution has some inherent defects, including high requirements for hardware, which increases the deployment difficulty of applications that want to run on ordinary consumer-level devices; performance problems may occur in the face of extremely high resolution or the need for extremely fine detail performance, resulting in frame rate reduction or picture delay and affecting user experience; although deep learning models are used to optimize fluid performance, they cannot guarantee the best visual effects and the smoothest user experience in all situations; and the dependence on the performance of specific hardware limits flexibility, and may still face computational bottlenecks when dealing with particularly complex scenes. SUMMARY

[0004] The present application provides a multi-level parallel world simulation model dynamic loading method and system to solve the problems in the prior art, such as high requirements for hardware, which increases the deployment difficulty of applications that want to run on ordinary consumer-level devices; performance problems may occur in the face of extremely high resolution or the need for extremely fine detail performance, resulting in frame rate reduction or picture delay and affecting user experience; although deep learning models are used to optimize fluid performance, they cannot guarantee the best visual effects and the smoothest user experience in all situations; and the dependence on the performance of specific hardware limits flexibility, and may still face computational bottlenecks when dealing with particularly complex scenes.

[0005] The first aspect of the present application provides a multi-level parallel world simulation model dynamic loading method, comprising:

[0006] Obtaining a multi-precision fluid dynamics data set containing fluid motion state information of different levels of detail;

[0007] Optically scanning a region in the multi-precision fluid dynamics data set where the fluid motion rate exceeds a preset rate threshold, to generate an optically marked region;

[0008] Converting the optically marked region into a visible light pattern, projecting the visible light pattern into a preset display plane to generate a visual focal point region;

[0009] According to a preset physical law constraint neural network, performing real-time optimization processing on the fluid motion state information of low levels of detail to generate detail enhancement data;

[0010] Combining the visual focal point region and the detail enhancement data to determine the data source of a preset region to be loaded according to the combination result;

[0011] According to the data source, selecting a high-detail data set in the multi-precision fluid dynamics data set or the detail enhancement data to complete real-time loading of a multi-level fluid simulation model using the selection result.

[0012] Optionally, obtaining a multi-precision fluid dynamics data set containing fluid motion state information of different levels of detail, comprises:

[0013] Dividing a preset target fluid domain to obtain a plurality of precision levels of divided fluid domains;

[0014] Assigning each of the divided fluid domains a corresponding set of preset simulation parameters to generate a plurality of parameterized fluid domains corresponding to each of the divided fluid domains;

[0015] Performing numerical simulation calculation using each of the parameterized fluid domains to generate fluid motion state information corresponding to each of the precision levels;

[0016] Connecting each of the fluid motion state information to establish a hierarchical correlation mapping relationship;

[0017] Aggregating each of the fluid motion state information and the hierarchical correlation mapping relationship to generate a multi-precision fluid dynamics data set.

[0018] Optionally, optically scanning a region in the multi-precision fluid dynamics data set where the fluid motion rate exceeds a preset rate threshold to generate an optically marked region, comprises:

[0019] calculating fluid motion change rate values of each adjacent spatial unit in the multi-precision fluid dynamics data set;

[0020] comparing each of the fluid motion change rate values with a preset change rate threshold to identify a change rate overrun region;

[0021] driving a preset micro-mirror array to reciprocally optically scan the change rate overrun region, to generate a beam deflection path;

[0022] applying the beam deflection path to drive a preset laser light source to cover the change rate overrun region, to generate an optical coverage region;

[0023] collecting optical feedback signals of the optical coverage region to generate an optical marking region.

[0024] Optionally, converting the optical marking region into a visible light pattern, projecting the visible light pattern into a preset display plane to generate a visual focal point region, including:

[0025] vectorizing the outline of the optical marking region to generate vector outline data;

[0026] inputting the vector outline data into a preset spatial light modulator to generate a phase modulation pattern;

[0027] applying the phase modulation pattern to wavefront shaping processing of a preset laser beam to generate a structured illumination beam;

[0028] projecting the structured illumination beam onto a preset display plane to generate a projected optical pattern;

[0029] identifying the light distribution characteristics of the projected optical pattern on the preset display plane to generate a visual focal point region.

[0030] Optionally, according to a preset physical law constraint neural network, real-time optimization processing of low-fidelity fluid motion state information is performed to generate detail enhancement data, including:

[0031] extracting key physical quantities from the low-fidelity fluid motion state information;

[0032] inputting the key physical quantities into a preset physical law constraint neural network to generate a physical law deviation amount;

[0033] adjusting internal parameters of the preset physical law constraint neural network according to the physical law deviation amount to generate an optimized neural network;

[0034] The optimized neural network is applied to resolution enhancement processing of the low-fidelity fluid motion state information, to generate preliminary enhanced data.

[0035] The preliminary enhanced data is fused with the high-fidelity fluid motion state information corresponding to the optical mark area, to generate detail enhanced data.

[0036] Optionally, the visual focus area and the detail enhanced data are combined, and a data source of a preset to-be-loaded area is determined according to a combination result, including:

[0037] The visual focus area is subjected to spatial priority evaluation, to generate a region priority sequence.

[0038] The region priority sequence is subjected to consistent matching with the detail enhanced data, to generate a matching result matrix.

[0039] The matching result matrix is used for data requirement analysis of a preset to-be-loaded area, to generate a data source allocation scheme.

[0040] Compatibility between the data source allocation scheme and the multi-precision fluid dynamics data set is verified, to generate a verified data source allocation scheme.

[0041] The verified data source allocation scheme is used for determining a data source of the preset to-be-loaded area.

[0042] Optionally, high-detail data sets in the multi-precision fluid dynamics data set or the detail enhanced data are selected according to the data source, to complete real-time loading of a multi-level fluid simulation model by using a selection result, including:

[0043] Loading instructions in the data source are parsed, to generate a data loading instruction set.

[0044] High-detail data sets in the multi-precision fluid dynamics data set or the detail enhanced data are selected according to the data loading instruction set, and a plurality of to-be-loaded data blocks are extracted from a selection result.

[0045] Each to-be-loaded data block is subjected to integrity verification, to generate a plurality of verified to-be-loaded data blocks.

[0046] Each verified to-be-loaded data block is subjected to parallel integration, to generate integrated simulation model data.

[0047] The integrated simulation model data is input into a preset real-time rendering pipeline, to complete real-time loading of a multi-level fluid simulation model.

[0048] In a second aspect, the present application provides a multi-level parallel world simulation model dynamic loading system, comprising:

[0049] An acquisition module is configured to acquire a multi-precision fluid dynamics data set containing fluid motion state information of different levels of detail.

[0050] A scanning module is configured to perform optical scanning on a region in the multi-precision fluid dynamics data set where the fluid motion rate exceeds a preset rate threshold, to generate an optical marking region.

[0051] A projection module is configured to convert the optical marking region into a visible light pattern, project the visible light pattern into a preset display plane, and generate a visual focus region.

[0052] An optimization module is configured to perform real-time optimization processing on the fluid motion state information of low levels of detail according to a preset physical law constraint neural network, to generate detail enhancement data.

[0053] A combination module is configured to combine the visual focus region and the detail enhancement data, and determine a data source of a preset to-be-loaded region according to a combination result.

[0054] A selection module is configured to select a high-detail data set in the multi-precision fluid dynamics data set or the detail enhancement data according to the data source, to complete real-time loading of a multi-level fluid simulation model by using a selection result.

[0055] In a third aspect, the present application provides a computing device comprising a processor and a memory, wherein the memory stores a computer program, and the processor is configured to run the computer program to perform the multi-level parallel world simulation model dynamic loading method according to any one of the first aspect.

[0056] In a fourth aspect, the present application provides a computer storage medium storing computer program instructions, wherein the computer program instructions are executed by a processor to implement the multi-level parallel world simulation model dynamic loading method according to any one of the first aspect.

[0057] The present application acquires a data set containing multi-precision information, automatically identifies a key region where fluid motion changes dramatically by using an optical scanning technology, converts the key region into a visible light pattern to form a visual focus, adopts a physical law constraint neural network to perform real-time optimization on low-precision data to generate enhanced details, and finally combines the visual focus and the enhanced data to dynamically select an optimal data source, thereby achieving high-fidelity real-time loading of a multi-level fluid simulation model, and effectively improving loading efficiency and the authenticity of visual effects.

[0058] Further, by extracting key physical quantities from the low-precision data and inputting them into a physical law constraint network to calculate a bias amount, the network internal parameters are adjusted accordingly and an optimized network is generated, and then the resolution enhancement processing is performed on the low-precision data to obtain a preliminary enhancement result, and then the preliminary enhancement result is fused with the high-precision data corresponding to the optical mark area, so that the physical accuracy and the detail richness of the fluid simulation data are significantly enhanced, and the high fidelity and dynamic response capability of the final loaded model are ensured.

[0059] These and other aspects of the present application will become more apparent from the following description in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0060] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, hereinafter, a brief introduction will be given to the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0061] Figure 1 A flow chart of a multi-level parallel world simulation model dynamic loading method provided by an embodiment of the present application;

[0062] Figure 2 A structural schematic diagram of a multi-level parallel world simulation model dynamic loading system provided by an embodiment of the present application;

[0063] Figure 3 A structural schematic diagram of a computing device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0064] In order to make the technical personnel in the art better understand the present application scheme, the technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application.

[0065] In some of the descriptions in the specification and claims of the present application and the above-mentioned drawings, a plurality of operations appearing in a specific order are included, but it should be clearly understood that these operations can be executed or in parallel without the order appearing in this text. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes can include more or fewer operations, and the operations can be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this text are used to distinguish different messages, devices, modules, etc., and do not represent the sequence, nor do "first" and "second" represent different types.

[0066] Clearly and completely, the technical solutions in the embodiments of the present application will be described below with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0067] Figure 1 A flow chart of a multi-level parallel world simulation model dynamic loading method is provided for the embodiments of the present application, as shown in Figure 1 The method comprises the following steps.

[0068] The existing simulation model dynamic loading technical solution based on computational fluid dynamics has three key defects: first, the traditional method cannot maintain the efficient operation of a large-scale scene and the high fidelity of physical details in a key area at the same time under limited computing resources; second, the conventional data scheduling strategy lacks real-time perception ability of fluid dynamic change characteristics, resulting in lag in visual focus area identification and unreasonable allocation of computing resources; third, the mainstream neural network enhancement method often deviates from the constraint of physical laws and relies only on data-driven generation of details, which is easy to produce visual artifacts that violate physical reality. To solve these problems, the research and development idea of the present application is: by fusing multi-precision fluid dynamics data and optical scanning technology, the fluid motion change rate is converted into a visible light marker to realize fast and accurate positioning of the visual focus; at the same time, the neural network is constrained by physical laws to optimize low-precision data in real time, ensuring that the enhanced details conform to the physical laws; finally, through the collaborative decision of the visual focus and the enhanced data, the optimal data source is dynamically selected, realizing the efficient real-time loading of the multi-level model under the premise of maintaining the physical reality. Based on this, the present application provides a multi-level parallel world simulation model dynamic loading method, as shown in Figure 1 , comprising the following steps.

[0069] Step 101: Obtain a multi-precision fluid dynamics data set, which contains fluid motion state information of different levels of detail.

[0070] In this step, the multi-precision fluid dynamics data set refers to a data set containing multiple resolution versions of the same fluid scene, including information levels reflecting the physical properties of fluid velocity, pressure, density, etc. with different levels of detail, which are used to provide multiple detail options for dynamic loading; the level of detail refers to the richness of physical details and the level of spatial resolution contained in the fluid motion state information, which is obtained based on the grid density and physical model complexity set during numerical simulation calculation; the fluid motion state information refers to the data used to describe the motion characteristics of the fluid at a specific time and spatial point, including velocity vector, pressure scalar, density value, etc. physical quantities, which are obtained based on computational fluid dynamics simulation.

[0071] In the embodiment of the present application, a multi-precision fluid dynamics data set is first generated by numerical simulation calculation, which specifically contains a plurality of fluid motion state information of different precision levels for the same fluid scene, where the precision level is mainly determined by the resolution of the data and the richness of physical details.

[0072] Step 102: optically scanning the region in the multi-precision fluid dynamics data set where the fluid motion rate exceeds a preset rate threshold, to generate an optically marked region.

[0073] In this step, the fluid motion rate refers to an index for quantifying the degree of change of fluid motion state over time or space, which is obtained by dividing the difference of physical quantities between adjacent time steps or spatial points by the corresponding interval; the preset rate threshold refers to a preset numerical limit for determining whether the fluid motion change is severe enough to require priority processing, which is determined based on the overall system resources and visual fidelity requirements; the optical scanning operation refers to the action of using a micro-electro-mechanical system mirror array to control the high-speed reciprocating scanning of a laser beam on a specific spatial region, for converting data features into physical optical signals; and the optically marked region refers to a physical spatial region specially marked by light, which is generated by optically scanning the fluid motion rate exceeding region, and is used to indicate the key region requiring high-fidelity loading.

[0074] In the embodiment of the present application, the fluid motion rate of each local region in the multi-precision fluid dynamics data set is first calculated, and then each calculated fluid motion rate value is compared with a preset rate threshold one by one, to identify all specific regions whose rate exceeds the preset threshold, and then the micro-electro-mechanical system mirror array is driven to perform high-speed reciprocating optical scanning on these identified rate exceeding regions, and finally an optical marked region is generated by the optical scanning operation.

[0075] Step 103: converting the optically marked region into a visible light pattern, projecting the visible light pattern into a preset display plane, to generate a visual focus region.

[0076] In this step, the conversion operation refers to the process of converting the profile information of the optical mark region from the optical signal to the digital vector instruction for generating the control signal for the spatial light modulator; the visible light pattern refers to the optical intensity distribution pattern generated by the spatial light modulator according to the vector instruction, which can be directly perceived by human vision; the projection operation refers to the process of imaging the visible light pattern from the spatial light modulator to a display plane by using the optical projection system; the preset display plane refers to a preset physical surface for receiving and displaying the projected optical pattern, which is usually the main output screen of the simulation rendering system; and the visual focus region refers to the area on the display plane that is illuminated and highlighted by the visible light pattern, which is used to indicate the simulation part that should be focused on by the user at present and needs high-fidelity rendering.

[0077] In the embodiment of the present application, the profile information possessed by the optical mark region is first subjected to vectorization conversion processing to generate corresponding digital control instructions, then the spatial light modulator is used to generate a corresponding phase modulation pattern according to these instructions, and then the phase modulation pattern is applied to perform wavefront shaping operation on the light beam emitted by a preset laser light source to generate a structured illumination light beam, and finally the structured illumination light beam is accurately projected onto a preset display plane to generate a clear visual focus region.

[0078] Step 104: Real-time optimization processing of the low-fineness fluid motion state information is performed according to the preset physical law constraint neural network to generate detail enhancement data.

[0079] In this step, the preset physical law constraint neural network refers to a neural network model in which an internal loss function is embedded in the fluid mechanics control equation, which is used to ensure that the output data conforms to the physical law; the low-fineness fluid motion state information refers to the part of data with lower resolution and sparser physical details in the multi-precision data set, which is used as the basis input for real-time optimization processing; the real-time optimization processing refers to the calculation process of quickly enhancing the low-precision data by using the physical law constraint neural network, which is used to improve the data details while ensuring the physical authenticity; and the detail enhancement data refers to the high-precision fluid data obtained by real-time optimization processing, which is rich in physical details and conforms to the physical law, and is used as a supplement or replacement for the high-detail data set.

[0080] In the embodiment of the present application, firstly, key physical quantities such as velocity and pressure are extracted from the low-fidelity fluid motion state information, then the key physical quantities are input into a preset physically law-constrained neural network to calculate a physical law deviation, and then the internal parameters of the neural network are dynamically adjusted according to the physical law deviation to generate an optimized neural network, then the optimized neural network is applied to the resolution enhancement processing of the low-fidelity fluid motion state information to generate preliminary enhanced data, and finally the preliminary enhanced data and the high-fidelity fluid motion state information corresponding to the optical mark region are subjected to consistency fusion operation to finally generate the required detail enhancement data.

[0081] Step 105: combining the visual focus region and the detail enhancement data, and determining the data source of the preset to-be-loaded region according to the combination result.

[0082] In this step, the combination operation refers to the process of associating and matching the spatial position information of the visual focus region with the physical content information of the detail enhancement data, which is used to provide a basis for subsequent decision-making; the combination result refers to the matrix or mapping table representing the matching relationship between the visual focus and the data enhancement region generated after the combination operation; the preset to-be-loaded region refers to the three-dimensional space range that needs to be loaded or updated according to the simulation process and the user's perspective; and the data source refers to the specific data supplier for filling the to-be-loaded region, including a pre-calculated high-detail data set or real-time generated detail enhancement data.

[0083] In the embodiment of the present application, firstly, the spatial range covered by the visual focus region is subjected to priority evaluation to generate a region priority sequence, then the region priority sequence is subjected to consistency matching operation with the spatial information contained in the detail enhancement data to generate a matching result matrix, then a preset to-be-loaded region is subjected to data demand analysis according to the matching result matrix to generate a data source allocation scheme, and finally the data source corresponding to the preset to-be-loaded region is determined according to the data source allocation scheme.

[0084] Step 106: selecting a high-detail data set in the multi-precision fluid dynamics data set or the detail enhancement data according to the data source, so as to complete the real-time loading of the multi-level fluid simulation model by using the selection result.

[0085] In this step, the high-detail data set refers to the pre-computed data level with the highest resolution and richest physical details in the multi-precision fluid dynamics data set; the selection result refers to the specific data instance finally determined for loading according to the data source decision; the multi-level fluid simulation model refers to a composite simulation model composed of model data with different levels of detail, which can dynamically switch the level of detail as needed; and the real-time loading operation refers to the process of quickly injecting the selected data block into the rendering pipeline and displaying it immediately, for realizing seamless update of the simulation picture.

[0086] In the embodiment of the present application, first, the loading instructions contained in the data source allocation scheme are parsed to generate a specific data loading instruction set, then one of the high-detail data set contained in the multi-precision fluid dynamics data set or the detail enhancement data generated in the foregoing step is selected according to the data loading instruction set and data reading is performed to generate a plurality of to-be-loaded data blocks, then integrity verification is performed on each to-be-loaded data block to generate a plurality of verified to-be-loaded data blocks, then parallel integration operation is performed on all the verified to-be-loaded data blocks to generate a complete integrated simulation model data, and finally the integrated simulation model data is injected into a preset real-time rendering pipeline to complete real-time loading of the multi-level fluid simulation model.

[0087] For example, first, a multi-precision fluid dynamics data set containing three grid precisions of coarse, medium and fine is generated by performing computational fluid dynamics simulation through a high-performance computing cluster. Second, the system automatically calculates the velocity variation rate of each grid cell in the data set and compares it with a preset threshold to identify vortex core areas, and then drives a micro-electromechanical system mirror array to perform laser scanning on these areas to generate optical markers. Third, an optical system converts the marker area profile into vector instructions to control a spatial light modulator to generate a phase pattern, and then modulates a laser beam and projects it onto a ring-shaped projection screen to form a high-light visual focus. At the same time, a physical information neural network extracts the velocity and pressure field from the coarse grid data, calculates the deviation from the Navier-Stokes equation, and adjusts the network parameters accordingly to perform super-resolution processing on the coarse grid data to generate preliminary enhancement results, and then fuses them with high-precision reference data in the corresponding area of the optical marker to generate physically accurate detail enhancement data. Subsequently, a processing engine matches the visual focus area coordinates with the spatial range of the detail enhancement data to generate a data source allocation scheme indicating that the vortex core area uses enhancement data and other areas use pre-computed medium-precision data. Finally, a renderer reads data blocks from the corresponding sources according to the scheme, verifies their integrity, and then integrates them into unified scene data in parallel, and submits them to the graphics application programming interface rendering pipeline to complete real-time simulation picture update of thirty frames per second.

[0088] The embodiment of the application realizes the efficient real-time loading of the multi-level fluid simulation model while maintaining high physical realism under limited computing power by integrating the optical scanning focusing and the physical law constraint neural network enhanced dual technical path, first, using optical means to quickly and accurately identify the key area with rapid changes in fluid motion and convert it into a visual focus, second, generating physically credible enhanced details by real-time optimization of low-precision data through a physical law constraint neural network, and finally intelligently fusing the visual focus and enhanced data to dynamically select the optimal data source, thereby effectively solving the key technical problems of unreasonable allocation of computing resources, and the difficulty of balancing visual details and physical authenticity in large-scale high-fidelity fluid simulation.

[0089] In order to solve the coordination problem in the multi-precision fluid data generation process, this step divides the target fluid domain by level and configures corresponding parameters, and finally integrates to generate a complete data set containing hierarchical relationships. The present application provides a specific embodiment, step 101, obtaining a multi-precision fluid dynamics data set, which contains fluid motion state information of different levels of detail, specifically including the following steps:

[0090] Step 111: dividing the preset target fluid domain to obtain a plurality of precision levels of the divided fluid domain.

[0091] In this step, the preset target fluid domain refers to the physical space region that needs to be calculated for fluid simulation, including the definition of specific geometric shape size and boundary conditions, which is determined based on the requirements of the simulation task; the division operation refers to the process of spatial discretization of the target fluid domain according to different grid density requirements, which is used to generate calculation regions of different precision; the precision level refers to the grid resolution level adopted by the divided fluid domain, which is reflected by the size of the grid size, and is obtained based on the grid density parameter set during the division operation; the divided fluid domain refers to each sub-computing region generated after the target fluid domain is divided, each region corresponding to a specific precision level.

[0092] In the embodiment of the present application, first, a pre-set target fluid domain space range is determined as the simulation object, then a grid generation technique is used to perform spatial discretization division operation on the target fluid domain, which is divided into a plurality of sub-regions with different grid densities according to the required level of detail, and finally a series of divided fluid domains corresponding to different precision levels are obtained.

[0093] Step 112: assigning a corresponding set of preset simulation parameters to each of the divided fluid domains to generate a plurality of parameterized fluid domains corresponding to each of the divided fluid domains.

[0094] In this step, the assignment operation refers to the process of establishing a correspondence between the simulation parameter set and the divided fluid domain, which is used to configure appropriate calculation parameters for each calculation region; the preset simulation parameter set refers to a set of calculation parameters configured in advance for a specific precision level, including time step, turbulence model parameters, convergence criteria, etc., which are determined based on the characteristics of the precision level and the limitations of the computing resources; the parameterized fluid domain refers to the divided fluid domain that has been assigned a simulation parameter set, which contains complete information of the calculation region definition and calculation parameter configuration.

[0095] In the embodiments of the present application, first, a pre-configured simulation parameter set is prepared for each divided fluid domain, which contains calculation parameters suitable for the grid precision level, and then each simulation parameter set is assigned to the corresponding divided fluid domain through the parameter binding operation, finally generating a series of parameterized fluid domains corresponding to each divided fluid domain.

[0096] Step 113: Numerical simulation calculation is performed using each parameterized fluid domain to generate fluid motion state information corresponding to each precision level.

[0097] In this step, the numerical simulation calculation operation refers to the process of solving fluid control equations using computational fluid dynamics methods, which is used to simulate the fluid motion state.

[0098] In the embodiments of the present application, first, the parameterized fluid domains are input into the computational fluid dynamics solver, and then numerical simulation calculation operations are performed for each parameterized fluid domain, respectively, to obtain fluid motion state information at each precision level by solving fluid control equations, and finally generate a fluid motion state information dataset corresponding to each precision level.

[0099] Step 114: Connect each fluid motion state information to establish a hierarchical association mapping relationship.

[0100] In this step, the connection operation refers to the process of establishing a correspondence between fluid motion state information of different precision levels, which is used to realize multi-precision data association; the hierarchical association mapping relationship refers to a data structure that records the spatial correspondence between fluid motion state information of different precision levels, which is established through the connection operation.

[0101] In the embodiments of the present application, first, the spatial position data in the fluid motion state information corresponding to each precision level is extracted, and then these data of different precision levels are connected through spatial indexing establishment technology, and finally a hierarchical association mapping relationship that can reflect the spatial correspondence between different precision data is established.

[0102] Step 115: Aggregate each fluid motion state information and the hierarchical association mapping relationship to generate a multi-precision computational fluid dynamics data set.

[0103] In this step, the aggregation operation refers to the process of merging and arranging multiple data sets into a complete data set, for generating a final multi-precision fluid dynamics data set.

[0104] In the embodiment of the present application, first, the fluid motion state information data sets of all precision levels are collected, then the data sets are integrated and packaged according to the established hierarchical correlation mapping relationship, and finally a complete multi-precision fluid dynamics data set is generated through data aggregation operation.

[0105] Through the systematic multi-precision data generation method, the embodiment of the present application first divides the target fluid domain into levels and configures corresponding calculation parameters, then generates fluid data of each precision level through parallel numerical simulation, and finally establishes hierarchical correlation and aggregates to form a complete data set, thereby realizing efficient generation and organic integration of multi-precision fluid dynamics data, providing a complete and correlated data basis for subsequent multi-level dynamic loading, and effectively solving the coordination and consistency problems in the multi-precision fluid data generation process.

[0106] In order to improve the recognition efficiency of fluid change area, this step calculates the change rate and drives the optical scanning device to mark the overrun area, and finally generates an accurate optical marking area. The present application provides a specific embodiment, step 102, optical scanning the area in the multi-precision fluid dynamics data set whose fluid motion change rate exceeds the preset change rate threshold, to generate an optical marking area, specifically including the following steps:

[0107] Step 201: Calculate the fluid motion change rate value of each adjacent spatial unit in the multi-precision fluid dynamics data set.

[0108] In this step, the spatial unit refers to the grid calculation unit in the multi-precision fluid dynamics data set, including tetrahedron or hexahedron grid forms, which is obtained based on grid division of computational fluid dynamics simulation; the fluid motion change rate value refers to a specific numerical value quantifying the degree of change of fluid motion state over time, which is obtained by calculating the change amount of fluid velocity or pressure per unit time.

[0109] In the embodiment of the present application, first, each grid unit in the multi-precision fluid dynamics data set that is adjacent in space is obtained, then the difference value of fluid motion state information between the current time step and the next time step of each grid unit is calculated, and finally the fluid motion change rate value of each adjacent spatial unit is obtained.

[0110] Step 202: Compare each fluid motion change rate value with a preset change rate threshold to identify the change rate overrun area.

[0111] In this step, the comparison operation refers to the process of determining the size of the fluid motion change rate value and the preset threshold value, which is used to determine whether a certain region belongs to a region with severe changes that requires special attention; the identification operation refers to the process of locating the specific region from all spatial units according to the results of the comparison operation, which is used to accurately find the range of the region with change rate exceeding the limit; the change rate exceeding the limit region refers to a set of continuous spatial units whose fluid motion change rate value exceeds the preset threshold value, reflecting the core region with the most severe changes in fluid motion.

[0112] In the embodiment of the present application, first, each calculated fluid motion change rate value is compared with a preset change rate threshold value one by one, then the spatial units whose fluid motion change rate values exceed the preset threshold value are screened out according to the comparison results, and finally these units are classified and identified as continuous change rate exceeding limit regions through cluster analysis.

[0113] Step 203: driving the preset micromirror array to reciprocally scan the change rate exceeding limit region to generate a light beam deflection path.

[0114] In this step, the reciprocally scanning operation refers to the coverage mode of controlling the laser beam to scan back and forth in the specific region, which is used to ensure complete optical coverage of the target region; the light beam deflection path refers to the path information describing the spatial movement trajectory in the scanning process of the laser beam, which is obtained by recording the deflection angle sequence of the micromirror array.

[0115] In the embodiment of the present application, first, a control signal is generated according to the spatial coordinates of the identified change rate exceeding limit region, then each micro mirror in the preset micromirror array is driven to deflect at a specific sequence, and finally the corresponding light beam deflection path is generated by recording the deflection trajectory of the lens.

[0116] Step 204: applying the light beam deflection path to drive the preset laser light source to cover the change rate exceeding limit region to generate an optical coverage region.

[0117] In this step, the optical coverage region refers to the physical space range covered by the laser beam scanning, which is formed by driving the laser light source to move according to the light beam deflection path.

[0118] In the embodiment of the present application, first, the generated light beam deflection path is converted into a driving signal, then the driving signal is applied to control the switching and intensity modulation of the preset laser light source, and finally the laser beam is scanned to cover the entire change rate exceeding limit region according to the predetermined path to form the optical coverage region.

[0119] Step 205: collecting the optical feedback signal of the optical coverage region to generate an optical marking region.

[0120] In this step, the optical feedback signal refers to the electrical signal converted by the sensor after receiving the optical signal reflected or scattered back by the optical coverage area, which is used to feedback the optical coverage effect.

[0121] In the embodiment of the present application, the optical signal reflected back by the optical coverage area is first received by the photosensor array, and then the optical signal is converted into an electrical signal and amplified and filtered, and finally the clear optical marking area is generated by profile extraction on the processed optical feedback signal.

[0122] In the embodiment of the present application, the optical signal reflected back by the optical coverage area is first received by the photosensor array, and then the optical signal is converted into an electrical signal and amplified and filtered, and finally the clear optical marking area is generated by profile extraction on the processed optical feedback signal.

[0123] In order to realize the accurate conversion of the optical marking to the visual focus, the optical signal is converted into a visible pattern by vectorization processing and wavefront shaping technology in this step, and finally the visual focus area is identified and generated. The present application provides a specific embodiment, step 103, converting the optical marking area into a visible light pattern, projecting the visible light pattern into a preset display plane to generate a visual focus area, specifically including the following steps:

[0124] Step 301: Vectorizing the profile of the optical marking area to generate vector profile data.

[0125] In this step, the profile refers to the external boundary shape information of the optical marking area, including the continuous curve features and geometric structure of the boundary, which is obtained based on edge extraction of the optical marking area; vectorization processing refers to the calculation process of converting continuous optical profile into mathematical description of vector path, which is used to generate machine-readable profile geometric data; vector profile data refers to the data set describing the profile geometric features in mathematical vector way, including path point coordinates and control point information, which is obtained based on vectorization processing of the profile.

[0126] In the embodiment of the present application, the boundary profile information of the optical marking area is first extracted, and then the continuous profile boundary is converted into a path described by a mathematical formula using a Bezier curve fitting algorithm, and finally the vector profile data containing all profile geometric features is generated.

[0127] Step 302: Input the vector profile data into the preset spatial light modulator to generate a phase modulation pattern.

[0128] In this step, the phase modulation pattern refers to the two-dimensional distribution pattern generated by the spatial light modulator for modulating the phase of light wave, which is formed by controlling the arrangement direction of liquid crystal molecules.

[0129] In the embodiment of the present application, the vector profile data is first converted into electrical signal format, then the electrical signals are input into the control unit of the preset spatial light modulator, and finally the corresponding light phase distribution pattern, i.e. phase modulation pattern, is generated by adjusting the orientation of liquid crystal molecules.

[0130] Step 303: applying the phase modulation pattern to perform wavefront shaping processing on the preset laser beam to generate a structured illumination beam.

[0131] In this step, the wavefront shaping processing refers to the technical process of changing the shape of the light wave front by using the phase modulation pattern, which is used to convert the ordinary laser beam into a specific structured beam; the structured illumination beam refers to the beam with a specific light field distribution after wavefront shaping processing, which is generated by the modulation of the laser beam by the phase modulation pattern.

[0132] In the embodiment of the present application, the phase modulation pattern is first placed in the preset laser beam propagation path, then the laser wavefront is phase modulated by using the pattern, and finally the structured illumination beam with a specific light field distribution is generated by changing the shape of the beam wavefront.

[0133] Step 304: projecting the structured illumination beam onto the preset display plane to generate a projection optical pattern.

[0134] In this step, the projection optical pattern refers to the light illumination distribution pattern formed by the structured illumination beam on the display plane, which is obtained by projecting the structured illumination beam onto the display plane.

[0135] In the embodiment of the present application, the structured illumination beam is first collimated and expanded, then the beam is imaged onto the preset display plane by the projection optical system, and finally a clear projection optical pattern is formed on the display plane.

[0136] Step 305: identifying the light illumination distribution characteristics of the projection optical pattern on the preset display plane to generate a visual focal point area.

[0137] In this step, the light illumination distribution characteristics refer to the spatial variation characteristics of the light intensity of the projection optical pattern on the display plane, including the brightness gradient and the contrast distribution, etc.; the identification operation refers to the technical action of analyzing and processing the light illumination distribution characteristics to extract a specific area, which is used to identify the visual focal point area from the optical pattern.

[0138] In the embodiment of the present application, the image sensor is first used to collect the projection optical pattern on the display plane, then the light intensity distribution characteristics of the pattern are analyzed, and finally the visual focal point area is generated by identifying the high-light area boundary through the threshold segmentation algorithm.

[0139] The embodiment of the present application realizes the accurate conversion from the optical mark to the visual focus by vectorizing the optical mark area and generating a phase modulation pattern, then generating a structured illumination beam by wavefront shaping and forming a projection optical pattern on the display plane, and finally generating a visual focus area by identifying the light distribution characteristics, thereby providing an accurate visual guidance basis for subsequent multi-level model dynamic loading.

[0140] In order to solve the problem of insufficient physical authenticity of low-precision data, the step is used to generate physically accurate detail enhancement data by real-time optimization and fusion processing of the neural network constrained by physical laws. The present application provides a specific embodiment, step 104, which generates detail enhancement data by real-time optimization processing of the low-fineness fluid motion state information according to the preset physical law constrained neural network, specifically including the following steps:

[0141] Step 401: Extracting key physical quantities from the low-fineness fluid motion state information.

[0142] In this step, the key physical quantity refers to the parameter extracted from the fluid motion state information, which is crucial for describing physical phenomena, including velocity, pressure, vorticity and other core physical variables, which are determined based on the basic principles of fluid mechanics.

[0143] In the embodiment of the present application, first, the physical parameters important to the physical law are identified from the low-fineness fluid motion state information, and then these parameters are separated from the original data by feature extraction technology, and finally the key physical quantities representing the essential characteristics of fluid motion are obtained.

[0144] Step 402: Inputting the key physical quantities into the preset physical law constrained neural network to generate a physical law deviation.

[0145] In this step, the physical law deviation refers to the difference between the neural network prediction result and the true physical law requirement, which is calculated by comparing the network output and the physical law constraint.

[0146] In the embodiment of the present application, first, the extracted key physical quantities are organized into a format suitable for neural network input, and then these data are input into the input layer of the preset physical law constrained neural network, and finally the output result, i.e. the physical law deviation, is calculated by the forward propagation in the network.

[0147] Step 403: Adjusting the internal parameters of the preset physical law constrained neural network according to the physical law deviation to generate an optimized neural network.

[0148] In this step, the internal parameters refer to variables in the neural network that need to be adjusted through learning, including weight matrices and bias vectors of each layer, which are initialized based on the network structure; the adjustment operation refers to the process of modifying the internal parameters of the neural network according to the error signal, which is realized by reducing the prediction error through optimization algorithms; the optimized neural network refers to the physical law constraint neural network with improved performance after parameter adjustment, which is obtained based on the internal parameter adjustment operation.

[0149] In the embodiments of the present application, the network prediction errors reflected by the physical law deviation are first analyzed, then the adjustment direction and amplitude of the network internal parameters are calculated based on these errors using the back propagation algorithm, and finally the optimization algorithm is applied to update the weight and bias parameters of the network to generate the optimized neural network.

[0150] Step 404: applying the optimized neural network to the low-fidelity fluid motion state information for resolution enhancement processing to generate preliminary enhanced data.

[0151] In this step, the resolution enhancement processing refers to the calculation process of increasing the spatial resolution and detail richness of the data, which is realized through the convolution and upsampling operations of the neural network; the preliminary enhanced data refers to the fluid motion state information after the resolution enhancement processing, which contains more detailed features than the original low-precision data.

[0152] In the embodiments of the present application, the low-fidelity fluid motion state information is first input into the optimized neural network, then the spatial resolution of the data is gradually increased through the multi-layer convolution and upsampling operations of the network, and finally the preliminary enhanced data containing more detailed information is generated.

[0153] Step 405: consistency fusion of the preliminary enhanced data and the high-fidelity fluid motion state information corresponding to the optical mark region to generate detail enhanced data.

[0154] In this step, the high-fidelity fluid motion state information refers to the data with the highest resolution and most rich physical details in the multi-precision data set, which is calculated by high-precision numerical simulation; the consistency fusion operation refers to the process of organically combining data from different sources to ensure that the fusion result maintains both physical accuracy and detail richness.

[0155] In the embodiments of the present application, the preliminary enhanced data and the high-fidelity fluid motion state information corresponding to the optical mark region are first aligned in spatial position, then the weighted fusion algorithm is used to combine the advantages of the two, and finally the detail enhanced data that maintains both physical accuracy and detail is generated through consistency verification.

[0156] The embodiment of the present application extracts key physical quantities from low-precision data and generates bias quantities by using physical law constraints on the neural network, and then optimizes the network parameters and improves the resolution of the low-precision data, and finally realizes the consistency fusion with the high-precision data, which realizes the significant improvement of the detail richness of the fluid motion data under the premise of keeping the correctness of the physical law, and provides a high-quality enhanced data source for the multi-level simulation model.

[0157] In order to improve the intelligent degree of data source allocation, the priority evaluation and consistency matching are used to generate an allocation scheme, and finally the optimal data source is determined. The present application provides a specific embodiment, step 105, combining the visual focus area and the detail enhancement data, determining the data source of the preset to-be-loaded area according to the combination result, specifically including the following steps:

[0158] Step 501: spatial priority evaluation is performed on the visual focus area to generate a region priority sequence.

[0159] In this step, the spatial priority evaluation operation refers to the process of analyzing and evaluating the importance of the visual focus area in space, including considering factors such as region position size and distance from the user's focus point; the region priority sequence refers to the list of region importance degree formed according to the spatial priority evaluation result, which is obtained through the sorting process of the evaluation operation.

[0160] In the embodiment of the present application, first, the spatial position distribution and importance of the visual focus area on the display plane are analyzed, then the priority score is calculated according to the distance between the region center and the user's viewpoint and the area size, and finally the region priority sequence is generated by sorting from high to low according to the score.

[0161] Step 502: consistency matching is performed between the region priority sequence and the detail enhancement data to generate a matching result matrix.

[0162] In this step, the consistency matching operation refers to the process of corresponding association between the region priority sequence and the detail enhancement data, which is used to establish the mapping relationship between the priority region and the enhancement data; the matching result matrix refers to a two-dimensional data table recording the corresponding relationship between the region priority and the detail enhancement data, which is generated by the consistency matching operation.

[0163] In the embodiment of the present application, first, each region in the region priority sequence is compared with the spatial distribution of the detail enhancement data, then the spatial overlap degree and the feature matching degree between each region and the enhancement data are calculated, and finally the matching result matrix reflecting the corresponding relationship between the priority and the enhancement data is generated.

[0164] Step 503: data requirement analysis is performed on the preset to-be-loaded area according to the matching result matrix to generate a data source allocation scheme.

[0165] In this step, the data requirement analysis operation refers to a processing procedure for analyzing the data characteristics required by the to-be-loaded region, including analyzing data precision format and real-time requirement, etc.; the data source allocation scheme refers to an allocation plan for specifying the data source used by each to-be-loaded region, which is formulated through the data requirement analysis operation.

[0166] In the embodiment of the present application, firstly, the data requirement characteristics of each to-be-loaded region in the matching result matrix are analyzed, then the availability and loading cost of different precision data sources are analyzed, and finally the optimal data source allocation scheme is generated according to the optimization target.

[0167] Step 504: verifying the compatibility between the data source allocation scheme and the multi-precision fluid dynamics data set, and generating a verified data source allocation scheme.

[0168] In this step, compatibility refers to the matching degree between the data source allocation scheme and the multi-precision fluid dynamics data set, including data existence and format consistency, etc.; the verification operation refers to a processing procedure for checking the feasibility of the data source allocation scheme, which is realized by comparing the scheme requirements with the actual situation of the data set; the verified data source allocation scheme refers to the allocation scheme that is confirmed to be executable after compatibility verification, which is generated through the verification operation.

[0169] In the embodiment of the present application, firstly, it is checked whether each data source specified in the data source allocation scheme exists and is available in the multi-precision fluid dynamics data set, then the compatibility of data format and precision level is verified, and finally a reliable data source allocation scheme that has passed verification is generated.

[0170] Step 505: determining the data source of the preset to-be-loaded region according to the verified data source allocation scheme.

[0171] In this step, the determination operation refers to a decision-making process for finally confirming the data source of the to-be-loaded region, which is executed based on the verified allocation scheme.

[0172] In the embodiment of the present application, firstly, the allocation instructions in the verified data source allocation scheme are read, then specific data sources are specified for each to-be-loaded region according to these instructions, and finally the determination of the data source of the entire to-be-loaded region is completed.

[0173] The embodiment of the present application realizes intelligent optimization and allocation of the data source of the to-be-loaded region by performing priority evaluation on the visual focus region, generating a matching matrix by matching with the detail enhancement data, analyzing data requirements to formulate an allocation scheme, verifying the compatibility of the allocation scheme, and finally determining the optimal data source, thereby ensuring the efficiency and reliability of the multi-level fluid simulation model loading process.

[0174] To ensure the reliability and real-time performance of the model loading, the step is to complete the real-time loading of the simulation model by parsing the instructions, verifying the data integrity and performing parallel integration. The application provides a specific embodiment, step 106, according to the data source, selecting a high-detail data set or the detail-enhanced data in the multi-precision fluid dynamics data set to complete the real-time loading of the multi-level fluid simulation model with the selection result, specifically including the following steps:

[0175] Step 601: Parsing the loading instructions in the data source to generate a data loading instruction set.

[0176] In this step, the loading instruction refers to a control command containing data source selection and loading parameters, including data source identifiers and loading ranges and other elements, which is generated based on a data source allocation scheme; the parsing operation refers to a processing process of decomposing and understanding the loading instructions, which is realized through syntax analysis and semantic recognition to standardize the instructions; the data loading instruction set refers to a standardized instruction set generated after parsing, which contains complete data loading parameters and execution order.

[0177] In the embodiment of the application, the original loading instructions contained in the data source are first read, and then syntax analysis and semantic parsing are performed on these instructions to identify the operation types and data source identifiers in the instructions, and finally a standardized data loading instruction set in a standardized format is generated.

[0178] Step 602: According to the data loading instruction set, selecting a high-detail data set or the detail-enhanced data in the multi-precision fluid dynamics data set, and extracting a plurality of to-be-loaded data blocks from the selection result.

[0179] In this step, the selection result refers to a specific data source object determined according to the data loading instruction set, which is obtained through a data source selection operation; the to-be-loaded data block refers to a discrete data unit to be processed extracted from the selected data source, which is obtained through a data extraction operation.

[0180] In the embodiment of the application, first, the type of data source to be accessed is determined according to the data loading instruction set, then the corresponding data source is selected from the high-detail data set or the detail-enhanced data in the multi-precision fluid dynamics data set, and finally a plurality of to-be-loaded data blocks are extracted from the selected data source according to the spatial region.

[0181] Step 603: Integrity verification is performed on each of the to-be-loaded data blocks to generate a plurality of verified to-be-loaded data blocks.

[0182] In this step, the integrity verification operation refers to a processing process of checking whether the data block is complete and available, which is realized through checksum comparison and data integrity checking; the verified to-be-loaded data block refers to a data block unit that passes the integrity check, which has an integrity guarantee flag.

[0183] In the embodiment of the present application, the checksums of each to-be-loaded data block are first calculated, then the checksums are compared with the pre-stored correct values, and finally the data blocks that pass the verification are marked to generate a plurality of verified to-be-loaded data blocks.

[0184] Step 604: The verified to-be-loaded data blocks are integrated in parallel to generate integrated simulation model data.

[0185] In this step, the parallel integration operation refers to the integration process of simultaneously processing a plurality of data blocks, and the integrated simulation model data refers to the complete simulation data after integration processing, which contains seamless connection data of all to-be-loaded areas.

[0186] In the embodiment of the present application, the verified to-be-loaded data blocks are first grouped according to spatial positions, then the spatial registration and edge processing of a plurality of data blocks are simultaneously processed by using parallel computing technology, and finally the integrated simulation model data with seamless connection is generated.

[0187] Step 605: The integrated simulation model data is input into a preset real-time rendering pipeline to complete real-time loading of the multi-level fluid simulation model.

[0188] In this step, the preset real-time rendering pipeline refers to a pre-configured graphics rendering process, including standard stages such as vertex processing, rasterization and pixel shading.

[0189] In the embodiment of the present application, the integrated simulation model data is first converted into a format recognizable by a rendering engine, then the data is submitted to the vertex processing and pixel shading stages of the preset real-time rendering pipeline, and finally the real-time loading and visual presentation of the multi-level fluid simulation model are completed.

[0190] The embodiment of the present application generates a standardized instruction set by analyzing the loading instructions, selects data sources and extracts data blocks according to the standardized instruction set, integrates the data blocks into complete model data in parallel after integrity verification, and finally completes the loading through the real-time rendering pipeline, thereby realizing efficient and reliable loading of the multi-level fluid simulation model and ensuring the smoothness of the simulation process and the real-time performance of the visual effect.

[0191] Figure 2 A structural schematic diagram of a multi-level parallel world simulation model dynamic loading system is provided for the embodiment of the present application, as shown in FIG. 1. Figure 2 The system comprises:

[0192] An acquisition module 21 is configured to acquire a multi-precision fluid dynamics data set, wherein the multi-precision fluid dynamics data set contains fluid motion state information of different levels of detail.

[0193] The scanning module 22 is configured to perform optical scanning on the region in the multi-precision fluid dynamics data set where the fluid motion change rate exceeds the preset change rate threshold, and generate an optical marked region.

[0194] The projection module 23 is configured to convert the optical marked region into a visible light pattern, project the visible light pattern into a preset display plane, and generate a visual focus region.

[0195] The optimization module 24 is configured to perform real-time optimization processing on the fluid motion state information of a low level of detail according to a preset physical law constraint neural network, and generate detail enhancement data.

[0196] The combination module 25 is configured to combine the visual focus region and the detail enhancement data, and determine a data source of a preset to-be-loaded region according to a combination result.

[0197] The selection module 26 is configured to select a high-detail data set in the multi-precision fluid dynamics data set or the detail enhancement data according to the data source, and complete real-time loading of the multi-level fluid simulation model by using a selection result.

[0198] Figure 2 The multi-level parallel world simulation model dynamic loading system can perform the following operations Figure 1 The multi-level parallel world simulation model dynamic loading method has the implementation principle and technical effects as described above, and will not be described in detail here. The specific operation manner of each module and unit in the multi-level parallel world simulation model dynamic loading system has been described in detail in the embodiments of the method, and will not be described in detail here.

[0199] In one possible design, Figure 2 The multi-level parallel world simulation model dynamic loading system can be implemented as a computing device, such as a computer. Figure 3 As shown in the figure, the computing device can include a storage component 31 and a processing component 32.

[0200] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32.

[0201] The processing component 32 is configured to: acquire a multi-precision fluid dynamics data set containing fluid motion state information of different levels of detail; perform optical scanning on a region in the multi-precision fluid dynamics data set where the rate of change of fluid motion exceeds a preset rate of change threshold, to generate an optical marked region; convert the optical marked region into a visible light pattern, project the visible light pattern into a preset display plane, and generate a visual focal point region; perform real-time optimization processing on the fluid motion state information of a low level of detail according to a preset physical law constraint neural network, to generate detail enhancement data; combine the visual focal point region and the detail enhancement data, and determine a data source of a preset to-be-loaded region according to a combination result; and select a high-detail data set in the multi-precision fluid dynamics data set or the detail enhancement data according to the data source, to complete real-time loading of a multi-level fluid simulation model by using a selection result.

[0202] The processing component 32 can include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component can also be one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic elements for executing the above method.

[0203] The storage component 31 is configured to store various types of data to support the operation of the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0204] Of course, the computing device can also include other components, such as an input / output interface, a display component, a communication component, etc.

[0205] The input / output interface provides an interface between the processing component and peripheral interface modules, which can be output devices, input devices, etc.

[0206] The communication component is configured to facilitate wired or wireless communication between the computing device and other devices, etc.

[0207] The computing device can be a physical device or an elastic computing host provided by a cloud computing platform, and the computing device can be a cloud server, and the processing component and the storage component can be basic server resources rented or purchased from the cloud computing platform.

[0208] The embodiment of the present application further provides a computer storage medium storing a computer program, and the computer program can realize the above-mentioned method when being executed by a computer. Figure 1 The embodiment of the present application further provides a computer storage medium storing a computer program, and the computer program can realize the above-mentioned method when being executed by a computer.

[0209] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-mentioned system, device and unit can refer to the corresponding processes in the foregoing method embodiments, and will not be described here.

[0210] The device embodiments described above are only schematic, and the units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, they can be located in one place, or can be distributed on multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiment scheme. Those skilled in the art can understand and implement without creative labor.

[0211] Through the description of the foregoing embodiments, those skilled in the art can clearly understand that each embodiment can be realized by means of software and necessary general hardware platform, and of course, can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of software product, which can be stored in a computer readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes a plurality of instructions to make a computer device (which can be a personal computer, server, or network device, etc.) execute the method described in each embodiment or some part of the embodiment.

[0212] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A dynamic loading method for a multi-level parallel world simulation model, characterized in that, include: Acquire a multi-precision fluid dynamics data set, which contains fluid motion state information with different levels of precision; Optical scanning is performed on regions in the multi-precision fluid dynamics dataset where the rate of change of fluid motion exceeds a preset rate of change threshold to generate optically marked regions. The optical marking area is converted into a visible light pattern, and the visible light pattern is projected onto a preset display plane to generate a visual focal area; The neural network is constrained by preset physical laws to perform real-time optimization processing on the low-precision fluid motion state information to generate detailed enhanced data. The visual focus area and the detail enhancement data are combined, and the data source of the preset area to be loaded is determined based on the combination result; Based on the data source, a high-detail dataset or the detail-enhanced data set from the multi-precision fluid dynamics dataset is selected to utilize the selection result for real-time loading of the multi-level fluid simulation model.

2. The method according to claim 1, characterized in that, Acquire a multi-precision fluid dynamics dataset, which contains fluid motion state information at different levels of detail, including: The preset target fluid domain is divided to obtain fluid domains with multiple accuracy levels. Assign a corresponding preset set of simulation parameters to each of the partitioned fluid domains, and generate multiple parameterized fluid domains corresponding to each of the partitioned fluid domains; Numerical simulation calculations are performed using the parameterized fluid domains described above to generate fluid motion state information corresponding to each of the accuracy levels described above. Connect the fluid motion state information to establish a hierarchical association mapping relationship; The fluid motion state information and the hierarchical association mapping relationship are aggregated to generate a multi-precision fluid dynamics data set.

3. The method according to claim 1, characterized in that, Optical scanning is performed on regions in the multi-precision fluid dynamics dataset where the rate of change of fluid motion exceeds a preset rate of change threshold to generate optically marked regions, including: Calculate the rate of change of fluid motion in each adjacent spatial unit in the multi-precision fluid dynamics dataset; The numerical values ​​of the fluid motion change rate are compared with a preset change rate threshold to identify the region where the change rate exceeds the limit. The preset micromirror array is driven to perform reciprocating optical scanning on the region where the rate of change exceeds the limit, thereby generating a beam deflection path; The beam deflection path is used to drive a preset laser source to cover the region where the rate of change exceeds the limit, thereby generating an optical coverage area. The optical feedback signal of the optical coverage area is collected to generate an optically marked area.

4. The method according to claim 1, characterized in that, Converting the optical marking area into a visible light pattern, and projecting the visible light pattern onto a preset display plane to generate a visual focal area, includes: The contour of the optical marking region is vectorized to generate vector contour data; The vector contour data is input into a preset spatial light modulator to generate a phase modulation pattern; The phase modulation pattern is applied to the preset laser beam to perform wavefront shaping to generate a structured illumination beam. The structured illumination beam is projected onto a preset display plane to generate a projected optical pattern; The illumination distribution characteristics of the projected optical pattern on the preset display plane are identified to generate a visual focal area.

5. The method according to claim 1, characterized in that, The neural network, constrained by preset physical laws, performs real-time optimization processing on the low-resolution fluid motion state information to generate enhanced detail data, including: Extract key physical quantities from the fluid motion state information at a low level of detail; The key physical quantities are input into a preset physical law constraint neural network to generate physical law deviation quantities; The internal parameters of the preset physical law-constrained neural network are adjusted according to the deviation of the physical law to generate an optimized neural network. The optimized neural network is applied to perform resolution enhancement processing on the low-resolution fluid motion state information to generate preliminary enhanced data; The preliminary enhancement data is consistently fused with the high-precision fluid motion state information corresponding to the optically marked area to generate detailed enhancement data.

6. The method according to claim 1, characterized in that, The visual focus area and the detail enhancement data are combined, and the data source for the preset region to be loaded is determined based on the combination result, including: Spatial priority evaluation is performed on the visual focus region to generate a region priority sequence; The region priority sequence is matched with the detail enhancement data to generate a matching result matrix. Based on the matching result matrix, a data demand analysis is performed on the preset region to be loaded, and a data source allocation scheme is generated; The compatibility between the data source allocation scheme and the multi-precision fluid dynamics dataset is verified, and a verified data source allocation scheme is generated. The data source for the preset region to be loaded is determined based on the verified data source allocation scheme.

7. The method according to claim 1, characterized in that, Based on the data source, a high-detail dataset or the enhanced detail data is selected from the multi-precision fluid dynamics dataset to utilize the selection result for real-time loading of a multi-level fluid simulation model, including: The loading instructions in the data source are parsed to generate a data loading instruction set; Based on the data loading instruction set, select the high-detail data set or the detail-enhanced data set from the multi-precision fluid dynamics data set, and extract multiple data blocks to be loaded from the selection result; Perform integrity verification on each of the data blocks to be loaded, and generate multiple verified data blocks to be loaded. The verified data blocks to be loaded are integrated in parallel to generate integrated simulation model data. The integrated simulation model data is input into a preset real-time rendering pipeline to complete the real-time loading of the multi-level fluid simulation model.

8. A dynamic loading system for a multi-level parallel world simulation model, characterized in that, include: The acquisition module is used to acquire a multi-precision fluid dynamics data set, which contains fluid motion state information with different levels of precision. The scanning module is used to perform optical scanning on regions in the multi-precision fluid dynamics dataset where the rate of change of fluid motion exceeds a preset rate of change threshold, and generate optically marked regions. The projection module is used to convert the optical marking area into a visible light pattern and project the visible light pattern onto a preset display plane to generate a visual focal area. The optimization module is used to constrain the neural network according to preset physical laws to perform real-time optimization processing on the low-precision fluid motion state information and generate detailed enhanced data. The module combines the visual focus area and the detail enhancement data, and determines the data source of the preset area to be loaded based on the combination result. The selection module is used to select either the high-detail dataset or the enhanced detail data from the multi-precision fluid dynamics dataset based on the data source, so as to use the selection result to complete the real-time loading of the multi-level fluid simulation model.

9. A computing device, characterized in that, It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to implement a dynamic loading method for a multi-level parallel world simulation model as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, The system contains a computer program that, when executed by a computer, implements a method for dynamically loading a multi-level parallel world simulation model as described in any one of claims 1 to 7.

Images