Three-dimensional construction method and system for bioaerosol simulation scene

Abstract

The invention discloses a three-dimensional construction method and system for a bioaerosol simulation scene, and relates to the technical field of modeling, and the method comprises the steps: recognizing a first training scene; configuring three-dimensional simulation parameters; performing three-dimensional simulation on the first training scene by adopting a three-dimensional simulation system, and outputting a first simulation scene; performing dynamic particle tracking on the first simulation scene, and identifying dynamic particle information; a simulation effect influence recognition model is established, the dynamic particle information is input into the model for analysis, identification dynamic particles are obtained, and the identification dynamic particles are particles with influence indexes larger than a first preset influence index; and establishing a granularity adjuster, performing feedback connection with the simulation effect influence identification model, performing fine granularity processing on the identification dynamic particles according to the granularity adjuster, and optimizing the first simulation scene. Therefore, the technical effect is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of modeling, and in particular to a three-dimensional construction method and system for bioaerosol simulation scenes. Background Art

[0002] Bioaerosols refer to microbial particles suspended in the air, such as bacteria, viruses, pollen and fungal spores. Existing technologies usually simulate and analyze bioaerosols by establishing simulation scene models in specific scenarios. When constructing simulation scene models, it is necessary to consider tracking a large number of particles. The configuration of the number and size of particles greatly affects the effect of simulation analysis and computing power cost. Existing construction methods have technical problems such as rough particle configuration and affecting simulation accuracy and efficiency. Summary of the invention

[0003] The present invention provides a three-dimensional construction method and system for bioaerosol simulation scenes to solve the technical problems in the prior art of rough particle configuration and affecting simulation accuracy and performance, and achieve the technical effects of improving simulation performance and enhancing simulation accuracy.

[0004] In a first aspect, the present invention provides a three-dimensional construction method for a bioaerosol simulation scene, wherein the method comprises: Identify a first training scene for bioaerosols; configure three-dimensional simulation parameters of the first training scene, the three-dimensional simulation parameters including spatial scale simulation parameters, time step simulation parameters and particle number simulation parameters; use a three-dimensional simulation system to perform three-dimensional simulation on the first training scene with the three-dimensional simulation parameters, and output a first simulation scene; perform dynamic particle tracking on the first simulation scene, and identify dynamic particle information, including the object to which the dynamic particles belong, the distribution of the number of dynamic particles, and the step length of dynamic particle changes; establish a simulation effect impact recognition model, input the object to which the dynamic particles belong, the distribution of the number of dynamic particles, and the step length of dynamic particle changes into the simulation effect impact recognition model for analysis, and obtain identified dynamic particles, wherein the identified dynamic particles are dynamic particles whose impact index is greater than a first preset impact index; establish a granularity adjuster, the granularity adjuster is feedback-connected to the simulation effect impact recognition model, and fine-grained processing is performed on the identified dynamic particles according to the granularity adjuster to optimize the first simulation scene.

[0005] In a second aspect, the present invention also provides a three-dimensional construction system for a bioaerosol simulation scene, wherein the system comprises: A scene recognition module, the scene recognition module is used to identify a first training scene for bioaerosols; a simulation parameter configuration module, the simulation parameter configuration module is used to configure the three-dimensional simulation parameters of the first training scene, the three-dimensional simulation parameters include spatial scale simulation parameters, time step simulation parameters and particle number simulation parameters; a scene simulation module, the scene simulation module is used to use a three-dimensional simulation system to perform a three-dimensional simulation of the first training scene with the three-dimensional simulation parameters, and output a first simulation scene; a dynamic tracking module, the dynamic tracking module is used to perform dynamic particle tracking on the first simulation scene, identify dynamic particle information, including the object to which the dynamic particle belongs, dynamic Particle quantity distribution, dynamic particle change step length; an impact identification module, the impact identification module is used to establish a simulation effect impact identification model, the object to which the dynamic particle belongs, the dynamic particle quantity distribution, and the dynamic particle change step length are input into the simulation effect impact identification model for analysis, and an identified dynamic particle is obtained, wherein the identified dynamic particle is a dynamic particle whose impact index is greater than a first preset impact index; a scene optimization module, the scene optimization module is used to establish a granularity adjuster, the granularity adjuster is feedback-connected to the simulation effect impact identification model, the identified dynamic particle is fine-grainedly processed according to the granularity adjuster, and the first simulation scene is optimized.

[0006] The invention discloses a three-dimensional construction method and system for a bioaerosol simulation scene, comprising: identifying a first training scene for bioaerosol; configuring three-dimensional simulation parameters of the first training scene, the three-dimensional simulation parameters including a spatial scale simulation parameter, a time step simulation parameter and a particle number simulation parameter; using a three-dimensional simulation system to perform three-dimensional simulation on the first training scene with the three-dimensional simulation parameters, and outputting the first simulation scene; performing dynamic particle tracking on the first simulation scene, and identifying dynamic particle information, including the object to which the dynamic particles belong, the distribution of the number of dynamic particles, and the step length of the dynamic particle change; establishing a simulation effect impact recognition model, inputting the object to which the dynamic particles belong, the distribution of the number of dynamic particles, and the step length of the dynamic particle change into the simulation effect impact recognition model for analysis, and obtaining identified dynamic particles, wherein the identified dynamic particles are dynamic particles whose impact index is greater than a first preset impact index; establishing a granularity adjuster, the granularity adjuster is feedback-connected with the simulation effect impact recognition model, and fine-grained processing is performed on the identified dynamic particles according to the granularity adjuster to optimize the first simulation scene. The three-dimensional construction method and system for bioaerosol simulation scenes disclosed in the present invention solve the technical problems of rough particle configuration and influence on simulation accuracy and performance, and achieve the technical effects of improving simulation performance and enhancing simulation accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 It is a schematic diagram of the process of the three-dimensional construction method of the present invention for bioaerosol simulation scene.

[0008] Figure 2 It is a schematic diagram of the structure of the three-dimensional construction system for bioaerosol simulation scene of the present invention.

[0009] Explanation of the reference numerals: scene recognition module 11 , simulation parameter configuration module 12 , scene simulation module 13 , dynamic tracking module 14 , impact recognition module 15 , scene optimization module 16 . DETAILED DESCRIPTION

[0010] The technical solution provided in the embodiments of the present invention is to solve the technical problems of the prior art that the particle configuration is rough and affects the simulation accuracy and efficiency. The overall idea adopted is as follows: First, identify the first training scenario for bioaerosol drills. Then, configure three-dimensional simulation parameters for the first training scenario, including simulation parameters of spatial scale, time step, and number of particles. Next, use a three-dimensional simulation system to simulate the first training scenario according to these three-dimensional simulation parameters and generate a first simulation scenario. Perform dynamic particle tracking on the first simulation scenario to identify the information of dynamic particles, including the source object, quantity distribution, and change step of the particles. Establish a simulation effect impact identification model, and input information such as the source object, quantity distribution, and change step of the particles into the model for analysis to identify the identified dynamic particles whose impact indicators exceed the preset threshold. Next, establish a granularity adjuster and connect it to the simulation effect impact identification model for feedback. Use the granularity adjuster to perform more detailed processing on the identified dynamic particles to optimize the first simulation scenario.

[0011] The above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods of the specification to better understand the above technical solution. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments used only to explain the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. In addition, it should be noted that, for the convenience of description, only the parts related to the present invention are shown in the drawings, rather than all of them. Example 1

[0012] Figure 1 The present invention is a schematic flow chart of a three-dimensional construction method for a bioaerosol simulation scene, which includes: Identify the first training scenario for bioaerosols.

[0013] Optionally, the training scene refers to a real space scene where bioaerosol simulation is required, and the training scene is associated with a corresponding training plan. Exemplarily, the training scene includes indoor scenes (such as hospitals, subway stations, office buildings, etc.) and outdoor scenes (such as city squares, stadiums, open-air markets, etc.).

[0014] Specifically, the first training scene is any training item in the training plan corresponding to the target training scene. In other words, the first training scene can be a scene corresponding to any sub-item or any multiple sub-items in the training plan under the target training scene, and the first training scene has both spatial characteristics and time range characteristics.

[0015] Among them, bioaerosol refers to biological particles suspended in the air, including bacteria, viruses, fungal spores, pollen, animal dander, etc. Different types of bioaerosols have different characteristics, including particle size distribution, concentration, transmission mode, survival time, etc.

[0016] Optionally, identifying a first training scenario for bioaerosols includes determining training scenario requirements, such as scenario objectives and application areas; obtaining training scenario elements of the first training scenario, such as spatial layout, environmental conditions, activity types, etc. Exemplarily, the spatial layout includes the spatial structure, enclosure, ventilation system structure, etc. of the site. Meteorological conditions include factors such as temperature, humidity, wind speed, and wind direction. Activity types include the activity level of space personnel (such as static and dynamic activities), personnel density, etc.

[0017] The three-dimensional simulation parameters of the first training scenario are configured, wherein the three-dimensional simulation parameters include a spatial scale simulation parameter, a time step simulation parameter, and a particle number simulation parameter.

[0018] Optionally, based on the recognition result of the first training scene, three-dimensional simulation parameters of the first training scene are extracted and configured, including determining spatial scale simulation parameters, time step simulation parameters and particle number simulation parameters.

[0019] Specifically, the spatial scale simulation parameters refer to the size and resolution of the simulation area, including length, width, height and boundary strips. Exemplarily, determining the spatial scale simulation parameters includes: determining the size of the three-dimensional simulation area, such as length, width and height, according to the actual spatial layout of the training scene; selecting an appropriate spatial resolution to ensure the simulation accuracy, and the units of spatial resolution include meters or centimeters, etc.; setting the boundary conditions of the simulation area, including the attributes of walls, doors, windows, vents and other locations, to reflect the impact of actual physical boundaries on aerosol propagation.

[0020] Specifically, the time step simulation parameter is used to define the time interval of each step in the simulation process and the total simulation time. Exemplarily, determining the time step simulation parameter includes: selecting an appropriate time step, such as seconds or milliseconds, according to the simulation target and the typical diffusion rate of aerosols. A smaller time step can improve the accuracy of the simulation, but will increase the amount of calculation; based on the scenario requirements of the first training scenario, determining the total simulation time to reflect the propagation and deposition process of aerosols in the training scenario.

[0021] Specifically, the particle number simulation parameters define the number and size distribution of aerosol particles used for simulation. Exemplarily, configuring the particle number simulation parameters includes: selecting the number of aerosol particles used for simulation in combination with computing resource limitations. A larger number of particles helps ensure the statistical reliability of the simulation results, and correspondingly increases the computing power requirements of the simulation process. Set the particle size distribution of aerosol particles to reflect the size range and proportion of actual aerosol particles. Specifically, the particle size distribution is determined based on experimental data or literature.

[0022] By setting appropriate three-dimensional simulation parameters, a data basis is provided for the subsequent accurate simulation of the diffusion and sedimentation process of bioaerosols in training scenarios and the assessment of aerosol transmission risks in different scenarios.

[0023] A three-dimensional simulation system is used to perform three-dimensional simulation on the first training scene with the three-dimensional simulation parameters, and a first simulation scene is output.

[0024] Optionally, select suitable three-dimensional simulation software as the above-mentioned three-dimensional simulation system, such as computational fluid dynamics (CFD) software, and import the configured three-dimensional simulation parameters (spatial scale simulation parameters, time step simulation parameters and particle number simulation parameters) into the three-dimensional simulation system; then, set the initial conditions of the simulation environment in the simulation system, including meteorological conditions such as temperature, humidity, wind speed and wind direction; initial particle state conditions such as the initial position and concentration distribution of aerosol particles, to ensure that the conditions at the beginning of the simulation are consistent with the actual training scenario.

[0025] Furthermore, based on the initialized three-dimensional simulation system, the three-dimensional simulation of the first training scene is started. The three-dimensional simulation system will calculate the aerosol diffusion path, concentration change, deposition position and other information according to the input simulation parameters and initial conditions. When the three-dimensional simulation reaches the time step simulation parameter, the aerosol diffusion trajectory, concentration distribution and deposition position and other information automatically collected during the simulation are output as the output of the first simulation scene.

[0026] Dynamic particle tracking is performed on the first simulation scene to identify dynamic particle information, including the object to which the dynamic particles belong, the number distribution of dynamic particles, and the step size of dynamic particle changes.

[0027] Specifically, in the three-dimensional simulation process, the more complex the scene and the more obstacles there are, the more complex the corresponding airflow state will be, and a larger number of particles will be required to ensure the accuracy of the simulation results.

[0028] Optionally, based on the first simulation scenario, dynamic particle tracking is performed on the obtained simulation results to obtain the motion trajectory and state changes of the particles in the fluid, and generate dynamic particle information. Specifically, according to the initial position and motion trajectory of the particles, the object or source to which the particles belong is identified, such as a specific pollution source or equipment, to achieve dynamic particle identification. The number of particles in different regions within each time step is counted, a particle density map is generated, the concentration changes of the particles are displayed, and the dynamic particle number distribution is obtained. The displacement changes of the particles within each time step are analyzed, and the velocity changes of the particles (such as instantaneous acceleration or average acceleration) are obtained to evaluate their dynamic behavior, and the dynamic particle change step is calculated based on the obtained dynamic behavior information.

[0029] Optionally, particle tracking algorithms such as Lagrangian method and Euler method are used to ensure the accuracy of particle motion trajectory, and combined with parallel computing technology to improve the calculation speed of particle tracking, which helps to improve the processing capacity of the system under multiple obstacles and complex airflow conditions. Through dynamic particle tracking, the motion trajectory and concentration changes of particles can be accurately obtained, providing accurate basic information for subsequent analysis and optimization.

[0030] In some embodiments, performing dynamic particle tracking on the first simulation scene to identify dynamic particle information includes: The first simulation scene is divided into regions according to scene objects to obtain multiple sub-regions, wherein each sub-region corresponds to a scene object; for the multiple sub-regions, multiple particle monitoring blocks are constructed, each particle monitoring block is used to monitor dynamic change data of particles in the corresponding sub-region; based on the multiple particle monitoring blocks, multiple particle monitoring data sets are generated; and particle dynamic analysis is performed on the multiple particle monitoring data sets to identify dynamic particle information.

[0031] Specifically, dynamic particle tracking is performed on the first simulation scene based on the particle tracking algorithm. First, the first simulation scene is divided into regions according to the scene objects to obtain multiple sub-regions. Each sub-region corresponds to a scene object, such as equipment, walls, channels, etc. Then, multiple particle monitoring blocks are constructed for the multiple sub-regions. Each particle monitoring block is used to monitor the dynamic change data of particles in the corresponding sub-region. Specifically, constructing multiple particle monitoring blocks includes setting the boundaries and sizes of each monitoring block, while ensuring that all sub-regions are covered and there is no overlap between blocks. Then, according to the multiple particle monitoring blocks, the dynamic change data of particles in the corresponding area are collected to generate multiple particle monitoring data sets, each data set corresponds to a monitoring block, and the multiple particle monitoring data sets reflect the dynamic change characteristics of particles in the multiple particle monitoring blocks.

[0032] Furthermore, a particle tracking algorithm is used to perform particle dynamic analysis on multiple particle monitoring data sets to identify dynamic particle information. For example, the dynamic particle information includes particle number distribution, particle velocity change, particle concentration change, and the like.

[0033] Optionally, the dynamic particle information is visualized in three dimensions to intuitively present the particle's motion trajectory and concentration distribution, thereby facilitating the understanding of the particle diffusion process.

[0034] A simulation effect impact identification model is established, and the object to which the dynamic particle belongs, the number distribution of the dynamic particles, and the change step of the dynamic particles are input into the simulation effect impact identification model for analysis to obtain an identified dynamic particle, wherein the identified dynamic particle is a dynamic particle whose impact index is greater than a first preset impact index.

[0035] The simulation effect impact identification model is used to evaluate the simulation effect impact of multiple particles based on the dynamic particle information, and identify the particles that have a significant impact on the simulation effect. Specifically, the simulation effect impact identification model takes the object to which the dynamic particle belongs, the number distribution of the dynamic particles, and the dynamic particle change step as input data, and outputs an identified dynamic particle, that is, a dynamic particle whose impact index is greater than a first preset impact index.

[0036] In some embodiments, establishing a simulation effect impact recognition model includes: A training sample data set is obtained, wherein the training sample data set includes samples of objects to which dynamic particles belong, samples of dynamic particle quantity distribution, samples of dynamic particle change step lengths, and identification information of identification influence indicators; an association influence network is established, and an association weight parameter of the association influence network is initialized; and the samples of objects to which the dynamic particles belong, samples of dynamic particle quantity distribution, samples of dynamic particle change step lengths, and identification information of identification influence indicators are trained with the association influence network until the accuracy reaches a preset accuracy, and the simulation effect influence recognition model is converged and output.

[0037] Optionally, a simulation effect impact identification model is established, and the simulation effect impact identification model can be a statistical model (such as multivariate regression analysis) or a machine learning model (such as random forest, support vector machine).

[0038] Optionally, a simulation effect impact identification model is established, including data collection and preprocessing, establishing an associated impact network, training a model, and evaluating and applying the model. Specifically, first collect and prepare a training sample data set for training the model, the training sample data set includes a sample of the object to which the dynamic particles belong, a sample of the number distribution of dynamic particles, a sample of the step length of dynamic particle changes, and identification information of the identification impact index. Then, design an associated impact network to map the input features to the output impact index. Exemplarily, the associated impact network is constructed based on a neural network structure, such as a multi-layer perceptron (MLP), a convolutional neural network (CNN), or a recurrent neural network (RNN), comprising an input layer, multiple hidden layers, and an output layer. At the same time, the associated weight parameters of the associated impact network are initialized based on a random initialization method.

[0039] Furthermore, the object samples to which the dynamic particles belong, the dynamic particle quantity distribution samples, the dynamic particle change step samples, and the identification information of the identification influence index are input into the association influence network, and the predicted influence index is calculated through the network, and the loss between the predicted influence index and the actual identification influence index is calculated. Then, the association weight parameters of the network are adjusted through the back propagation algorithm to minimize the loss function, and the training process is continuously iterated until the accuracy of the association influence network reaches the preset accuracy requirement. Specifically, when the training loss of the association influence network stabilizes at a low level and the accuracy on the validation set reaches the preset accuracy, it is considered that the association influence network has converged, and the trained association influence network is saved as a simulation effect influence recognition model for subsequent particle influence analysis.

[0040] Optionally, data enhancement technology is used to increase the diversity of training samples and improve the generalization ability of the association influence network; during the training process, the cross-validation method is used to evaluate the performance of the model, and the L1 or L2 regularization method is used to prevent the model from overfitting and improve the generalization ability of the model.

[0041] By building an associated influence network and training a simulation effect impact identification model, it is helpful to accurately identify the key dynamic particles that have a significant impact on the simulation effect, thereby improving the subsequent optimization efficiency.

[0042] In some embodiments, establishing the simulation effect impact recognition model further includes: The object samples to which the dynamic particles belong are divided to obtain first-category object samples and second-category object samples, wherein the first-category objects represent objects whose reactive bioaerosols have changed, and the second-category objects represent objects whose non-reactive bioaerosols have changed; according to the first-category object samples and the second-category object samples, a first penalty hyperparameter and a second penalty hyperparameter are configured, wherein the first penalty hyperparameter is greater than the second penalty hyperparameter; a penalty feedback network is constructed with the first penalty hyperparameter and the second penalty hyperparameter, and the simulation effect impact identification model is optimized with the penalty feedback network.

[0043] Optionally, the object samples to which the dynamic particles belong are divided into two categories, and the first category of object samples characterizes objects that react to changes in bioaerosols. In other words, when bioaerosols change, objects that are greatly affected are the first category of object samples. The second category of object samples characterizes objects that do not react to changes in bioaerosols. When bioaerosols change, the second category of object samples are less affected or not affected.

[0044] Optionally, penalty hyperparameters are configured for the first type of object samples and the second type of object samples, respectively, wherein the first penalty hyperparameter assigned to the first type of object samples is larger, indicating that the error penalty for such samples is larger, and is used to enhance the sensitivity of the model to objects that react to bioaerosol changes. The second penalty hyperparameter assigned to the second type of object samples is smaller than the first penalty hyperparameter, indicating that the error penalty for such samples is smaller, and is used to reduce the attention paid to non-reacting objects.

[0045] Furthermore, a penalty feedback network is constructed, which is used to introduce a penalty mechanism into the simulation effect impact recognition model to help the simulation effect impact recognition model better balance the degree of attention paid to the first category of object samples and the second category of object samples.

[0046] Specifically, the first penalty hyperparameter and the second penalty hyperparameter are input into the penalty feedback network to initialize the penalty feedback network. Then, in the training process of the simulation effect impact recognition model, the penalty loss of each sample is calculated based on the penalty feedback network, and the parameters of the simulation effect impact recognition model are adjusted in combination with the penalty loss to make the model more sensitive to the errors of the first category of object samples, thereby optimizing the model performance.

[0047] In some implementations, after obtaining the first type of object samples, the method further includes: Perform secondary classification on the first category object samples to obtain first category direct objects and first category indirect objects; wherein the first category direct objects are objects that characterize changes in direct reaction bioaerosols, and the first category indirect objects are objects that characterize changes in indirect reaction bioaerosols; configure first direct penalty hyperparameters and first indirect penalty hyperparameters corresponding to the first category direct objects and the first category indirect objects, respectively; wherein the first direct penalty hyperparameter is greater than the first indirect penalty hyperparameter, which is greater than the second penalty hyperparameter.

[0048] Optionally, based on the degree of directness of the relationship with the bioaerosol change, the first-category object samples are secondary classified into first-category direct objects and first-category indirect objects. The first-category direct objects are objects that directly respond to bioaerosol changes, such as the aerosol colloid itself and its diffusion concentration. The first-category indirect objects are objects that indirectly respond to bioaerosol changes, such as objects that cause aerosol changes due to other environmental factors (temperature, convection, etc.).

[0049] Optionally, a first direct penalty hyperparameter is configured for the first type of direct object, and a first indirect penalty hyperparameter is configured for the first type of indirect object, wherein the first direct penalty hyperparameter is greater than the first indirect penalty hyperparameter, and the first indirect penalty hyperparameter is greater than the second penalty hyperparameter. Through the above-mentioned parameter size differentiation configuration, it is ensured that the error penalty for the first type of direct object is large, the sensitivity of the model to the direct response of aerosol changes is strengthened, the error penalty for the first type of indirect object is moderate, so as to pay moderate attention to the objects whose aerosol changes are caused by environmental influences, and the error penalty for the second type of object is minimized, thereby avoiding excessive attention to non-responsive objects.

[0050] By performing secondary classification on the first-class object samples, objects that directly and indirectly respond to changes in bioaerosols can be more accurately identified. Different penalty hyperparameters are then assigned to different object samples to improve the efficiency and accuracy of the model when processing dynamic particle data.

[0051] A granularity adjuster is established, the granularity adjuster is feedback-connected with the simulation effect impact identification model, and the identified dynamic particles are fine-grainedly processed according to the granularity adjuster to optimize the first simulation scene.

[0052] Optionally, a granularity adjuster is used to adjust the granularity of the identified dynamic particles, thereby optimizing the three-dimensional simulation parameters of the first training scene, wherein the granularity adjuster is feedback-connected to the simulation effect impact identification model, and then the optimized three-dimensional simulation parameters of the first training scene are fed back to the simulation effect impact identification model to achieve iterative optimization of the three-dimensional simulation parameters of the first training scene.

[0053] Exemplarily, to establish a granularity adjuster, first, initialize the granularity adjuster and configure the basic parameters of the granularity adjuster, including the granularity level and the adjustment range. Then, the granularity adjuster is feedback-connected to the simulation effect impact recognition model to ensure that the adjuster can receive and process feedback information from the simulation effect impact recognition model and feed back the adjustment result to the simulation effect impact recognition model.

[0054] Optionally, the granularity adjuster receives the identified dynamic particle information from the simulation effect impact recognition model, and performs fine-grained processing on the identified dynamic particles, that is, adjusting the behavior and distribution range of the particles. Exemplarily, the granularity adjuster performs fine-grained processing on the identified dynamic particles, including reducing the time step simulation parameter of the identified dynamic particles or increasing the particle concentration around the identified dynamic particles, thereby optimizing the first simulation scene.

[0055] Through fine-grained processing, the particle size level and distribution range can be adjusted as needed to enhance the flexibility and adaptability of the simulation system, making the simulation closer to the actual situation and improving the reliability and effectiveness of the simulation system.

[0056] In some embodiments, establishing the granularity adjuster further includes: Obtain computing resource configuration parameters of the three-dimensional simulation system; perform particle support analysis using the computing resource configuration parameters and output an upper limit of supported particles; generate granularity constraints based on the upper limit of supported particles, and constrain the granularity adjuster based on the granularity constraints.

[0057] For example, to establish a granularity adjuster, first, obtain the computing resource configuration parameters of the three-dimensional simulation system, including collecting the configuration information of the CPU, GPU, memory, storage, etc. of the three-dimensional simulation system, and then analyze the processing capacity of the three-dimensional simulation system to determine the maximum number and complexity of particles that the system can support. Next, according to the upper limit of supported particles, set different granularity levels and corresponding adjustment ranges, generate granularity constraints, and ensure that the granularity adjuster does not exceed the computing power of the system during the optimization process. Finally, configure the parameters of the granularity adjuster with the granularity constraints so that it can dynamically adjust the granularity during the optimization process, while ensuring that the granularity adjuster does not exceed the computing power of the system during operation.

[0058] Through the above steps and technical means, by acquiring and analyzing the computing resource configuration parameters, establishing and configuring the granularity adjuster, it is helpful to efficiently utilize the resources of the three-dimensional simulation system in the optimization process and improve the accuracy and reliability of the simulation results.

[0059] Furthermore, the method further includes: decomposing the bioaerosol training scene to obtain N training scenes, and performing migration modeling management on the N training scenes based on the three-dimensional migration management system, including: Acquire multiple training scenarios that have completed three-dimensional simulation; connect multiple simulation scenarios corresponding to the multiple training scenarios, and share multiple simulation data sets corresponding to the multiple simulation scenarios; acquire the training scenarios to be migrated, perform data migration modeling based on the multiple simulation data sets, and construct simulation scenarios corresponding to the training scenarios to be migrated.

[0060] Optionally, first, the bioaerosol training scenario is decomposed into N training scenarios, and data is collected for the bioaerosol training scenarios that have completed three-dimensional simulation. Specifically, each training scenario corresponds to a specific simulation scenario and simulation data set; then, the multiple training scenarios collected above are connected to form a shared simulation data set network, and multiple simulation scenarios can access each other's data sets to achieve data sharing.

[0061] Further, a migration training scenario is obtained, which is a target for migration modeling. The training scenario to be migrated will obtain data from the completed training scenario and use multiple shared simulation data sets to perform migration modeling.

[0062] Specifically, suitable data is selected from the shared simulation data set for migration, including data sets with similar scene features or scene requirements. Then, the migrated data is used to build a new simulation scene, and the migrated model is optimized and adjusted to ensure that it is suitable for the new training scene.

[0063] Through the above method, existing simulation data and models can be reused efficiently, reducing the time and cost of re-modeling. At the same time, new training scenarios can be modeled based on more data, improving the accuracy and reliability of simulation results.

[0064] Furthermore, the method also includes: obtaining non-identification dynamic particles, the non-identification dynamic particles are dynamic particles whose influence index is less than or equal to the first preset influence index; obtaining dynamic particles whose influence index is less than or equal to a second preset influence index among the non-identification dynamic particles, wherein the first preset influence index is greater than the second preset influence index; and performing coarse-grained processing on the dynamic particles whose influence index is less than or equal to the second preset influence index according to the granularity adjuster to optimize the first simulation scene.

[0065] Optionally, dynamic particles whose impact index is less than or equal to the first preset impact index are identified, i.e., non-identified dynamic particles. Then, among the non-identified dynamic particles, dynamic particles whose impact index is less than or equal to the second preset impact index are further identified. Specifically, it can be considered that dynamic particles whose impact index is less than or equal to the second preset impact index have little impact on the three-dimensional simulation process. For example, dynamic particles whose impact index is less than or equal to the second preset impact index may be dynamic particles in a dead zone with poor air circulation in the first training scene or in a space far away from the bioaerosol.

[0066] Optionally, based on the above-mentioned granularity adjuster, coarse-grained processing is performed on dynamic particles that are less than or equal to the second preset influence index, including correspondingly reducing particle concentration, increasing corresponding time step simulation parameters, etc., thereby reducing inefficient computing power consumption on dynamic particles that are less than or equal to the second preset influence index and optimizing the first simulation scene.

[0067] In summary, the three-dimensional construction method for bioaerosol simulation scene provided by the present invention has the following technical effects: By identifying the first training scene for bioaerosols; configuring the three-dimensional simulation parameters of the first training scene, the three-dimensional simulation parameters include spatial scale simulation parameters, time step simulation parameters and particle number simulation parameters; using the three-dimensional simulation system to perform three-dimensional simulation on the first training scene with the three-dimensional simulation parameters, and outputting the first simulation scene; performing dynamic particle tracking on the first simulation scene, identifying dynamic particle information, including the object to which the dynamic particles belong, the distribution of the number of dynamic particles, and the step length of the dynamic particle change; establishing a simulation effect impact recognition model, inputting the object to which the dynamic particles belong, the distribution of the number of dynamic particles, and the step length of the dynamic particle change into the simulation effect impact recognition model for analysis, and obtaining the identified dynamic particles, wherein the identified dynamic particles are dynamic particles whose impact index is greater than the first preset impact index; establishing a granularity adjuster, the granularity adjuster is feedback-connected with the simulation effect impact recognition model, and fine-grained processing is performed on the identified dynamic particles according to the granularity adjuster to optimize the first simulation scene. Thus, the technical effect of improving simulation efficiency and simulation accuracy is achieved. Example 2

[0068] Figure 2 Schematic diagram of the structure of the three-dimensional construction system for bioaerosol simulation scene of the present invention. For example, Figure 1 The schematic flow chart of the three-dimensional construction method for the bioaerosol simulation scene of the present invention can be shown as follows: Figure 2 The structure shown is implemented.

[0069] Based on the same concept as the three-dimensional construction method for bioaerosol simulation scene in the above embodiment, the present invention also provides a three-dimensional construction system for bioaerosol simulation scene, including: The scene recognition module 11 is used to recognize a first training scene for bioaerosol.

[0070] The simulation parameter configuration module 12 is used to configure the three-dimensional simulation parameters of the first training scenario, and the three-dimensional simulation parameters include space scale simulation parameters, time step simulation parameters and particle number simulation parameters.

[0071] The scene simulation module 13 is used to use a three-dimensional simulation system to perform three-dimensional simulation on the first training scene with the three-dimensional simulation parameters, and output a first simulation scene.

[0072] The dynamic tracking module 14 is used to perform dynamic particle tracking on the first simulation scene and identify dynamic particle information, including the object to which the dynamic particles belong, the number distribution of dynamic particles, and the step length of dynamic particle changes.

[0073] The impact identification module 15 is used to establish a simulation effect impact identification model, input the object to which the dynamic particle belongs, the number distribution of the dynamic particles, and the change step of the dynamic particles into the simulation effect impact identification model for analysis, and obtain an identified dynamic particle, wherein the identified dynamic particle is a dynamic particle whose impact index is greater than a first preset impact index.

[0074] The scene optimization module 16 is used to establish a granularity adjuster, which is feedback-connected with the simulation effect impact identification model, and performs fine-grained processing on the identified dynamic particles according to the granularity adjuster to optimize the first simulation scene.

[0075] The dynamic tracking module 14 includes: A region segmentation unit is used to segment the first simulation scene into regions according to scene objects to obtain a plurality of sub-regions, wherein each sub-region corresponds to a scene object.

[0076] The monitoring block construction unit is used to construct multiple particle monitoring blocks for the multiple sub-areas, and each particle monitoring block is used to monitor the dynamic change data of particles in the corresponding sub-area.

[0077] The particle monitoring unit is used to generate a plurality of particle monitoring data sets according to the plurality of particle monitoring blocks.

[0078] The particle dynamic analysis unit is used to perform particle dynamic analysis on the multiple particle monitoring data sets to identify dynamic particle information.

[0079] Furthermore, the impact identification module 15 includes: The sample data unit is used to obtain a training sample data set, wherein the training sample data set includes a sample of an object to which a dynamic particle belongs, a sample of a distribution of the number of dynamic particles, a sample of a change step of a dynamic particle, and identification information for identifying an influencing index.

[0080] The association network configuration unit is used to establish an association influence network and initialize the association weight parameters of the association influence network.

[0081] The simulation effect influence identification model training unit is used to train the object samples to which the dynamic particles belong, the dynamic particle quantity distribution samples, the dynamic particle change step samples and the identification information of the identification influence index with the associated influence network until the accuracy reaches the preset accuracy and converges to output the simulation effect influence identification model.

[0082] Furthermore, the impact identification module 15 also includes: The sample division unit is used to divide the object samples to which the dynamic particles belong, and obtain first-category object samples and second-category object samples, wherein the first-category objects represent objects that react to changes in bioaerosols, and the second-category objects represent objects that react to changes in non-bioaerosols.

[0083] A hyperparameter configuration unit is used to configure a first penalty hyperparameter and a second penalty hyperparameter according to the first category of object samples and the second category of object samples, wherein the first penalty hyperparameter is greater than the second penalty hyperparameter.

[0084] A model optimization unit is used to construct a penalty feedback network using the first penalty hyperparameter and the second penalty hyperparameter, and to optimize the simulation effect impact identification model using the penalty feedback network.

[0085] In some implementations, the sample division unit in the impact identification module 15 includes: The secondary classification unit is used to perform secondary classification on the first-category object samples to obtain first-category direct objects and first-category indirect objects, wherein the first-category direct objects are objects that characterize changes in direct-reaction bioaerosols, and the first-category indirect objects are objects that characterize changes in indirect-reaction bioaerosols.

[0086] The hyperparameter group configuration unit is used to configure a first direct penalty hyperparameter and a first indirect penalty hyperparameter corresponding to the first type of direct object and the first type of indirect object, respectively, wherein the first direct penalty hyperparameter is greater than the first indirect penalty hyperparameter and greater than the second penalty hyperparameter.

[0087] Furthermore, the system also includes a migration modeling management unit, which is used to: Decompose the bioaerosol training scenario to obtain N training scenarios, and perform migration modeling management on the N training scenarios based on the three-dimensional migration management system, including: obtaining multiple training scenarios that have completed three-dimensional simulation. Connect multiple simulation scenarios corresponding to the multiple training scenarios, and share multiple simulation data sets corresponding to the multiple simulation scenarios. Obtain the training scenario to be migrated, perform data migration modeling based on the multiple simulation data sets, and construct the simulation scenario corresponding to the training scenario to be migrated.

[0088] Furthermore, the scene optimization module 16 includes: The computing resource acquisition unit is used to acquire computing resource configuration parameters of the three-dimensional simulation system.

[0089] The particle support analysis unit is used to perform particle support analysis with the computing resource configuration parameters and output an upper limit of supported particles.

[0090] The granularity constraint unit is used to generate a granularity constraint condition according to the upper limit of supported particles, and constrain the granularity adjuster according to the granularity constraint condition.

[0091] It should be understood that the embodiments mentioned in this specification focus on their differences from other embodiments. The specific embodiments in the aforementioned embodiment one are also applicable to the three-dimensional construction system for bioaerosol simulation scenes described in embodiment two. For the sake of brevity of the specification, they will not be further expanded here.

[0092] It should be understood that the embodiments disclosed in the present invention and the above description can enable those skilled in the art to use the present invention to implement the present invention. At the same time, the present invention is not limited to the above-mentioned embodiments. It should be understood that those skilled in the art can still modify the technical solutions recorded in the above-mentioned embodiments, or replace some of the technical features therein by equivalents; 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 invention, and should be included in the protection scope of the present invention.

Claims

1. A three-dimensional construction method for bioaerosol simulation scenes, characterized in that: The method comprises: Identify the first training scenario for bioaerosols; Configuring three-dimensional simulation parameters of the first training scenario, wherein the three-dimensional simulation parameters include a spatial scale simulation parameter, a time step simulation parameter, and a particle number simulation parameter; Using a three-dimensional simulation system to perform three-dimensional simulation on the first training scene with the three-dimensional simulation parameters, and outputting a first simulation scene; Performing dynamic particle tracking on the first simulation scene to identify dynamic particle information, including the object to which the dynamic particles belong, the number distribution of dynamic particles, and the step length of dynamic particle changes; Establishing a simulation effect impact identification model, inputting the object to which the dynamic particle belongs, the number distribution of the dynamic particles, and the change step of the dynamic particles into the simulation effect impact identification model for analysis, and obtaining an identified dynamic particle, wherein the identified dynamic particle is a dynamic particle whose impact index is greater than a first preset impact index; A granularity adjuster is established, the granularity adjuster is feedback-connected with the simulation effect impact identification model, and the identified dynamic particles are fine-grainedly processed according to the granularity adjuster to optimize the first simulation scene.

2. The three-dimensional construction method for bioaerosol simulation scene according to claim 1, characterized in that: Performing dynamic particle tracking on the first simulation scene to identify dynamic particle information includes: Performing region segmentation on the first simulation scene according to the scene objects to obtain a plurality of sub-regions, wherein each sub-region corresponds to a scene object; For the multiple sub-areas, construct multiple particle monitoring blocks, each particle monitoring block is used to monitor dynamic change data of particles in the corresponding sub-area; generating a plurality of particle monitoring data sets according to the plurality of particle monitoring blocks; Perform particle dynamic analysis on the multiple particle monitoring data sets to identify dynamic particle information.

3. The three-dimensional construction method for bioaerosol simulation scene according to claim 1, characterized in that: Establish a simulation effect impact identification model, the method includes: Acquire a training sample data set, wherein the training sample data set includes a sample of an object to which dynamic particles belong, a sample of a distribution of the number of dynamic particles, a sample of a change step of dynamic particles, and identification information of an identification influencing index; Establishing an association influence network and initializing association weight parameters of the association influence network; The associated influence network is used to train the object samples to which the dynamic particles belong, the dynamic particle quantity distribution samples, the dynamic particle change step samples, and the identification information of the identification influence index until the accuracy reaches the preset accuracy, and the simulation effect influence identification model is converged and output.

4. The three-dimensional construction method for bioaerosol simulation scene according to claim 3, characterized in that: Establishing a simulation effect impact identification model, the method also includes: Dividing the object samples to which the dynamic particles belong to obtain first-category object samples and second-category object samples, wherein the first-category objects represent objects that react to changes in bioaerosols, and the second-category objects represent objects that react to changes in non-bioaerosols; According to the first category of object samples and the second category of object samples, configuring a first penalty hyperparameter and a second penalty hyperparameter, wherein the first penalty hyperparameter is greater than the second penalty hyperparameter; A penalty feedback network is constructed using the first penalty hyperparameter and the second penalty hyperparameter, and the penalty feedback network is used to perform model optimization on the simulation effect impact identification model.

5. The three-dimensional construction method for bioaerosol simulation scene according to claim 4, characterized in that: After obtaining the first type of object samples, it also includes: Performing secondary classification on the first-category object samples to obtain first-category direct objects and first-category indirect objects; Wherein, the first type of direct objects are objects that characterize changes in direct-reaction bioaerosols, and the first type of indirect objects are objects that characterize changes in indirect-reaction bioaerosols; Configure a first direct penalty hyperparameter and a first indirect penalty hyperparameter corresponding to the first type of direct objects and the first type of indirect objects respectively; Among them, the first direct penalty hyperparameter is greater than the first indirect penalty hyperparameter which is greater than the second penalty hyperparameter.

6. The three-dimensional construction method for bioaerosol simulation scene according to claim 1, characterized in that: Decompose the bioaerosol training scenario to obtain N training scenarios, and perform migration modeling management on the N training scenarios based on the three-dimensional migration management system, including: Acquire multiple training scenarios that have completed three-dimensional simulation; Connecting multiple simulation scenarios corresponding to the multiple training scenarios, and sharing multiple simulation data sets corresponding to the multiple simulation scenarios; Acquire the training scenario to be migrated, perform data migration modeling based on the multiple simulation data sets, and construct a simulation scenario corresponding to the training scenario to be migrated.

7. The three-dimensional construction method for bioaerosol simulation scene according to claim 1, characterized in that: Establishing the granularity adjuster, the method further comprises: Obtaining computing resource configuration parameters of the three-dimensional simulation system; Perform particle support analysis using the computing resource configuration parameters and output an upper limit of supported particles; A granularity constraint condition is generated according to the upper limit of supported particles, and the granularity adjuster is constrained according to the granularity constraint condition.

8. The three-dimensional construction method for bioaerosol simulation scene according to claim 1, characterized in that: Acquire a non-identification dynamic particle, where the non-identification dynamic particle is a dynamic particle whose influence index is less than or equal to the first preset influence index; Acquire dynamic particles whose influence index is less than or equal to a second preset influence index among the non-identification dynamic particles, wherein the first preset influence index is greater than the second preset influence index; The first simulation scene is optimized by performing coarse-grained processing on the dynamic particles that are less than or equal to the second preset impact index according to the granularity adjuster.

9. A three-dimensional construction system for bioaerosol simulation scenes, characterized in that: The system is used to execute the three-dimensional construction method for bioaerosol simulation scene according to any one of claims 1 to 8, and the system comprises: A scene recognition module, the scene recognition module is used to identify a first training scene for bioaerosol; A simulation parameter configuration module, wherein the simulation parameter configuration module is used to configure three-dimensional simulation parameters of the first training scenario, wherein the three-dimensional simulation parameters include a spatial scale simulation parameter, a time step simulation parameter, and a particle number simulation parameter; A scene simulation module, the scene simulation module is used to use a three-dimensional simulation system to perform three-dimensional simulation on the first training scene with the three-dimensional simulation parameters, and output a first simulation scene; A dynamic tracking module, the dynamic tracking module is used to perform dynamic particle tracking on the first simulation scene, and identify dynamic particle information, including the object to which the dynamic particles belong, the number distribution of dynamic particles, and the step length of dynamic particle changes; An impact identification module, the impact identification module is used to establish a simulation effect impact identification model, input the object to which the dynamic particle belongs, the number distribution of the dynamic particles, and the change step of the dynamic particles into the simulation effect impact identification model for analysis, and obtain an identified dynamic particle, wherein the identified dynamic particle is a dynamic particle whose impact index is greater than a first preset impact index; A scene optimization module is used to establish a granularity adjuster, the granularity adjuster is feedback-connected with the simulation effect impact identification model, and the identified dynamic particles are fine-grainedly processed according to the granularity adjuster to optimize the first simulation scene.

Images