Energy controllable turbulence simulation method and device based on continuous convolution

Abstract

The invention provides an energy-controllable turbulence simulation method and device based on continuous convolution, and relates to the technical field of computer graphics fluid simulation. The method comprises the following steps: generating particle method physical turbulence data through various physical solvers, sampling and calculating particle speed to obtain various model data sets; training the continuous convolution network model to obtain a plurality of continuous convolution approximation model units, and performing particle displacement prediction according to the model units; and inputting the particle displacement prediction result into an energy interpolation network to obtain an interpolation proportionality coefficient, and carrying out hybrid interpolation on the particle displacement prediction results of the plurality of continuous convolution approximation model units according to the interpolation proportionality coefficient to obtain a final simulation result. The turbulence energy change level generated by the method keeps positive correlation with the control parameters, the effectiveness of the energy intensity coefficient on turbulence energy control is proved, and the obvious trend of transition from laminar flow to turbulence is presented in the visual effect.

Description

Technical Field

[0001] This invention relates to the field of computer graphics fluid simulation technology, and in particular to an energy-controlled turbulence simulation method and apparatus based on continuous convolution. Background Technology

[0002] Turbulence is a collective term for various complex and subtle fluid phenomena such as vortices, smoke, and waves. Computer simulation of turbulence is a hot topic in computer graphics. Compared to laminar flow, turbulence offers much richer detail. The challenge in turbulence simulation lies in balancing the level of detail in vortices with the time and space costs of simulation, while simultaneously preserving the natural generation and decay of vortices.

[0003] Currently, various particle-based turbulence simulation methods exist to simulate various turbulent phenomena with vortices. Conventional smoothed particle hydrodynamics methods, when simulating turbulence, suffer from vortex loss due to pressure projection, leading to the loss of turbulent details. Eddy particle methods, by introducing a vortex equation and calculating and compensating for vortex loss in each frame, can comprehensively track the vortex information of each particle. However, the Biot-Savart summation for converting vortex compensation to a velocity field incurs expensive computational time costs. Furthermore, as the positive feedback of the correction accumulates, the water surface is prone to prolonged oscillations, resulting in a lack of surface calm. In addition, due to strict physical constraints, the velocity displacement generated in each frame by the physical solver is fixed, and the energy lacks a controllable range, limiting the scalability of the simulation. For previous turbulence simulators, excessive velocity compensation can lead to energy oscillations and non-convergence, resulting in a fluid surface that remains unstable for extended periods. Such solvers typically rely on adjusting initial parameters and repeating simulations to indirectly affect energy; there is usually no direct correspondence between simulation parameters and energy, making direct energy control impossible.

[0004] Continuous convolution, as a method for extracting features from particle data, can be used for feature learning and particle displacement prediction in turbulent data. By using continuous convolution kernels to extract information about the particle's neighborhood and penalizing deviations in particle motion displacement, fluid motion characteristics are learned. Compared to physical solvers, the resulting corrections do not need to strictly adhere to the Navier-Stokes equations, exhibiting looser constraints. This provides the possibility of dynamically adjusting turbulent energy at each time step. Summary of the Invention

[0005] To address the technical problems of existing physical solvers in simulating turbulence, such as insufficient turbulence details due to vorticity loss, difficulty in directly controlling turbulence energy, or prolonged water surface instability due to excessive energy compensation, this invention provides an energy-controllable turbulence simulation method and apparatus based on continuous convolution. The technical solution is as follows:

[0006] On the one hand, an energy-controlled turbulence simulation method based on continuous convolution is provided. This method is implemented by an energy-controlled turbulence simulation device based on continuous convolution, and includes: S1. For the turbulent scenario to be simulated, particle method physical turbulence data corresponding to each physical solver is generated through multiple physical solvers. The particle velocity is then sampled and calculated for each type of particle method physical turbulence data to obtain multiple model datasets.

[0007] S2. Train a continuous convolutional network model based on each model dataset in the multiple model datasets to obtain multiple continuous convolutional approximation model units. Perform particle displacement prediction based on each continuous convolutional approximation model unit to obtain the particle displacement prediction results of multiple continuous convolutional approximation model units.

[0008] S3. Construct an energy interpolation network and input the particle displacement prediction results of multiple continuous convolution approximation model units into the energy interpolation network to obtain the interpolation ratio coefficients.

[0009] S4. Based on the interpolation scaling factor, perform mixed interpolation on the particle displacement prediction results of multiple continuous convolution approximation model units to obtain the final simulation result.

[0010] Optionally, in S1, the physical turbulence data for each particle method are sampled and the particle velocities are calculated to obtain multiple model datasets, including: Each type of particle physics turbulence data is sampled at intervals. The particle velocity is calculated based on the particle position in each frame of the sampling results. The model dataset corresponding to each type of particle physics turbulence data is constructed based on the particle position and velocity in each frame, thus forming multiple model datasets.

[0011] Optionally, each continuous convolutional approximation model unit in S2 includes: a continuous convolutional kernel for extracting features of particles and performing particle displacement prediction.

[0012] Optionally, in S3, the particle displacement prediction results of multiple consecutive convolutional approximation model units are input into the energy interpolation network to obtain interpolation scaling coefficients, including: The target intensity of turbulent energy is obtained based on the particle displacement prediction results of multiple continuous convolution approximation model units. The particle displacement prediction results of multiple continuous convolution approximation model units and the target intensity of turbulent energy are input into the energy interpolation network to obtain the interpolation ratio coefficient.

[0013] Optionally, the target intensity of turbulence energy is obtained based on the particle displacement prediction results of multiple consecutive convolutional approximation model units, including: The energy of turbulence is simplified to the kinetic energy of particles. The kinetic energy of particles is calculated from the predicted velocity of particles, and the predicted velocity of particles is calculated from the predicted displacement of particles. Then, the change function of the predicted particle displacement of multiple consecutive convolutional approximation model units is regarded as the change of turbulent energy. Based on the changes in turbulent energy, the upper and lower bounds of energy intensity of multiple continuous convolutional approximation model units are obtained. An energy error is constructed, and an energy interpolation network is used to penalize the energy error based on the upper and lower bounds of energy intensity, thereby approximating the target intensity of turbulent energy.

[0014] Optionally, in S4, the particle displacement prediction results of multiple consecutive convolutional approximation model units are mixed and interpolated based on the interpolation scaling factor to obtain the final simulation results, including: S41. The predicted velocity of the particle is calculated based on the particle displacement prediction results of multiple consecutive convolution approximation model units.

[0015] S42. Calculate the intermediate velocity generated by the advection process in the turbulent scene based on the predicted velocity of the particles.

[0016] S43. The intermediate displacement generated by the advection process is calculated based on the particle displacement prediction results, the predicted velocity of the particles, and the intermediate velocity.

[0017] S44. Calculate the mixed particle displacement prediction based on the interpolation scaling factor and the particle displacement prediction result, and calculate the particle displacement prediction result for the next frame based on the mixed particle displacement prediction and the intermediate displacement.

[0018] S45. Repeat steps S41-S44 to produce the final simulation results.

[0019] On the other hand, an energy-controlled turbulence simulation apparatus based on continuous convolution is provided. This apparatus is applied to an energy-controlled turbulence simulation method based on continuous convolution. The apparatus includes: The dataset construction module is used to generate particle-based physical turbulence data for each physical solver for the turbulence scenario to be simulated. The particle velocities are then calculated and sampled for each type of particle-based physical turbulence data to obtain multiple model datasets.

[0020] The displacement prediction module is used to train a continuous convolutional network model based on each model dataset in multiple model datasets, obtain multiple continuous convolutional approximation model units, and perform particle displacement prediction based on each continuous convolutional approximation model unit to obtain the particle displacement prediction results of multiple continuous convolutional approximation model units.

[0021] The interpolation scaling factor calculation module is used to construct an energy interpolation network. It inputs the particle displacement prediction results of multiple continuous convolution approximation model units into the energy interpolation network to obtain the interpolation scaling factor.

[0022] The output module is used to perform mixed interpolation on the particle displacement prediction results of multiple consecutive convolution approximation model units according to the interpolation scaling factor to obtain the final simulation result.

[0023] Optionally, the dataset building module is further used for: Each type of particle physics turbulence data is sampled at intervals. The particle velocity is calculated based on the particle position in each frame of the sampling results. The model dataset corresponding to each type of particle physics turbulence data is constructed based on the particle position and velocity in each frame, thus forming multiple model datasets.

[0024] Optionally, each continuous convolutional approximation model unit includes: a continuous convolutional kernel for extracting features of particles and performing particle displacement prediction.

[0025] Optionally, the interpolation scaling factor calculation module is further used for: The target intensity of turbulent energy is obtained based on the particle displacement prediction results of multiple continuous convolution approximation model units. The particle displacement prediction results of multiple continuous convolution approximation model units and the target intensity of turbulent energy are input into the energy interpolation network to obtain the interpolation ratio coefficient.

[0026] Optionally, the interpolation scaling factor calculation module is further used for: The energy of turbulence is simplified to the kinetic energy of particles. The kinetic energy of particles is calculated from the predicted velocity of particles, and the predicted velocity of particles is calculated from the predicted displacement of particles. Then, the change function of the predicted particle displacement of multiple consecutive convolutional approximation model units is regarded as the change of turbulent energy. Based on the changes in turbulent energy, the upper and lower bounds of energy intensity of multiple continuous convolutional approximation model units are obtained. An energy error is constructed, and an energy interpolation network is used to penalize the energy error based on the upper and lower bounds of energy intensity, thereby approximating the target intensity of turbulent energy.

[0027] Optionally, the output module is further used for: S41. The predicted velocity of the particle is calculated based on the particle displacement prediction results of multiple consecutive convolution approximation model units.

[0028] S42. Calculate the intermediate velocity generated by the advection process in the turbulent scene based on the predicted velocity of the particles.

[0029] S43. The intermediate displacement generated by the advection process is calculated based on the particle displacement prediction results, the predicted velocity of the particles, and the intermediate velocity.

[0030] S44. Calculate the mixed particle displacement prediction based on the interpolation scaling factor and the particle displacement prediction result, and calculate the particle displacement prediction result for the next frame based on the mixed particle displacement prediction and the intermediate displacement.

[0031] S45. Repeat steps S41-S44 to produce the final simulation results.

[0032] On the other hand, an energy-controlled turbulence simulation device based on continuous convolution is provided. The energy-controlled turbulence simulation device based on continuous convolution includes: a processor; a memory, wherein computer-readable instructions are stored in the memory, and when the computer-readable instructions are executed by the processor, any one of the above-described energy-controlled turbulence simulation methods based on continuous convolution is implemented.

[0033] On the other hand, a computer-readable storage medium is provided, wherein at least one instruction is stored in the storage medium, the at least one instruction being loaded and executed by a processor to implement any of the above-described energy-controlled turbulence simulation methods based on continuous convolution.

[0034] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: In this invention, an energy-adjustable turbulence simulation method based on continuous convolution feature extraction is designed. By providing adjustable energy parameters, the energy intensity of the turbulence simulation is quantitatively controlled, which makes up for the shortcomings of traditional turbulence solvers in energy control, while still maintaining visual rationality and improving the accuracy and controllability of the simulation results.

[0035] By blending multiple different continuous convolutional approximation models, the framework achieves the fusion of visual features from different methods. It can be freely extended to more datasets, training the corresponding continuous convolutional approximation model units and then uniformly fusing them to provide richer results.

[0036] In terms of efficiency, the average inference time of each model unit is close to that of a physical turbulence solver. As long as the dataset contains particle velocity and displacement information, this invention has no theoretical restrictions on the construction method and source of the particle dataset, and can be extended to any other particle-based turbulence simulation method. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 This is a flowchart of an energy-controllable turbulence simulation method based on continuous convolution provided by an embodiment of the present invention; Figure 2 This is a scene diagram used in the verification experiment provided in the embodiments of the present invention; Figure 3 This is a graph showing the variation of turbulent kinetic energy with frame number provided in an embodiment of the present invention; Figure 4 This is an experimental scenario diagram for testing long-term simulation stability provided in an embodiment of the present invention; Figure 5 This is a block diagram of an energy-controllable turbulence simulation device based on continuous convolution provided in an embodiment of the present invention; Figure 6 This is a schematic diagram of the structure of an energy-controllable turbulence simulation device based on continuous convolution provided in an embodiment of the present invention. Detailed Implementation

[0039] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0040] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0041] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0042] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0043] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0044] This invention provides an energy-controlled turbulence simulation method based on continuous convolution. This method can be implemented using an energy-controlled turbulence simulation device based on continuous convolution, which can be a terminal or a server. Figure 1 The flowchart shown is for an energy-controlled turbulence simulation method based on continuous convolution. The processing flow of this method may include the following steps:

[0045] S1. For the turbulent scenario to be simulated, particle method physical turbulence data corresponding to each physical solver is generated through multiple physical solvers. The particle velocity is then sampled and calculated for each type of particle method physical turbulence data to obtain multiple model datasets.

[0046] Optionally, in S1, the physical turbulence data for each particle method are sampled and the particle velocities are calculated to obtain multiple model datasets, including: Each type of particle physics turbulence data is sampled at intervals. The particle velocity is calculated based on the particle position in each frame of the sampling results. The model dataset corresponding to each type of particle physics turbulence data is constructed based on the particle position and velocity in each frame, thus forming multiple model datasets.

[0047] In one feasible implementation, a particle turbulence data frame is first generated using a physics solver, followed by interval sampling. The corresponding velocity is calculated based on the particle position information of each frame to obtain a model dataset. This process is performed on multiple physics solvers to construct datasets with the characteristics of the corresponding methods.

[0048] S2. Train a continuous convolutional network model based on each model dataset in the multiple model datasets to obtain multiple continuous convolutional approximation model units. Perform particle displacement prediction based on each continuous convolutional approximation model unit to obtain the particle displacement prediction results of multiple continuous convolutional approximation model units.

[0049] Optionally, each continuous convolutional approximation model unit in S2 includes: a continuous convolutional kernel for extracting features of particles and performing particle displacement prediction.

[0050] In one feasible implementation, a continuous convolutional network model is trained on the dataset to obtain a single continuous convolutional approximation model corresponding to each physical turbulence solver. An energy interpolation network is then trained to infer correction coefficients based on the target energy intensity.

[0051] Specifically, a continuous convolutional model is trained. First, data is input into this module: from the fluid dataset simulated by SPH (Smoothed Particle Hydrodynamics), the particle position and velocity dataset for each frame is obtained through sampling.

[0052] (1) In the formula, , They represent particles respectively In Position and velocity at any given moment The number of particles in the flow field.

[0053] Furthermore, a fundamental component of a continuous convolutional network is a continuous convolutional kernel. The model uses continuous convolutional kernels to extract fluid features and performs displacement prediction by applying the kernels to the particle features:

[0054] (2) In the formula, Indicates in Time of the first The network predicts the displacement of each particle. and Both indicate that the convolution kernel is in a relative position. The weight value at that point, Indicates the first The feature vector of each particle Indicates the first The feature vector of each particle can be encoded as , Indicates in Time of the first The network prediction speed for each particle For particles The neighborhood within a certain radius. This is the distance importance function, used to... Particles within the space are assigned weights based on their distance. This is a mapping from a spherical domain to a cubic domain. Indicates in Time of the first The network predicts the displacement of each particle.

[0055] In continuous convolutional networks, a loss penalty can be imposed on the network based on the dataset. Obtain the required model parameters. The penalty term is the particle's displacement in the next frame plus its displacement after one iteration:

[0056] (3) In the formula, Indicates the current moment. Indicates the time step between adjacent frames. This indicates the number of particles in the flow field.

[0057] S3. Construct an energy interpolation network and input the particle displacement prediction results of multiple continuous convolution approximation model units into the energy interpolation network to obtain the interpolation ratio coefficients.

[0058] Optionally, in S3, the particle displacement prediction results of multiple consecutive convolutional approximation model units are input into the energy interpolation network to obtain interpolation scaling coefficients, including: The target intensity of turbulent energy is obtained based on the particle displacement prediction results of multiple continuous convolution approximation model units. The particle displacement prediction results of multiple continuous convolution approximation model units and the target intensity of turbulent energy are input into the energy interpolation network to obtain the interpolation ratio coefficient.

[0059] Optionally, the target intensity of turbulence energy is obtained based on the particle displacement prediction results of multiple consecutive convolutional approximation model units, including: The energy of turbulence is simplified to the kinetic energy of particles. The kinetic energy of particles is calculated from the predicted velocity of particles, and the predicted velocity of particles is calculated from the predicted displacement of particles. Then, the change function of the predicted particle displacement of multiple consecutive convolutional approximation model units is regarded as the change of turbulent energy. Based on the changes in turbulent energy, the upper and lower bounds of energy intensity of multiple continuous convolutional approximation model units are obtained. An energy error is constructed, and an energy interpolation network is used to penalize the energy error based on the upper and lower bounds of energy intensity, thereby approximating the target intensity of turbulent energy.

[0060] In one feasible implementation, an energy interpolation network is trained to infer correction coefficients based on the target energy intensity.

[0061] Specifically, the energy interpolation network is constructed using a multi-layer linear sensing stacking method: (4) In the formula, Indicates the first Layer features, This represents a non-linear filtering function. Represents the weight matrix. This represents the bias vector. Indicates network depth. Input features. Linear correction and control energy intensities given for the model elements:

[0062] (5) In the formula, This represents the displacement prediction generated by each model element. Indicates the number of consecutive convolutional models. Indicates the target intensity of turbulent energy. This represents the model interpolation scaling factor. Turbulent energy. This can be simplified to the kinetic energy of the particles, assuming the particles have equal mass. In this case, particle kinetic energy for:

[0063] (6) In the formula, Indicates in Time of the first The predicted velocity of each particle. This represents the number of particles in the flow field. The change in flow field energy generated in each frame can be considered as the predicted displacement change. function :

[0064] (7) and, (8) in The time step between adjacent frames As mentioned earlier, displacement predictions are generated for each model unit. The upper and lower bounds of energy intensity that the model can predict for each frame are also included. , The extreme value of energy change that a single model unit can produce is used as the benchmark:

[0065] (9) (10) in This represents the sequence number of the continuous convolutional model. Energy error is used. Penalized interpolation networks approximate the required strength :

[0066] (11) in, As mentioned earlier, this represents the target intensity of turbulent energy. .when When , it indicates that the lowest level achievable by a single solver is selected (with only laminar components, close to a non-turbulent level). At that time, the energy reaches the highest level achievable by selecting a single solver.

[0067] S4. Based on the interpolation scaling factor, perform mixed interpolation on the particle displacement prediction results of multiple continuous convolution approximation model units to obtain the final simulation result.

[0068] Optionally, step S4 above includes: S41. The predicted velocity of the particle is calculated based on the particle displacement prediction results of multiple consecutive convolution approximation model units.

[0069] S42. Calculate the intermediate velocity generated by the advection process in the turbulent scene based on the predicted velocity of the particles.

[0070] S43. The intermediate displacement generated by the advection process is calculated based on the particle displacement prediction results, the predicted velocity of the particles, and the intermediate velocity.

[0071] S44. Calculate the mixed particle displacement prediction based on the interpolation scaling factor and the particle displacement prediction result, and calculate the particle displacement prediction result for the next frame based on the mixed particle displacement prediction and the intermediate displacement. Repeat the above process to produce the final simulation result.

[0072] In one feasible implementation, energy interpolation inference and iterative prediction are performed by integrating multiple trained continuous convolutional approximation model units into the network, using an interpolation network to predict the interpolation ratio, which is determined based on the energy correction that a single model can provide, and performing mixed interpolation in each frame to predict displacement.

[0073] Specifically, non-stress term prediction. Particle displacement is predicted using a hybrid interpolation ratio, and particle motion prediction is iteratively executed. The final prediction result for the scene and its corresponding energy range are obtained. This is then approximated to the desired energy range using a prediction network. During the iterative prediction process, the intermediate velocity generated by the advection process is first calculated:

[0074] (12) In the formula, Represents gravity. Indicates intermediate speed. Indicates the time step. and All of these represent the entire network (i.e., all modules described in this invention combined) at the [missing information] th [missing information]. The prediction speed of the frame output is calculated by equation (8) above, specifically based on... Calculations show that The calculation method can be found in the following formula (15).

[0075] Then calculate the displacement caused by advection: (13) in, Indicates intermediate displacement. This indicates that the entire network is in the [number]th [phase]. Predicted displacement of the frame.

[0076] Energy interpolation coefficient inference and pressure term prediction. After obtaining individual model units, these model units are combined and fed into an energy interpolation network for coefficient inference to approximate the required energy range. A hybrid correction term is then applied to obtain the final particle displacement prediction result for the next frame. :

[0077] (14) (15) in, This represents the number of continuous convolutional approximation models. Hybrid linear particle displacement correction given for the interpolation network. As mentioned earlier, this represents the displacement correction produced by a single continuous convolution model. middle This represents the interpolation ratio coefficient of the model. It is a subscript used for summation.

[0078] This invention achieves hybrid interpolation by summing the above equation (14), and calculates the displacement prediction result according to the above equation (15). For example, suppose... , ,but Therefore, the order of calculation is: first calculate... Then calculate Finally, the calculation was completed. (And so on, to calculate) →Calculate →Calculate →Calculate …).

[0079] Furthermore, the accuracy of the energy approximation strategy was verified through experiments. Figure 2 To verify the scene used in the experiment, color depth was used to represent the speed of local particles, with areas of higher grayscale values ​​(i.e., lighter colors) corresponding to areas of higher speed. Water blocks started from the left side of a rectangular container and moved to the right, impacting a hemispherical obstacle. They then bounced off the obstacle's rear boundary, creating wave and vortex details. The scene was initially set with a gradient-distributed relative energy intensity. ,from Figure 2 As can be seen, with the increase in energy control intensity, the resulting turbulence also exhibits more pronounced structural characteristics. Initially, laminar flow predominated, with a smooth fluid surface, but as... The improvement of the control coefficient leads to a transition of the fluid surface towards turbulence with more fine structures, a phenomenon that verifies the control coefficient. It can effectively enhance turbulence details. From Figure 3The curve showing the change of turbulent kinetic energy with frame number in the image shows that the required energy curve changes with the given coefficient. The upward movement due to the increase in turbulence energy verifies the relationship between the intensity of turbulence energy and its intensity coefficient. There is an approximately positive correlation between them.

[0080] Figure 4 To test the stability of a long-term simulated scenario, a baffle was inserted into the water and rotated around its center to create waves. After a period of time, it was removed, and the energy curves were compared with those obtained from several physics solvers. Figure 4 As can be seen, this invention effectively addresses the problems of excessively rapid energy convergence and energy instability in previous methods: During rotation, DFSPH (Divergence-Free Smoothed Particle Hydrodynamics) fails to generate vortices due to insufficient turbulent energy, while MCVSPH (Monte-Carlo Vortex Smoothed Particle Hydrodynamics), although visually producing many small vortices, exhibits a fairly uniform distribution and is not naturally generated by the rotation of the baffle. In contrast, this invention produces more natural, circular, outward-spreading vortices. After removing the baffle, the MCVSPH method still exhibits a water surface that remains unstable for a long time, while the method of this invention gradually becomes stable. The corresponding energy curves show that during long-term simulations, the energy of MCVSPH is not stable but exhibits alternating oscillations. DFSPH, on the other hand, shows insufficient energy. In contrast, the method of this invention maintains high kinetic energy while achieving stable final convergence.

[0081] This embodiment proposes an energy-controlled turbulence simulation method based on continuous convolution. First, multiple physical turbulence simulation datasets are constructed, and corresponding continuous convolution displacement prediction models are trained. At each time step, the displacement prediction of each model is calculated, and the prediction weights are mixed using an energy interpolation network to make the turbulence energy approximate the specific fluid energy level provided by the user. The resulting turbulence energy change level maintains a positive correlation with the control parameters, demonstrating the effectiveness of the energy intensity coefficient for turbulence energy control, and visually exhibiting a clear trend of transition from laminar to turbulent flow.

[0082] In this embodiment of the invention, an energy-adjustable turbulence simulation method based on continuous convolution feature extraction is designed. By providing adjustable energy parameters, the energy intensity of the turbulence simulation is quantitatively controlled, which makes up for the shortcomings of traditional turbulence solvers in energy control, while still maintaining visual rationality and improving the accuracy and controllability of the simulation results.

[0083] By blending multiple different continuous convolutional approximation models, the framework achieves the fusion of visual features from different methods. It can be freely extended to more datasets, training the corresponding continuous convolutional approximation model units and then uniformly fusing them to provide richer results.

[0084] In terms of efficiency, the average inference time of each model unit is close to that of a physical turbulence solver. As long as the dataset contains particle velocity and displacement information, this invention has no theoretical restrictions on the construction method and source of the particle dataset, and can be extended to any other particle-based turbulence simulation method.

[0085] Figure 5 This is a block diagram of an energy-controlled turbulence simulation apparatus based on continuous convolution, according to an exemplary embodiment. The apparatus is used in an energy-controlled turbulence simulation method based on continuous convolution. (Refer to...) Figure 5 The device includes a dataset construction module 310, a displacement prediction module 320, an interpolation scaling factor calculation module 330, and an output module 340. Wherein:

[0086] The dataset construction module 310 is used to generate particle method physical turbulence data for each physical solver for the turbulence scenario to be simulated, and to sample and calculate the particle velocity for each type of particle method physical turbulence data to obtain multiple model datasets.

[0087] The displacement prediction module 320 is used to train a continuous convolutional network model based on each model dataset in multiple model datasets to obtain multiple continuous convolutional approximation model units, and to perform particle displacement prediction based on each continuous convolutional approximation model unit to obtain the particle displacement prediction results of multiple continuous convolutional approximation model units.

[0088] The interpolation scaling factor calculation module 330 is used to construct an energy interpolation network. It inputs the particle displacement prediction results of multiple continuous convolution approximation model units into the energy interpolation network to obtain the interpolation scaling factor.

[0089] Output module 340 is used to perform mixed interpolation on the particle displacement prediction results of multiple consecutive convolution approximation model units according to the interpolation scaling factor to obtain the final simulation result.

[0090] In this embodiment of the invention, an energy-adjustable turbulence simulation method based on continuous convolution feature extraction is designed. By providing adjustable energy parameters, the energy intensity of the turbulence simulation is quantitatively controlled, which makes up for the shortcomings of traditional turbulence solvers in energy control, while still maintaining visual rationality and improving the accuracy and controllability of the simulation results.

[0091] By blending multiple different continuous convolutional approximation models, the framework achieves the fusion of visual features from different methods. It can be freely extended to more datasets, training the corresponding continuous convolutional approximation model units and then uniformly fusing them to provide richer results.

[0092] In terms of efficiency, the average inference time of each model unit is close to that of a physical turbulence solver. As long as the dataset contains particle velocity and displacement information, this invention has no theoretical restrictions on the construction method and source of the particle dataset, and can be extended to any other particle-based turbulence simulation method.

[0093] Figure 6 This is a schematic diagram of the structure of an energy-controllable turbulence simulation device based on continuous convolution provided in an embodiment of the present invention, as shown below. Figure 6 As shown, an energy-controllable turbulence simulation device based on continuous convolution can include the above-mentioned Figure 5 The illustrated energy-controlled turbulence simulation device is based on continuous convolution. Optionally, the energy-controlled turbulence simulation device 410 based on continuous convolution may include a first processor 2001.

[0094] Optionally, the energy-controlled turbulence simulation device 410 based on continuous convolution may also include a memory 2002 and a transceiver 2003.

[0095] The first processor 2001, memory 2002, and transceiver 2003 can be connected via a communication bus.

[0096] The following is combined Figure 6 A detailed description of each component of the energy-controlled turbulence simulation device 410 based on continuous convolution is provided below: The first processor 2001 is the control center of the energy-controllable turbulence simulation device 410 based on continuous convolution. It can be a single processor or a collective term for multiple processing elements. For example, the first processor 2001 can be one or more central processing units (CPUs), application-specific integrated circuits (ASICs), or one or more integrated circuits configured to implement embodiments of the present invention, such as one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs).

[0097] Optionally, the first processor 2001 can perform various functions of the energy-controlled turbulence simulation device 410 based on continuous convolution by running or executing software programs stored in the memory 2002 and calling data stored in the memory 2002.

[0098] In a specific implementation, as one example, the first processor 2001 may include one or more CPUs, for example... Figure 6 CPU0 and CPU1 are shown in the diagram.

[0099] In a specific implementation, as one example, the energy-controllable turbulence simulation device 410 based on continuous convolution may also include multiple processors, for example... Figure 6 The first processor 2001 and the second processor 2004 are shown in the diagram. Each of these processors can be a single-core processor or a multi-core processor. Here, a processor can refer to one or more devices, circuits, and / or processing cores used to process data (such as computer program instructions).

[0100] The memory 2002 is used to store the software program that executes the present invention, and is controlled by the first processor 2001 to execute it. The specific implementation method can be referred to the above method embodiment, and will not be repeated here.

[0101] Optionally, the memory 2002 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory 2002 may be integrated with the first processor 2001 or may exist independently, and may be connected via the interface circuit of the energy-controlled turbulence simulation device 410 based on continuous convolution. Figure 6 (Not shown in the image) is coupled to the first processor 2001, and this embodiment of the invention does not specifically limit this.

[0102] The transceiver 2003 is used to communicate with network devices or with terminal devices.

[0103] Alternatively, transceiver 2003 may include a receiver and a transmitter. Figure 6 (Not shown separately). The receiver is used to implement the receiving function, and the transmitter is used to implement the transmitting function.

[0104] Optionally, the transceiver 2003 can be integrated with the first processor 2001 or exist independently, and can be connected via the interface circuit of the energy-controlled turbulence simulation device 410 based on continuous convolution. Figure 6 (Not shown in the image) is coupled to the first processor 2001, and this embodiment of the invention does not specifically limit this.

[0105] It should be noted that, Figure 6 The structure of the energy-controlled turbulence simulation device 410 based on continuous convolution shown does not constitute a limitation on the router. Actual knowledge structure identification devices may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0106] Furthermore, the technical effects of the energy-controlled turbulence simulation device 410 based on continuous convolution can be referred to the technical effects of the energy-controlled turbulence simulation method based on continuous convolution described in the above method embodiments, and will not be repeated here.

[0107] It should be understood that the first processor 2001 in the embodiments of the present invention may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0108] It should also be understood that the memory in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0109] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0110] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0111] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0112] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0113] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0114] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0115] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0116] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0117] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0118] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0119] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An energy-controlled turbulence simulation method based on continuous convolution, characterized in that, The method includes: S1. For the turbulent scenario to be simulated, particle method physical turbulence data corresponding to each physical solver is generated through multiple physical solvers. The particle method physical turbulence data is sampled and the particle velocity is calculated to obtain multiple model datasets. S2. Train a continuous convolutional network model based on each model dataset in multiple model datasets to obtain multiple continuous convolutional approximation model units. Perform particle displacement prediction based on each continuous convolutional approximation model unit to obtain the particle displacement prediction results of multiple continuous convolutional approximation model units. S3. Construct an energy interpolation network and input the particle displacement prediction results of multiple consecutive convolutional approximation model units into the energy interpolation network to obtain the interpolation ratio coefficients. S4. Based on the interpolation scaling factor, perform mixed interpolation on the particle displacement prediction results of multiple continuous convolution approximation model units to obtain the final simulation result.

2. The energy-controllable turbulence simulation method based on continuous convolution according to claim 1, characterized in that, In step S1, the physical turbulence data for each particle method are sampled and the particle velocity is calculated to obtain multiple model datasets, including: Each type of particle physics turbulence data is sampled at intervals. The particle velocity is calculated based on the particle position in each frame of the sampling results. The model dataset corresponding to each type of particle physics turbulence data is constructed based on the particle position and velocity in each frame, thus forming multiple model datasets.

3. The energy-controllable turbulence simulation method based on continuous convolution according to claim 1, characterized in that, Each continuous convolutional approximation model unit in S2 includes: a continuous convolutional kernel for extracting features of particles and performing particle displacement prediction.

4. The energy-controlled turbulence simulation method based on continuous convolution according to claim 1, characterized in that, In step S3, the particle displacement prediction results of multiple consecutive convolutional approximation model units are input into the energy interpolation network to obtain interpolation scaling coefficients, including: The target intensity of turbulent energy is obtained based on the particle displacement prediction results of multiple continuous convolution approximation model units. The particle displacement prediction results of multiple continuous convolution approximation model units and the target intensity of turbulent energy are input into the energy interpolation network to obtain the interpolation ratio coefficient.

5. The energy-controllable turbulence simulation method based on continuous convolution according to claim 4, characterized in that, The step of obtaining the target intensity of turbulent energy based on particle displacement prediction results from multiple consecutive convolutional approximation model units includes: The energy of turbulence is simplified to the kinetic energy of particles. The kinetic energy of particles is calculated from the predicted velocity of particles, and the predicted velocity of particles is calculated from the predicted displacement of particles. Then, the change function of the predicted particle displacement of multiple consecutive convolutional approximation model units is regarded as the change of turbulent energy. Based on the changes in turbulent energy, the upper and lower bounds of energy intensity of multiple continuous convolutional approximation model units are obtained. An energy error is constructed, and an energy interpolation network is used to penalize the energy error based on the upper and lower bounds of energy intensity, thereby approximating the target intensity of turbulent energy.

6. The energy-controllable turbulence simulation method based on continuous convolution according to claim 1, characterized in that, In step S4, the particle displacement prediction results of multiple consecutive convolutional approximation model units are mixed and interpolated according to the interpolation scaling factor to obtain the final simulation result, including: S41. The predicted velocity of the particle is calculated based on the particle displacement prediction results of multiple consecutive convolutional approximation model units. S42. Calculate the intermediate velocity generated by the advection process in the turbulent scene based on the predicted velocity of the particles. S43. The intermediate displacement generated by the advection process is calculated based on the particle displacement prediction results, the predicted velocity of the particles, and the intermediate velocity. S44. Calculate the mixed particle displacement prediction based on the interpolation scaling factor and the particle displacement prediction result, and calculate the particle displacement prediction result for the next frame based on the mixed particle displacement prediction and the intermediate displacement. S45. Repeat steps S41-S44 to produce the final simulation results.

7. An energy-controlled turbulence simulation device based on continuous convolution, wherein the energy-controlled turbulence simulation device based on continuous convolution is used to implement the energy-controlled turbulence simulation method based on continuous convolution as described in any one of claims 1-6, characterized in that, The device includes: The dataset construction module is used to generate particle method physical turbulence data for each physical solver for the turbulence scenario to be simulated. The particle method physical turbulence data is sampled and the particle velocity is calculated for each type of particle method physical turbulence data to obtain multiple model datasets. The displacement prediction module is used to train a continuous convolutional network model based on each model dataset in multiple model datasets, obtain multiple continuous convolutional approximation model units, and perform particle displacement prediction based on each continuous convolutional approximation model unit to obtain the particle displacement prediction results of multiple continuous convolutional approximation model units. The interpolation scaling factor calculation module is used to construct an energy interpolation network. It inputs the particle displacement prediction results of multiple consecutive convolutional approximation model units into the energy interpolation network to obtain the interpolation scaling factor. The output module is used to perform mixed interpolation on the particle displacement prediction results of multiple consecutive convolution approximation model units according to the interpolation scaling factor to obtain the final simulation result.

8. The energy-controllable turbulence simulation device based on continuous convolution according to claim 7, characterized in that, The process involves sampling and calculating particle velocities for each type of particle-based physical turbulence data to obtain multiple model datasets, including: Each type of particle physics turbulence data is sampled at intervals. The particle velocity is calculated based on the particle position in each frame of the sampling results. The model dataset corresponding to each type of particle physics turbulence data is constructed based on the particle position and velocity in each frame, thus forming multiple model datasets.

9. An energy-controllable turbulence simulation device based on continuous convolution, characterized in that, The energy-controlled turbulence simulation device based on continuous convolution includes: processor; A memory storing computer-readable instructions that, when executed by the processor, implement the method as described in any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium contains program code that can be invoked by a processor to execute the method as described in any one of claims 1 to 6.

Images