CN-121997798-A - Energy-controllable turbulence simulation method and device based on continuous convolution

CN121997798ACN 121997798 ACN121997798 ACN 121997798ACN-121997798-A

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 steps of generating physical turbulence data of a particle method through a plurality of physical solvers, sampling and calculating particle speeds to obtain a plurality of model data sets, training a continuous convolution network model to obtain a plurality of continuous convolution approximate model units, predicting particle displacement according to the model units, inputting a particle displacement prediction result into an energy interpolation network to obtain an interpolation proportionality coefficient, and carrying out mixed interpolation on the particle displacement prediction results of the plurality of continuous convolution approximate model units according to the interpolation proportionality coefficient to obtain a final simulation result. The turbulence energy variation level generated by the invention maintains positive correlation with the control parameter, proves the effectiveness of the energy intensity coefficient for controlling the turbulence energy, and shows obvious trend of transition from laminar flow to turbulent flow on visual effect.

Inventors

WANG XIAOKUN
Yan Zhexi
XU YANRUI
Zeng Haokai
ZHANG YALAN
YAO CHAO

Assignees

北京科技大学顺德创新学院

Dates

Publication Date: 20260508
Application Date: 20251217

Claims (10)

1. A method of energy-controllable turbulence simulation based on continuous convolution, the method comprising: s1, for a turbulence scene to be simulated, generating particle method physical turbulence data corresponding to each physical solver through a plurality of physical solvers, respectively sampling each particle method physical turbulence data and calculating particle speed to obtain a plurality of model data sets; S2, training a continuous convolution network model according to each model data set in a plurality of model data sets to obtain a plurality of continuous convolution approximate model units, and respectively carrying out particle displacement prediction according to each continuous convolution approximate model unit to obtain particle displacement prediction results of the plurality of continuous convolution approximate model units; s3, constructing an energy interpolation network, and inputting particle displacement prediction results of a plurality of continuous convolution approximate model units into the energy interpolation network to obtain an interpolation proportionality coefficient; And S4, performing mixed interpolation on the particle displacement prediction results of the plurality of continuous convolution approximate model units according to the interpolation proportion coefficient to obtain a final simulation result.
2. The continuous convolution based energy controllable turbulence simulation method according to claim 1, wherein the step of sampling each particle method physical turbulence data and calculating particle speeds in the step S1 respectively to obtain a plurality of model data sets comprises the following steps: And respectively carrying out interval sampling on the physical turbulence data of each particle method, calculating the particle speed according to the particle position of each frame in the sampling result, and constructing a model data set corresponding to the physical turbulence data of each particle method according to the particle position and the particle speed of each frame so as to further form a plurality of model data sets.
3. The continuous convolution based energy controllable turbulence simulation method according to claim 1, wherein each continuous convolution approximation model unit in S2 comprises a continuous convolution kernel for extracting features of particles and performing particle displacement prediction.
4. The continuous convolution-based energy-controllable turbulence simulation method according to claim 1, wherein the step of inputting the particle displacement prediction results of the plurality of continuous convolution approximation model units into the energy interpolation network in S3 to obtain an interpolation scaling factor comprises: and obtaining the target intensity of the turbulence energy according to the particle displacement prediction results of the continuous convolution approximate model units, and inputting the particle displacement prediction results of the continuous convolution approximate model units and the target intensity of the turbulence energy into an energy interpolation network to obtain an interpolation proportionality coefficient.
5. The continuous convolution based energy controllable turbulence simulation method according to claim 4, wherein the obtaining the target intensity of turbulence energy according to the particle displacement prediction result of the plurality of continuous convolution approximation model units comprises: simplifying turbulent energy into kinetic energy of particles, wherein the kinetic energy of the particles is calculated by the predicted speed of the particles, the predicted speed of the particles is calculated by the predicted result of the particle displacement, and then the change function of the predicted result of the particle displacement of a plurality of continuous convolution approximate model units is regarded as turbulent energy change; and obtaining the energy upper limit intensity and the energy lower limit intensity of a plurality of continuous convolution approximate model units according to the turbulence energy change, constructing an energy error, punishing an energy interpolation network based on the energy upper limit intensity and the energy lower limit intensity by using the energy error, and further approaching the target intensity of the turbulence energy.
6. The energy-controllable turbulence simulation method based on continuous convolution according to claim 1, wherein the step S4 of performing hybrid interpolation on the particle displacement prediction results of the plurality of continuous convolution approximation model units according to the interpolation scaling factor to obtain a final simulation result comprises: s41, calculating the prediction speed of particles according to the particle displacement prediction results of a plurality of continuous convolution approximate model units; S42, calculating to obtain the intermediate speed generated in the advection process of the turbulence scene according to the predicted speed of the particles; S43, calculating to obtain intermediate displacement generated in the advection process according to the particle displacement prediction result, the particle prediction speed and the intermediate speed; S44, calculating to obtain mixed particle displacement prediction according to the interpolation proportionality coefficient and the particle displacement prediction result, and calculating to obtain a particle displacement prediction result of the next frame according to the mixed particle displacement prediction and the intermediate displacement; s45, repeatedly executing the steps S41-S44 to generate a final simulation result.
7. An energy-controllable turbulence simulation device based on continuous convolution for implementing the energy-controllable turbulence simulation method based on continuous convolution according to any one of claims 1-6, characterized in that the device comprises: The data set construction module is used for generating particle method physical turbulence data corresponding to each physical solver through a plurality of physical solvers for a turbulence scene to be simulated, respectively sampling each particle method physical turbulence data and calculating particle speed to obtain a plurality of model data sets; the displacement prediction module is used for training a continuous convolution network model according to each model data set in a plurality of model data sets to obtain a plurality of continuous convolution approximate model units, and respectively carrying out particle displacement prediction according to each continuous convolution approximate model unit to obtain particle displacement prediction results of the plurality of continuous convolution approximate model units; the interpolation proportionality coefficient calculation module is used for constructing an energy interpolation network, and inputting particle displacement prediction results of a plurality of continuous convolution approximate model units into the energy interpolation network to obtain an interpolation proportionality coefficient; And the output module is used for carrying out mixed interpolation on the particle displacement prediction results of the plurality of continuous convolution approximate model units according to the interpolation proportion coefficient to obtain a final simulation result.
8. The continuous convolution based energy controllable turbulence simulation apparatus according to claim 7, wherein the sampling each kind of particle method physical turbulence data and calculating particle velocity respectively to obtain a plurality of model data sets comprises: And respectively carrying out interval sampling on the physical turbulence data of each particle method, calculating the particle speed according to the particle position of each frame in the sampling result, and constructing a model data set corresponding to the physical turbulence data of each particle method according to the particle position and the particle speed of each frame so as to further form a plurality of model data sets.
9. An energy controllable turbulence simulation device based on continuous convolution, characterized in that the energy controllable turbulence simulation device based on continuous convolution comprises: A processor; A memory having stored thereon computer readable instructions which, when executed by the processor, implement the method of any of claims 1 to 6.
10. A computer readable storage medium having stored therein program code which is callable by a processor to perform the method of any one of claims 1 to 6.

Description

Energy-controllable turbulence simulation method and device based on continuous convolution Technical Field The invention relates to the technical field of computer graphics fluid simulation, in particular to an energy-controllable turbulence simulation method and device based on continuous convolution. Background Turbulence is a collective term for various complex and delicate fluid phenomena such as eddies, smoke, waves, and the like. Computer modeling of turbulence is one of the hot spot orientations in the field of computer graphics. Turbulence has a much richer detail than laminar flow. The difficulty with turbulence modeling is how to balance the degree of refinement of the vortex with the time-space cost of the simulation while maintaining the natural generation and attenuation of the vortex. Currently, there are a number of particle-based turbulence simulation methods that simulate a variety of turbulence phenomena with eddies. The conventional smooth particle fluid dynamic method can completely track the vorticity information of each particle by introducing a vorticity equation on the basis of the vortex particle method, calculating the vorticity loss in each frame and performing compensation, but the sum of Biot-Savart (Biot-Savart) converted from the vorticity compensation to a velocity field can bring about expensive calculation time cost, and the water surface is easy to generate long-time oscillation behavior along with the accumulation of corrected positive feedback, so that the water surface cannot be calm for a long time. In addition, the physical solver determines the speed displacement generated by each frame due to strict physical constraint, and the controllable range of energy does not exist, so that the expansion of the simulation effect is limited. For previous turbulence simulators, excessive velocity compensation may result in oscillations and inability to converge in energy, thereby creating a fluid surface that is unable to calm for a long period of time. Such solvers generally rely on adjusting initial parameters and repeating simulations to indirectly affect energy, and there is typically no direct correspondence between simulation parameters and energy, which is not a direct control of energy. The continuous convolution is used as a mode for extracting the characteristics of the particle data, and can be used for characteristic learning of turbulent flow data and a particle displacement prediction process. Particle neighborhood information is extracted by using a continuous convolution kernel and deviations in particle motion displacement are penalized to learn fluid motion characteristics. The resulting correction does not need to follow exactly the Navier-Stokes equation compared to a physical solver, with a more relaxed conditional constraint, which offers the possibility of dynamically adjusting the turbulence energy in each time step. Disclosure of Invention In order to solve the technical problems that the existing physical solver is insufficient in turbulence details and difficult to directly control turbulence energy due to vortex loss or the water surface cannot be stabilized for a long time due to excessive energy compensation when simulating turbulence, the embodiment of the invention provides an energy-controllable turbulence simulation method and device based on continuous convolution. The technical scheme is as follows: in one aspect, there is provided a method of energy-controllable turbulence simulation based on continuous convolution, the method being implemented by an energy-controllable turbulence simulation device based on continuous convolution, the method comprising: s1, for a turbulence scene to be simulated, generating particle method physical turbulence data corresponding to each physical solver through a plurality of physical solvers, respectively sampling each particle method physical turbulence data, and calculating particle speeds to obtain a plurality of model data sets. S2, training a continuous convolution network model according to each model data set in a plurality of model data sets to obtain a plurality of continuous convolution approximate model units, and respectively carrying out particle displacement prediction according to each continuous convolution approximate model unit to obtain particle displacement prediction results of the plurality of continuous convolution approximate model units. S3, constructing an energy interpolation network, and inputting particle displacement prediction results of a plurality of continuous convolution approximate model units into the energy interpolation network to obtain an interpolation proportionality coefficient. And S4, performing mixed interpolation on the particle displacement prediction results of the plurality of continuous convolution approximate model units according to the interpolation proportion coefficient to obtain a final simulation result. Optionally, in S1, each particle method physical