Discrete element simulation method based on adjoint computational domain

Abstract

The present invention discloses a discrete element simulation calculation method based on an adjoint computational domain, comprising the following steps: (1) importing a target device according to an actual industrial process and setting a simulation environment; (2) setting the grid size of the adjoint computational domain according to the particle size of the particle system in the actual industrial process; (3) setting the boundary extension value and size value of the adjoint computational domain according to the actual situation; (4) starting discrete element simulation calculation using the adjoint computational domain, and the adjoint computational domain is updated in real time according to the position of the target device at the current time step to simulate the particle system in the actual industrial process and obtain a simulation result. The simulation analysis method provided by the present invention solves the problems of huge memory consumption and low calculation efficiency in large-scale calculations. The method using the adjoint computational domain can greatly reduce memory consumption and improve calculation efficiency, meeting the needs of actual simulations; applying the simulation results to the optimization guidance of the actual industrial process can improve production efficiency.

Description

Technical Field

[0001] The present invention relates to the field of numerical calculations, and particularly to a discrete element simulation method based on an adjoint computational domain. Background Art

[0002] Granular materials are one of the substances widely used in current industrial processes. Common ones such as rocks, minerals, sediment, catalysts, biomass, pellets, tablets, etc. can all be regarded as granular substances. The movement law of particulate matter in industrial processes is relatively complex, and its influence on the process is very crucial. However, obtaining the behavior of particles through experimental methods is often costly and it is difficult to obtain information at the particle scale. Therefore, using the discrete element method for simulation has become an important means to understand the behavior of particulate matter in current industrial processes. For example, Chinese Patent with publication number CN110414139A discloses a simulation calculation method for the thermal processing of granular nuts in a drum, including the following steps: obtaining the physical property parameters of the granular nuts, the material property parameters and structural parameters of the drum, and the interaction parameters between the drum and the nuts and between the nuts; establishing a discrete element model of the granular nuts that conforms to the actual situation in a discrete element software according to the obtained physical property parameters of the granular nuts; establishing a three-dimensional solid model of the drum according to the structural parameters of the drum; converting the format of the three-dimensional solid model of the drum, importing it into the discrete element software, and then establishing a discrete element simulation model for the thermal processing of the granular nuts by adding conditions such as a heat exchange contact model and the rotation speed of the drum; and designing an orthogonal experiment to simulate the movement state and heat conduction situation of the nuts in the drum under different working conditions, and obtaining the optimal working condition through variance analysis.

[0003] When the industrial process is relatively complex, such as when the equipment scale is large or the movement is complex, the traditional computational domain generation method will include all the positions where particles may appear during the entire simulation process in the computational domain. This often leads to the situation that only a local area in the entire computational domain contains particles at a certain moment, while the contact detection between particles is still carried out on the entire computational domain. Since the grid size during particle contact detection is comparable to the particle size in the particle system, when the particle size is small, using the above traditional computational domain generation method will cause a large amount of memory consumption and increase the computational time cost, greatly reducing the simulation efficiency.

[0004] Industrial processes often involve a large number of particle systems, such as in the fields of minerals, civil engineering, energy, chemical engineering, and pharmaceuticals. In specific industrial processes, the number of particles involved is often extremely large, and the scale of the particles is relatively small compared to the size of the equipment. For example, in the ball milling process in the minerals field, the materials and media in the ball mill are relatively small in size compared to the equipment, while the number of materials and media in the ball mill is extremely large. This results in a large size of the computational domain in the discrete element simulation of the ball milling process, but a small grid size for contact detection, ultimately generating a large number of grids within the computational domain, greatly increasing the memory consumption. Moreover, for some processes with more complex equipment movements, such as the ball milling process in a planetary ball mill. A planetary ball mill is an essential device for mixing, fine grinding, small sample preparation, nanomaterial dispersion, new product development, and small-batch production of high-tech materials, and is widely used in departments such as geology, minerals, metallurgy, electronics, building materials, ceramics, chemical engineering, light industry, medicine, beauty, and environmental protection. During the normal operation of a planetary ball mill, the ball milling tank not only revolves around the axis with the bottom plate but also rotates around its own axis. Therefore, in discrete element simulation, if the traditional computational domain generation method is adopted, the positions passed by the ball milling tank throughout the simulation need to be included in the computational domain, resulting in an overly large computational domain. And within each time step, the particles in the computational domain are only concentrated in a relatively small local space, thus causing a waste of a large amount of memory and also leading to an increase in time costs. Summary of the Invention

[0005] The purpose of the present invention is to provide a discrete element simulation analysis method based on an adjoint computational domain to solve the problems of an overly large computational domain in complex industrial processes, thereby causing memory waste and low computational efficiency; improving the memory utilization rate and computational efficiency in large-scale calculations, and meeting the simulation needs of actual industrial processes.

[0006] The present invention provides the following technical solutions:

[0007] A discrete element simulation method based on an adjoint computational domain, comprising the following steps:

[0008] (1) Import the target equipment according to the actual industrial process and set the simulation environment;

[0009] (2) Set the grid size of the adjoint computational domain according to the particle size of the particle system in the actual industrial process;

[0010] (3) Set the boundary expansion value and size value of the adjoint computational domain according to the actual situation. The boundary expansion value is the size by which the adjoint computational domain expands outward based on the minimum position of the target equipment in the global coordinate system, and the size value is the size of the length, width, and height of the adjoint computational domain;

[0011] (4) Use the adjoint computational domain to start the discrete element simulation calculation. The adjoint computational domain is updated in real time according to the position of the target device at the current time step, simulating the particle system in the actual industrial process to obtain the simulation results.

[0012] Among them, during the simulation process, the position and size of the adjoint computational domain are determined according to the minimum position of the target device in the global coordinate system, the set boundary extension value, and the size value at each time step. The obtained adjoint computational domain needs to envelope the target device in the global coordinate system.

[0013] In the invention, the industrial process contains a large number of particles with small sizes. Using the discrete element simulation analysis method based on the adjoint computational domain can greatly reduce the memory required for the simulation and lower the time cost, which is very important for large-scale industrial-level simulation operations.

[0014] In step (1), the set simulation environment is a discrete element simulation.

[0015] In step (1), the actual industrial process is an industrial process containing a large number of particles such as mineral ball milling, particle crushing, material dispersion, or pellet coating.

[0016] In step (1), the target device is the main device in the corresponding industrial process, such as a ball mill, a bucket wheel, a mixer, or a coating device.

[0017] In step (2), the mesh size of the adjoint computational domain is comparable to the particle size in the particle system.

[0018] In step (3), the boundary extension value is the distance extended outward based on the minimum position of the device at the current time step. Under the action of the boundary extension value, the boundaries of the computational domain in these three directions will surely ensure the envelope of the target device. If the particles are only located inside the target device during the simulation process, the setting of the boundary extension value only needs to meet the stability of the calculation, so it can be set to a relatively small value; if the particles will appear outside the device during the simulation process, such as during the discharging process, the setting of the boundary extension value needs to ensure the accuracy of the reproduction of this process while ensuring the calculation stability. At this time, the boundary extension value needs to be set according to the actual situation. Generally speaking, the boundary extension value is less than one-tenth of the size of the target device.

[0019] In step (3), the dimension values are the values accompanying the length, width, and height of the computational domain. In the discrete element method calculation, the computational domain needs to contain all the particles within the system, and the particles outside the computational domain will not participate in the discrete element calculation. The accompanying computational domain also needs to meet this requirement. Therefore, the dimension values of the accompanying computational domain are closely related to the dimensions and motion mode of the device. If the device does not rotate during the motion process, the dimension values of the accompanying computational domain should be greater than the lengths of the target device along the three coordinate axes in the global coordinate system; if the target device rotates, the dimension values of the accompanying computational domain should be greater than the maximum values of the lengths of the target device along the three coordinate axes during the motion process.

[0020] In step (3), at each time step, the lowest points of the target device along the x, y, and z coordinate axes in the global coordinate system are obtained, that is, the minimum values of the target device along the three coordinate axes in the global coordinate system are obtained. The boundary position of the accompanying computational domain in the global coordinate system is controlled by the boundary extension value. The size of the entire accompanying computational domain is controlled by the dimension values.

[0021] In step (4), according to the minimum position of the target device in the global coordinate system and the boundary extension value at each time step, the positions of the bottom surface, side surfaces, and rear surface of the accompanying computational domain are determined, and then combined with the dimension values to obtain a hexahedron as the accompanying computational domain at each time step.

[0022] Regarding the positions of the bottom surface, side surfaces, and rear surface, it can be understood as: the positions of the side surface (perpendicular to the x-axis), bottom surface (perpendicular to the y-axis), and rear surface (perpendicular to the z-axis) in the global coordinate system.

[0023] In step (4), at each time step, the target device will complete the update according to the set motion mode. The target device sets the motion mode according to the actual situation. In the present invention, the position of the accompanying computational domain at each time step is related to the position of the target device in the corresponding time step. When the target device completes the position update according to the set motion mode at the current time step, the accompanying computational domain will also complete the update according to the current position of the target device.

[0024] The simulation results obtained according to step (4) can be used to optimize and guide the corresponding real industrial process. That is, the simulation results are applied to the actual large-scale industrial process to provide technical support for the real industrial process. For example, for the ball milling process in a planetary ball mill, the simulation results are the wear conditions of the particles and the device, which can provide technical support for the rotation speed and filling amount of the ball mill tank involved in the ball milling process.

[0025] Due to the complex interactions between particulate matters and the significant impact of particle collisions on the motion behavior of particles, in the simulation analysis using the discrete element method, the contact detection between particles is a very important step in the simulation process and also a process that consumes a large amount of memory and time. The conventional contact detection uses the grid method. First, the computational domain is defined, and then the entire computational domain is divided into a certain number of grids, with the grid size being comparable to the particle size in the particle system. When performing contact detection, only when the grid where a particle is located is adjacent to the grid where the target particle is located, will it be further determined whether the two particles are in contact. Therefore, the number of grids has a great impact on the computational efficiency and memory usage of the overall simulation. The accompanying computational domain method proposed in the present invention has the same setting of the grid size as that in the traditional computational domain method, both being comparable to the particle size in the particle system. However, since the accompanying computational domain proposed in the present invention generates a local computational domain according to the position of the device at the current time step, compared with the traditional computational domain technology, it can greatly reduce the total computational memory. Therefore, under the same grid size, by using the accompanying computational domain method proposed in the present invention, the number of grids required for the contact detection process can be greatly reduced, thereby reducing the memory consumption during the calculation process and improving the memory utilization rate and computational efficiency.

[0026] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0027] By updating the computational domain at each time step, the present invention combines the accompanying computational domain with the discrete element simulation, solving the problem of excessive memory consumption and time cost in discrete element simulation under complex conditions. Combining the accompanying computational domain method with the discrete element simulation can greatly reduce the memory consumption during the simulation process, improve the computational efficiency, enhance the efficiency of discrete element simulation in dealing with complex processes, and has guiding significance for the simulation optimization of large-scale industrial processes. Brief Description of the Drawings

[0028] Figure 1 is the flowchart of the discrete element simulation method based on the accompanying computational domain;

[0029] Figure 2 is the schematic diagram using the traditional computational domain;

[0030] Figure 3 is the schematic diagram using the accompanying computational domain;

[0031] Figure 4 is the schematic diagram of the boundary expansion value and size value of the accompanying computational domain;

[0032] Figure 5 is the schematic diagram of the simulation of the planetary ball mill. Detailed Embodiments

[0033] The following further elaborates on the content of the present invention in conjunction with the accompanying drawings and embodiments.

[0034] As Figure 1 shown, the discrete element simulation method based on the adjoint computational domain provided by the present invention includes the following steps:

[0035] (1) Import the target device according to the actual industrial process and set the simulation environment;

[0036] (2) Set the grid size of the adjoint computational domain according to the particle size of the particle system in the actual industrial process;

[0037] (3) Set the boundary extension value and size value of the adjoint computational domain according to the actual situation. The boundary extension value is the size by which the adjoint computational domain extends outward based on the minimum position of the target device in the global coordinate system, and the size value is the size of the length, width, and height of the adjoint computational domain;

[0038] (4) Use the adjoint computational domain to start the discrete element simulation calculation. The adjoint computational domain is updated in real time according to the position of the target device at the current time step, simulating the particle system in the actual industrial process to obtain the simulation results.

[0039] Among them, during the simulation process, the position and size of the adjoint computational domain are determined according to the minimum position of the target device in the global coordinate system, the set boundary extension value, and size value at each time step. The obtained adjoint computational domain needs to envelope the target device in the global coordinate system.

[0040] Embodiment 1

[0041] The following explains the above discrete element simulation calculation method based on the adjoint computational domain in combination with the actual process. The process involved is the ball milling process, and the device used is a small planetary ball mill on a laboratory scale.

[0042] The planetary ball mill is an essential device for mixing, fine grinding, small sample preparation, nanomaterial dispersion, new product development, and small-scale production of high-tech materials. It is widely used in departments such as geology, minerals, metallurgy, electronics, building materials, ceramics, chemical engineering, light industry, medicine, beauty, and environmental protection. During the normal operation of the planetary ball mill, the ball mill tank not only revolves around the axis with the bottom plate but also rotates around its own axis. Therefore, in discrete element simulation, if the traditional computational domain generation method is used, all the positions passed by the ball mill tank during the simulation process need to be included in the computational domain, resulting in a huge computational domain. During the calculation process, the particles are only concentrated in a relatively small local space at each time step, causing a large waste of memory and an increase in time cost. As Figure 2 shown, Figure 2 is a schematic diagram when using the traditional computational domain. From Figure 2It can be seen that with the traditional calculation domain generation method, the calculation domain remains unchanged throughout the simulation process, and all positions passed by the ball mill during the simulation process need to be included in the calculation domain. However, at each time step, the particles only exist in the local space where the ball mill is located, and there are no particles in a large amount of space in the traditional calculation domain, resulting in a large amount of memory waste.

[0043] Figure 3 The following shows a schematic diagram when using the adjoint calculation domain. As can be seen from Figure 3 it that at each time step, the calculation domain is updated in real time according to the position of the current device, ensuring that the calculation domain contains all the particles in the simulation while avoiding a large amount of wasted calculation memory and greatly improving the memory utilization rate.

[0044] Example 2

[0045] The following explains the setting of the adjoint calculation domain in combination with a specific simulation example.

[0046] First, determine the grid size in the calculation domain according to the particle size in the system, and determine the adjoint relationship between the calculation domain and the ball mill. That is, in the subsequent simulation calculations, the position of the adjoint calculation domain will change with the change of the position of the ball mill.

[0047] Figure 4 The following shows the state diagram of the device at a certain time step. Only the state in the x-y plane is shown in the figure, and the settings in the z direction are the same as those in the x and y directions. Due to the adjoint relationship between the calculation domain and the ball mill, the adjoint calculation domain will be generated based on the position of the target device, the ball mill. As Figure 4 shown, at the current time step, after obtaining the lowest points of the device in the three coordinate axes directions, through three boundary extension values, the boundary positions of the associated calculation domain in the three coordinate axes directions can be obtained, that is, the positions of the left side (perpendicular to the x axis), the bottom (perpendicular to the y axis), and the back (perpendicular to the z axis) of the calculation domain. After the three boundaries are determined, the length, width, and height of the calculation domain are determined by three size values. The three boundary extension values plus the three size values can obtain a hexahedron-shaped adjoint calculation domain. It should be noted that the set associated calculation domain needs to be able to completely enclose all the particles in the simulation.

[0048] Example 3

[0049] The following explains the reduction of memory consumption and the improvement of calculation efficiency caused by using the adjoint calculation domain in combination with a specific simulation example.

[0050] Figure 5 is a simulation schematic diagram of a planetary ball mill. The simulation is carried out using the traditional calculation domain and the adjoint calculation domain respectively, and the differences in memory consumption and calculation efficiency between the two methods are compared.

[0051] Figure 5 Shown is a schematic diagram of a ball mill tank in the simulation. The average diameter of the particles in the tank is 0.5 mm and the number is 500,000. If the traditional computational domain generation method is used, the size of the computational domain required in the simulation is as Figure 2 shown. The size in the x direction ranges from -0.239 m to 0.364 m, the size in the y direction ranges from -0.4 m to 0.4 m, and the size in the z direction ranges from -0.0611 m to 0.0609 m. The memory consumption after starting the calculation is 12261.8 Mb.

[0052] When using the adjoint computational domain, the boundary extension sizes in all three directions are set to 0.001 m. The size value in the x direction is 0.125 m, the size value in the y direction is 0.085 m, and the size value in the z direction is 0.122 m. The schematic diagram of the adjoint computational domain is as Figure 3 shown. When the simulation is carried out with the same parameters, the memory consumption after starting the calculation is 1220.7 Mb. It can be seen that, compared with the traditional computational domain generation method, the memory consumption is reduced to one-tenth of the original. In summary, it can be seen that using the adjoint computational domain can greatly reduce the memory consumption.

[0053] Next, the traditional computational domain method and the adjoint computational domain method are used for single-core calculations respectively. Both calculations are carried out on the same computer using a single core. After statistics, it can be found that if the calculation time of the traditional computational domain method is used for normalization, then the time required to calculate the same number of time steps using the adjoint computational domain method is about 0.8. It can be found that using the adjoint computational domain can improve the calculation efficiency.

[0054] It should be noted that the current simulation only uses a scaled laboratory-scale device. When the device size is larger, and the particle diameter is smaller and the number is larger, reaching the micron level, the memory consumption of using the traditional computational domain generation method will show an exponential increase, and in some cases, it will cause an unaffordable huge memory consumption. At this time, using the adjoint computational domain method proposed in the present invention can greatly reduce the memory consumption.

Claims

1. A discrete element simulation method based on an adjoint computational domain, characterized in that It includes the following steps: (1) Import the target device according to the actual industrial process and set up the simulation environment; (2) Set the grid size of the adjoint computational domain according to the particle size of the particle system in the actual industrial process; (3) Set the boundary extension value and size value of the adjoint computational domain according to the actual situation. The boundary extension value is the size by which the adjoint computational domain extends outward based on the minimum position of the target device in the global coordinate system, and the size value is the size of the length, width, and height of the adjoint computational domain; (4) Start the discrete element simulation calculation using the adjoint computational domain. The adjoint computational domain is updated in real time according to the position of the target device at the current time step to simulate the particle system in the actual industrial process and obtain the simulation results; Among them, during the simulation process, the position and size of the adjoint computational domain are determined according to the minimum position of the target device in the global coordinate system, the set boundary extension value, and size value at each time step to ensure that the adjoint computational domain envelopes the target device in the global coordinate system.

2. The discrete element simulation method based on an adjoint computational domain according to claim 1, wherein In step (1), the actual industrial process is mineral ball milling, particle crushing, material dispersion, or pellet coating.

3. The discrete element simulation method based on the adjoint computational domain according to claim 1, characterized in that In step (1), the target device is a ball mill, a bucket wheel, a mixer, or a coating device.

4. The discrete element simulation method based on an adjoint computational domain according to claim 1, characterized in that In step (2), the grid size of the adjoint computational domain is comparable to the particle diameter in the particle system.

5. The discrete element simulation method based on the adjoint computational domain according to claim 1, wherein In step (3), the boundary extension value is less than one-tenth of the target device.

6. The discrete element simulation method based on the adjoint computational domain according to claim 1, wherein In step (3), if the target device does not rotate during the movement process, the size value of the adjoint computational domain should be greater than the lengths of the target device along the three coordinate axes in the global coordinate system; if the target device rotates, the size value of the adjoint computational domain should be greater than the maximum value of the lengths of the target device along the three coordinate axes during the movement process.

7. The discrete element simulation method based on an adjoint computational domain according to claim 1, wherein In step (4), determine the positions of the adjoint computational domain on the bottom, side, and back according to the minimum position of the target device in the global coordinate system and the boundary extension value at each time step, and then combine with the size value to obtain a hexahedron as the adjoint computational domain at each time step.

8. The discrete element simulation method based on an adjoint computational domain according to claim 1, wherein, In step (4), the target device moves according to the set motion mode at each time step.

Images