Gas-solid two-fluid simulation method for bubbling fluidized bed

Abstract

A gas-solid two-fluid simulation method for a bubbling fluidized bed includes: S1. Using Solidworks to establish a three-dimensional model of the bubbling bed and importing the model into open source software; S2. Setting physical property parameters and establishing a microscale model of particle dynamics; S3. Refining the simulation based on the two-fluid method, setting different working conditions for simulation, verifying the effectiveness of TFM simulation when the grid is smaller than the particle diameter, and establishing a data connection between the coarse grid simulation and the fine grid simulation; S4. Calculating the two-phase drag force using the fine grid simulation data; S5. Correlation analysis between the refined data and the two-phase drag force, determining the input and output parameters of the training, and normalizing the data; S6. Building a complete artificial neural network training model; using Python tools to establish a nonlinear fitting relationship between the input and output parameters through the artificial neural network, and constructing a drag model; S7. Implanting the self-built model through secondary development, applying the model to large-scale industrial equipment, and verifying the model.

Description

Technical Field

[0001] The invention belongs to the technical field of computer simulation, and in particular relates to a method for simulating gas-solid dual fluids in a bubbling fluidized bed. Background Art

[0002] In recent years, the simulation of gas-solid flows within large-scale industrial equipment has garnered widespread attention. The two-fluid model's unrestricted particle number has led to its widespread application in the simulation of industrial equipment (fluidized beds, boilers, etc.). Currently, two-fluid simulation equipment is primarily laboratory-scale and has not been extensively applied to large-scale industrial equipment. When combined with small-scale reactors, the classic two-fluid model allows for the observation of detailed structures, both for single particles and particle aggregates, in fine-grid simulations, and provides relatively accurate flow field information. However, when combined with large-scale reactors, fine-grid simulations miss some of the heterogeneous structures caused by particle agglomeration; while coarse-grid simulations make it virtually impossible to observe heterogeneous structures within the flow field, such as particle agglomeration. This suggests that despite its widespread application, the two-fluid model is still subject to numerous limitations. Industrial fluidized bed reactors are typically large, often requiring coarse grids for simulation. However, two-fluid models that fail to account for the heterogeneous gas-solid structures within the grid struggle to predict reasonable fluid velocity and pressure characteristics.

[0003] In actual engineering applications, real-time simulation is not feasible to resolve various operational issues encountered by engineering equipment. Achieving high precision on a large scale is challenging due to both simulation duration and spatial storage requirements. This requires not only improving simulation resolution but also revising the model. While closing the model, computational speeds are accelerated to improve simulation efficiency. This addresses the shortcomings of dual-fluid simulations at multiple and mesoscales, making them more suitable for industrial-scale simulations. Summary of the Invention

[0004] The present invention aims to overcome the above-mentioned shortcomings of the prior art and provides a gas-solid two-fluid simulation method for a bubbling fluidized bed.

[0005] The bubbling fluidized bed gas-solid two-fluid simulation method of the present invention comprises the following specific steps:

[0006] 1. Use Solidworks to build a 3D model of the bubbling bed and import the model into open source software.

[0007] 2. Set physical property parameters and establish a micro-scale model of particle dynamics:

[0008]

[0009]

[0010]

[0011]

[0012] ε is the gas-solid two-phase concentration, ρ is the gas-solid two-phase density, v / u is the gas-solid two-phase velocity, p is the gas-solid two-phase pressure, d p is the particle diameter, and τ is the shear stress.

[0013] 3. Based on the two-fluid method, the simulation is refined and different working conditions are set for simulation to verify the effectiveness of TFM simulation when the grid is smaller than the particle diameter, and to establish the data connection between the coarse grid simulation and the fine grid simulation.

[0014] 4. Calculate the two-phase drag force using fine grid simulation data.

[0015]

[0016]

[0017]

[0018] Re is the Reynolds number, which is related to the two-phase density, concentration and gas phase viscosity;

[0019] 5. Perform correlation analysis on the refined data and the two-phase drag force, determine the input and output parameters of the training, and normalize the data.

[0020] 6. Use the third-party tool Python to build a complete artificial neural network training model; use Python tools to establish a nonlinear fitting relationship between input and output parameters through artificial neural networks to construct a drag calculation model.

[0021] 7. Through secondary development, the self-built model is implanted and applied to large-scale industrial equipment to verify the effectiveness and reliability of the model.

[0022] The working principle of this invention is to combine an artificial neural network with a traditional two-fluid simulation of bubbling bed flow. The traditional simulation obtains flow field structure information, which is then trained using a neural network to obtain a mathematical model that can completely reflect the flow field structure. This model is then added to the custom model section of the traditional model.

[0023] There are currently three problems with traditional simulation methods: insufficient storage capacity; long simulation time; and poor simulation accuracy. In order to solve these existing problems, the present invention proposes an artificial neural network combined with a dual-fluid reactor flow field simulation. The combination of the two can calculate the flow characteristic distribution in the bubbling bed at a fine scale, and obtain the two-phase motion velocity and trajectory. By combining the artificial neural network, a fine flow field characteristic distribution model is obtained for secondary development. Applying the distribution model at the microscopic scale to a large-scale industrial reactor at a macroscopic scale, on the one hand, ensures the calculation accuracy, and on the other hand, saves space storage and calculation time. Compared with the traditional dual-fluid simulation, the simulation results of the dual-fluid model combined with the secondary development of the artificial neural network are closer to the actual situation. At the same time, the combination of artificial neural networks and dual-fluid simulation methods is of great significance for improving the particle dynamics model.

[0024] The advantages of this method are that it can, to a certain extent, overcome the three major issues currently associated with traditional two-fluid simulation. For large-scale industrial reactors, this method can be used to simulate coarse grids, resolving issues of simulation accuracy and time. Application of this method in the field of CFD can also provide a reference for subsequent researchers to refine their models, thereby improving traditional two-fluid simulation methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a schematic diagram comparing the simulation value and the fitting value of the present invention.

[0026] Figure 2 It is a schematic diagram of bed expansion height under different grid sizes of the present invention.

[0027] Figure 3 Schematic diagram of radial distribution of particles at different grid sizes of the present invention.

[0028] Figures 4(a) and 4(b) are comparison diagrams of the fitted values ​​and calculated values ​​of the present invention, wherein Figure 4(a) is a schematic diagram of the Gidaspow model and Figure 4(b) is a schematic diagram of the theoretical values.

[0029] Figures 5(a) and 5(b) show the instantaneous concentration distribution at 5 seconds of the present invention, wherein Figure 5(a) is a schematic diagram of the optimization model and Figure 5(b) is a schematic diagram of the Gidaspow model. DETAILED DESCRIPTION

[0030] The technical solution of the present invention is further described below with reference to the accompanying drawings.

[0031] The specific steps of the gas-solid two-fluid simulation method of the bubbling fluidized bed are as follows:

[0032] Step 1: Create a bubbling fluidized bed model using Solidworks and import it into open source software.

[0033] Step 2: Set specific parameters for bubbling bed simulation, select the Gidaspow drag model, G0 as the radial distribution function, select the two-fluid simulation (TFM) solution method, and construct the traditional particle dynamics equation.

[0034] Step 3: Perform global and local simulations on the bubbling bed. p =2.5mm) and set seven global simulation conditions with different mesh sizes: 10mm, 7.5mm, 5mm, 2.5mm, 2mm, 1.25mm, and 1mm. Physical properties such as particle size and restitution coefficient were set identically. TFM local fine-scale simulations were performed using periodic boundaries, also with different mesh size settings. The mesh sizes were 250µm, 150µm, and 100µm.

[0035] Step 4: Analyze the global and local simulation results to verify the effectiveness of TFM simulation when the mesh is smaller than the particle diameter, and calculate the two-phase drag force based on the simulation results.

[0036] Step 5: Use Origin to perform a Pearson correlation analysis between the simulation result parameters (primarily two-phase velocities and concentrations) and the calculated viscosity drag to determine the correlation between the result parameters and the drag. The gas-solid two-phase velocities and concentrations serve as the neural network input layer, and the two-phase drag serves as the neural network output layer. Sigmoid and Tanh functions are used to normalize the input and output parameters so that all parameters are within the range of 0 to 1, ensuring that the raw data for neural network training are of the same magnitude.

[0037] Step 6: Select the Rule function as the activation function for artificial neural network training, ensuring that the number of neurons is greater than the number of input parameters and the learning rate is within the range of 0 to 1. The neural network training structure is continuously adjusted to reduce the loss function value. Repeat the training multiple times to establish a clear nonlinear relationship between velocity, concentration, and drag force, ultimately obtaining a mathematical model of drag force with respect to velocity and concentration.

[0038] Step 7: Use Fortran language to write the model into the uesr.f file of the open source software. Select the uesr.f custom model to perform coarse grid simulation of large industrial equipment (the simulation size is determined according to the actual size of the factory) and analyze the gas-solid two-phase flow simulation results.

[0039] In this invention, the solid phase is glass beads and the gas phase is air. A two-fluid simulation method is used to study the gas-solid two-phase flow in the bubbling bed. The specific simulation parameters are as follows:

[0040] Solid phase density 2500kg / m 3, the superficial velocity is 1.5 times the fluidization velocity, the system simulation pressure is atmospheric pressure 101325Pa, the system temperature is 298K, and the time step is 10 -4 s, both microscale and macroscale are simulated using the two-fluid method in combination with the particle dynamics equations.

[0041] The above description is based on a specific fluidized bed and is not limited to the simulation conditions described herein. The physical parameters described in this specification include velocity, concentration, two-phase viscosity, particle diameter, temperature, initial pressure, and time settings. Without departing from the principles of this method, any fluidized bed simulation conditions can be substituted. By utilizing the simulation results to timely adjust the neural network training structure, optimal results can be achieved.

Claims

1. A bubbling fluidized bed gas-solid two-fluid simulation method comprising the following steps: S1. Use Solid Works to build a 3D model of the bubbling bed and import the model into open source software; S2. Set physical property parameters and establish a micro-scale model of particle dynamics: ε is the gas-solid two-phase concentration, ρ is the gas-solid two-phase density, v / u is the gas-solid two-phase velocity, p is the gas-solid two-phase pressure, d p is the particle diameter, τ is the shear stress; S3. Refine the simulation based on the two-fluid method, set different working conditions for simulation, verify the effectiveness of TFM simulation when the grid is smaller than the particle diameter, and establish the data connection between the coarse grid simulation and the fine grid simulation; S4. Calculate the two-phase drag force using fine grid simulation data; Re is the Reynolds number, which is related to the two-phase density, concentration and gas phase viscosity; S5. performing correlation analysis on the refined data and the two-phase drag force, determining the input and output parameters of the training, and normalizing the data; S6. Use Python, a third-party tool, to build a complete artificial neural network training model. Use Python to establish a nonlinear fitting relationship between input and output parameters through the artificial neural network and construct a drag calculation model. S7. Through secondary development, the self-built model is implanted and applied to large-scale industrial equipment to verify the effectiveness and reliability of the model.

Images