How fluent simulates zero-lift angle of attack

The author came into contact with FLUENT two weeks ago. Before that, I always thought that FLUENT was just a solver. As long as the parameters are set properly, it can produce more realistic results than other small algorithms. Therefore, I focused on its English interface. I bought a tutorial online and saw a simple fluid-structure interaction analysis case (wind blowing a tree), and I thought I had mastered the skills of FLUENT instantly.

However, after two weeks of contact, I found that what I had mastered was just an ordinary model. It would be difficult if it really involved aerodynamics.

The current learning results of these two weeks are summarized:

Simplifying the model

The simplification of the model is very important and is related to the mesh division and solution time. , further affecting the accuracy of calculation. The author believes that simplification should be based on his own actual experience. For example: if a person who does not understand aircraft design gives him an airplane model, he may simplify the wing into a flat plate. Of course this is a joke. As far as the author is concerned, all the drones I have come into contact with so far have roughly the same outline of the fuselage. Therefore, I think it is necessary to simplify the fuselage, but the connection between the fuselage and the wing is still relatively small. It is important to keep it properly. However, the simplification cannot be overdone, and the windward area of ??some external accessories must be appropriately increased. Because of interference and parasitic resistance during actual flight of the drone, the author cannot guarantee that his calculations can be so accurate.

Selection of solver

At present, I don’t know much about this aspect, so I will just remember a few points.

Calculate the Reynolds number based on the object's speed and size characteristics, and determine whether the airflow flowing through it is turbulent or laminar. Air viscosity must be considered. I read in a book that because the existence of air viscosity is ignored, the calculated results differ from the actual ones by about 8%. There is also the issue of the incoming flow angle of attack. In FLUENT, the incoming flow angle of attack problem is converted into a velocity vector, which means that passive is converted into active. This is not difficult to understand. However, the resistance has been negative several times before. After Baidu, they said it was because the force vector setting was wrong. You can draw a vector triangle yourself and you will understand.

Meshing

Before, I have been using the divider that comes with workbench to divide it. As you can imagine, this is enough for simple objects, but for drones To put it bluntly, the author experienced residual value divergence due to grid problems, which caused the lift and drag coefficients to change from positive to negative as the number of iteration steps changed, and the value increased by more than 10,000 times. As you can imagine. I am currently learning icem (it is true that the more I learn, the more ignorant I feel). I hope that soon I will be able to draw grids as I like.

Solution time

Regarding the solution time, the longer it is, the more accurate it is, because everyone knows convergence and divergence, and some calculations can be used once a certain error value is reached. Just observe its convergence. Some results are not accurate as the solution time increases. This is due to computer accuracy issues.

5. Flow field volume

The author is also aware of this. At the beginning, the volume of the flow field is roughly 10 times the size of the object to be calculated. However, after reading the book, I found that the size of the flow field area also affects the accuracy of the solution. You can think about how much airflow around a 787 changes when it flies in the air. Therefore, in the case of computer NB, the bigger the better, the grid can change from dense to sparse with the size.

This week I studied icem meshing for a week, and finally I can draw a good quality boundary layer mesh for the airfoil. However, there is still a small problem, that is, the mesh mapping of the airfoil leading edge is not ideal, and importing the mesh file into fluent always fails. (It was later discovered that the reason was that the leading edge of the airfoil was not related to the line)

Icem's meshing method is indeed a bit abstract, and I have never been exposed to mapping and topology. But now I almost understand it. For three-dimensional entities, block represents the grid distribution in a certain area.

File saving:

The author does not have good saving habits when meshing. If the computer crashes, there is a certain chance that your meshing format will crash, so you must develop this habit. This will also make the next analysis faster.

Grid quality:

Good grid quality means that the analysis results of this analysis are highly credible (after all, it is a simulation analysis, so there will definitely be errors). You must be sure of yourself The divided mesh is checked for standards.

Grid number control

After two weeks of intermittent speculation, the author can finally draw the boundary layer grid of a simple wing. However, another problem was discovered, which is the distribution of grid nodes. Generally speaking, the boundary layer mesh is controlled by the E#￥%￥% function. First, determine the number of nodes on the entire target line, then determine the boundary layer growth rate, and then determine the minimum node size. Non-related positions are generally controlled by the B#￥%￥% function, which means even distribution.

When drawing the grid of an analysis model, you should first determine how many grid nodes need to be controlled. This is determined by the computer configuration. Since the author is not a wealthy person, I generally control the grid nodes to 500,000. the following. Next determine whether a boundary layer mesh is needed and how many layers are needed. Because this part of the grid is not easy to calculate. Then use addition, subtraction, multiplication and division to determine the grid node distribution.

Tetrahedral grid

The author came into contact with the drawing of tetrahedral grid not long ago, which originated from the paper on the aerodynamic analysis of UAV that I saw before. The students in it used Tetrahedral mesh, boundary layer thickness of 5 layers. After calculation today, we found that the grid quality is difficult to control and the generation speed is slow. But it can be imported directly into fluent without conversion.

About the distortion in fluent calculations with angle of attack model results

I have been trying to use fluent to solve the problem of incoming flow angle of attack during this period, but the results have always been little different from those without angle of attack. There is no credibility. The current analysis is related to these factors:

Grid quality

Number of boundary layers

The solver setting is wrong (it is it)

< p>Solution model selection for aircraft model

At present, all the problems about the grid have finally come to an end. I have been learning fluent for more than two months. Among them, I reviewed {Introduction to Air and Gas Dynamics} and read a lot of introductory books on fluent and icem. Now it is possible to draw aircraft meshes of a certain degree of difficulty, but there is another problem with the selection of the solution model.

Since the tutorials are all about aircraft aerodynamic calculations under subsonic operating conditions, they all choose pressure-steady solvers, spalart-allmaras equation turbulence models, and far-field pressure inlets. The author used the k-Epslion equation turbulence model before, with velocity inlet and pressure outlet.

Summary of the first phase

The author has been in contact with icem fluent for almost 4 months. So far, I can finally solve the aerodynamic analysis of low Reynolds number UAVs, as well as some simple post-processing. . But I always feel that it shouldn’t take so long. Now list the outline of UAV aerodynamic analysis:

Note: It is best to keep a backup after each step

Simplify the model

List reliable calculations (Wind tunnel numerical comparison is required) and is similar to it.

Verify the algorithm, including grid independence verification, Y+ calculation experiments, calculation examples of lift and drag moment coefficients, and comparison of pressure distribution at a certain section. Errors within a certain range are allowed.

Based on the above experience, conduct grid correlation, Y+ iteration attempts, divide the grid (if the geometry is simple, divide the structural grid, if it is complex, then divide the structural grid), and calculate the first layer grid size , try to control the number of grids within a certain critical value according to the hardware configuration.

Start meshing. The computational domain is generally at least 10 times the characteristic length. If it is incompressible, it can be imported slightly closer to the model.

Check the grid quality 2*2*2 standard as much as possible to be greater than 0.2, and the angle as much as possible to be greater than 18 degrees. It is necessary to smooth the mesh.

Generally, the spalart-allmaras turbulence equation is used, ideal viscous gas, and the relaxation factor is adjusted after trying it.

At present, the continuity equation is not very convergent, but if the lift and drag coefficients converge, the results can be considered to have certain credibility.

Post-processing and comparison with actual experience. Import the ansys post-processing software and extract the required parameters.

(End of the first stage)

Summary of Icem usage skills

During this period, the author had many tasks, so the models I came into contact with were only some simple outline models, but I found that it is necessary to If you want to draw a grid efficiently, you should sometimes use your brain (it feels like nonsense). Under normal circumstances, don't stick to the rules. For example, if a water pipe has a strange interior, it can be generated by mesh stretching if it is not strange. It will be much simpler to divide the strange places into blocks.

Currently, I am trying to calculate a flying wing model. I first used structural mesh to draw it. After many adjustments, when the number of meshes was kept at about 2 million, the mesh quality was greater than 0.35. It can be said to be quite good. of. However, the author found that it was difficult to converge during the solution process. After trying to solve the problem overnight, it was determined that the number of convergence steps was approximately 4,000. Analysis may be due to the following reasons:

The number of grids exceeds the computer load, resulting in a slow solution process.

This model includes the fuselage, so the flow is inherently complicated.

The volume gap between adjacent grids is too large.

The wall grid node allocation is unreasonable.

The above reasons cannot be verified for a while, so the author has tried to draw unstructured external field grids and control the number of grids (with the same number of grids, the unstructured solution is slower).