CFD for Beginners

[This was auto-generated. There may be mispellings.]

Good morning, I believe I should say, because some of you might be in the evening. The webinar will start right now. My name is Ahmed Al-Ghandour. I'm an application engineer at Allsign Engineering. I'll be providing this webinar to you. My colleague Adam is in the presentation as well.

In case I lose connection for any reason or any technical issue, he will interfere. He will be there to answer a question in case I lose connection. So I think we can start right now. We have enough people to start. We will start by giving a presentation about what is CFD for beginners.

And at the end of the presentation, we will have a quick demo on Ansys Fluent, which is one of the Ansys CFD software. All right, so let's start. So to start, we need to know first what's CFD.

CFD stands for computational fluid dynamics, which is basically the science that people use to study the fluid flow, the heat and mass transfer, chemical reaction and related phenomena.

If you need to understand the system that includes any fluids, have some heat transfer for fluid basically, or have any particles or mass transfer in the system, CFD is your tool. CFD is basically solving the equation of conservation of mass, Newton's second law and conservation of energy.

These are the equations that are used in the background to solve your system. Oops, sorry, I think I didn't share the screen. My bad. Oops. Okay, let's go up. Okay, everybody sees this, I believe everybody now. Okay, so let's go up. Okay, everybody sees this, I believe everybody now.

Okay, so let's go up. Okay, everybody sees this, I believe everybody now. Okay, everybody sees this, I believe everybody now. See the screen. Okay, sorry for that. See the screen. Okay, sorry for that. So let's start over.

Okay, so again, CFD is a computational fluid dynamics and is the science of predicting fluid flow, heat and mass transfer, chemical reaction and related phenomena. Okay, perfect. You can see now. Thank you.

So people, when they use CFDs, do you know that in the background, they're solving particular equations such as conservation of mass, Newton's Law, second law and conservation of energy? Once you solve this, you try to understand your flow.

And you need to understand what is distribution of pressure, velocity, and temperature? If your system has external flow, what is the flow organized? Essentially? This is what we call airflow, noise, and gas flow.

No, we said what you say, What is the lift and drag forces in case you're studying something like automotive or airplanes, etc.?

If you have a system with multiple phases, you want to understand where the phases will be and what is the volume fraction or the participation of each fluid or phase, regardless if it's gas and liquid or gas and solid or liquid and solid.

If you're doing chemical, you want to understand the reactions, the compositions in your system, and CFD will help you to understand all of that if you have a model that's built correctly. You might ask when in the process of any design that we can use a CFD.

CFD has any other computational system like FEA. You can use it in all the phases of engineering process. So you can use it while you're starting a new course.

So during the conceptual design and investigating the different ideas, CFD can be used to help you to understand the performance of each concept.

Once you have a design and during the development process, you still can use CFD to help you to modify your system and understand the effect of any modification you have done. It can help you in the optimization process, troubleshooting process, as well as the redesign based on the design.

So it can be used on the results, assuming you have a solid model with accurate representation. Why CFD? Because CFD will help you to do a lot of testing, you avoid experimentation, reduce the time, reduce the money and experimentation. We have a lot of experiments in the lab.

Adam, I'm going to call Evachmed. It sounds like he may have just run into some technical difficulties. I think he mentioned he may have lost his internet connection. So hopefully he'll be joining us again in just a second. I'll keep you updated on that. He'll be right back though.

Shouldn't take more than just a minute or so. So in the meantime, I know he kind of just started talking about the theory, but if you have any questions or anything, feel free to let me know in the chat. Yeah, he's saying he'll be back in just a minute.

So yeah, if you have any questions already, feel free to know in the chat. If not, we'll just hopefully wait. Hopefully he'll just be back in just a second. Okay. So I'm going to go ahead and get started. Okay. So I'm going to go ahead and get started. Okay. Looks like we have a question.

I remember an ANSYS person telling me there's a difference in the way ANSYS AIM and ANSYS Fluent solve the CFD equations. What is the difference? So ANSYS AIM is a little bit of a more lightweight kind of solver.

If your ANSYS AIM is built as a more easy to use and more comprehensive solution, you can run that. But if your ANSYS AIM is more complex, it's a little bit more challenging. But it's a little bit easier for you to get around. And it's a little bit easier to solve than some other programs.

So there's more details on that. You can find out. We can talk about that all day. But in terms of the exact equations they're solving and what the actual differences are, but generally that's kind of the difference between AIM and Fluent. Okay I'm back, sorry for that.

Thank you, Adam, for jumping in, I appreciate that. I see you were answering the question. Yep, thank you, give me one minute to go back to the slide where I stopped. Okay, so now I'm sharing my screen again.

Okay, all right, so as I was saying, you divide the volume into finite control volumes and then it will be solved over this equation to satisfy the continuity equation, momentum equation, and energy equation over the whole volume of your domain which we're going to define that in future.

Okay, so let's go back to the slide and I'll show you how to do that in a few seconds, in a few minutes. Okay, so how do we solve a problem? The first thing is we define the modeling goals. What do you want from this analysis? What's your interest? Are you looking for the temperature distribution?

Are you looking for the pressure drop? Are you interested in the mass flow rate or some other goals? So you need to define your goals. Why? Because once you define your goals, you need to define your assumptions and simplification. How far can you go with your simplification?

Do you even need a simplification? How accurate do you want your model to be? Okay, how detailed you want your model to be? How fast you want to solve your problem and get your results? All of these will help you to define how we're going to set up your model as we're going to see later.

And even before that is seeing if this is an appropriate tool for you or not. Once you define your goals, you can start setting up your model. So the first thing you do is to identify your domain.

Sometimes you have a very big model, a very big system, complicated system, but you're really interested in a very small part where you are interested in. So as you can see here, we have a system with a lot of pipes, a lot of details, but do we really need all of that? No.

We just need to define the system interested in. So I will consider this my particular domain to be used for solving my problem. Once you define your domain, you start thinking about your boundary conditions. Where will be my boundary conditions? How I'm going to model them?

And we're going to talk about that in a few slides. Also based on your domain, you can decide if you're going to simplify your model, solve it as 2D or 3D. Is it symmetric or axisymmetric? So I can benefit from that, and reduce my model, and consequently reduce my time.

So once you define your domain, you start working with it, creating a solid model that will be used to be meshed later. So how do you create your model? There's a different ways to do that. The first one is to be provided to you as a CAD model, simple and easy.

Second one is to extract the fluid region. You have a solid, and then using one of your CAD, like space claim for example, you can extract the internal fluid as you can see here or the very simplest one to build it yourself using ANSYS space claim or design model or even any external CAD model.

Once you have the model, you can simplify it, you can remove the unneeded or unnecessary features just like fillets, bolts, small details that will be hard to mesh but it has no impact in your analysis so just get rid of it to simplify your model if you know this part has no impact on your performance.

Also, you need to define if you have symmetry or periodicity. We can benefit from this if we are using CFD because it will give you the same exact result but with a smaller model and faster time of solution. Okay, once you have the model, you start working with the mesh.

Meshing CFD is a very important significant step. You have to do it very carefully and with a lot of consideration to make sure you have good results. Why? Because you need to make sure that you model all the features on the geometry that has a significant influence on your analysis.

Okay, so you need to make sure that you have fine mesh where it's needed, coarse mesh where it's not needed. You want to make sure that the mesh itself is representing everything correctly so you can capture all the small details in pressure, temperature, etc.

Also, you need to define what kind of mesh you're going to use. Are you going to use hexahedral mesh, tetra elements? Do I need to have something like this? Are you going to use a non-conformal interface?

Sometimes you need this to make it easier for you to mesh complicated geometry and sometimes you have a simple model so just make everything conformal so which is you call it also nodal connectivity. Okay, another important question which is the last one here.

Do you have sufficient computer resources? Why is that? Because sometimes some people say okay I'm gonna make everything fine mesh, everything is perfect meshing and then you have millions and millions of cells. Do you have resources to solve this or no?

That's why I need to think about your hardware while you're doing your mesh and also to make sure that you don't waste a lot of time if it's not really needed. Okay, so now we have the model, you have the mesh, now we start preparing for the solving. What are non-conformal interfaces?

I see a question is what are non-conformal interfaces? It's basically, let me go back to that, when you have two faces here, you have an interface. When you have the interface, that means the mesh on the right does not have to be compatible with the mesh on the other side.

Basically, you can have let's say 10 elements here and 7 elements here, so you don't have a node-to-node connectivity. In this case, you will have to define an interface between the two surfaces to transfer the information between the two surfaces.

If we have conformal mesh, that means every node on the right side has matching node on the left side, that's called conformal mesh. Okay, all right, so now we go back to the second slide, which is the setup of the solver.

This is where you need to use your engineering judgment and knowledge very carefully. You need to define all the needed information to represent your model from the practical to the numerical. You need to define your material, what kind of material I'm using.

What is going to be the appropriate physical model? Does my model need a laminar flow or the turbulent flow? Is there is combustion? Is there multi-phase? Do you need to define all these and then you have to get the needed information. Sometimes it has more than one. Sometimes it has all of them.

You need to understand your system to select the appropriate representation. Okay, what's going to be your boundary conditions? How many inlets do I have? And what are the information I have for each inlets? What is the interior problem or boundary condition I need to define?

Okay, another question is how to decide what our simulation has to be carried out in 2D or in 3D? Okay, whether simulation has to be carried out in 2D or in 3D. Okay, that's a good question. It depends on the system. It depends on the simplicity of your model.

Sometimes if the three-dimensional effect has no influence, it's not going to be a good thing. Sometimes if the three-dimensional effect has no influence, then you can just go with 2D.

Like, for example, this photo here, you can see the flows going basically in this plane, and I'm not interested in the out-of-plane effect.

Okay, we assume it's constant and it has no influence on the flow where this part, which is the main interest, that the third dimension has no influence on that. Then in this case, I'll just model it as 2D.

We can see here it's ready for complex problems, solving a series of problems, simplifying the 2D problem. We provide valuable experience with the models and solve settings for your problem in a short amount of time. But you cannot do that for all the models.

Especially, as I said, if you have, like, for example, you have an inlet in the X direction, inlet in the Y direction, and outlet in the Z direction. You cannot solve this in 2D. Okay, so back to the model. Again, you need, in the setup, you need to define your boundary condition.

You need to define initial values or previous solutions. Sometimes you can start from scratch. Sometimes you can start from where you stopped, using previous solution as your initial value. This sometimes saves a lot of time in the solution as well.

You need to define the solver controls, how many iterations you're going to use, what are the time steps you're going to use. All of these setups, the more accurate it is, the more accurate your model is. Everything you define affects your results accuracy and solution time.

You also need to set up convergence monitors. I'm going to talk about convergence in details later, but it's one of the most important things to define, which is a convergence monitor.

When we talk about boundary condition, it's important to understand what kind of definition I should give to the model. So, for example, if I have a model that has a boundary condition, I can define the boundary condition. The most common definition is the most robust one.

This is the most common in most of the models people use this definition of boundary condition. Another robust one is using mass flow rate. Let's say you have a flow rate of an inlet.

Instead of defining the velocity, you say I have this, for example, cubic kilometer per hour or whatever units you're using at an inlet information, and you define the static pressure at the outlet.

Another information you can define, but it's less robust and might cause troubles, is if you define pressure at the inlet, pressure at the outlet, which is not really recommended.

You can define the pressure at the inlet, pressure at the outlet, and then mass flow at the outlet or mass flow at the inlet and outflow. This is not recommended because it will have trouble solving this in the background of the software. That's why we always recommend to use a boundary condition.

So, if you have a boundary condition, you are going to use the first one, which is velocity inlet, pressure out. Okay, so now you start the solution. Now we define the set-up, the mesh, everything is ready. You want to solve. Once you want to solve, you need to look at the convergence of your model.

So, the equation you have will be solved at each iteration until you converge. So, what does it mean until you converge? Basically, you have some parameters you look at to see where the solution is going. Is it going in the right direction or not? The first thing we look at are the residuals.

You want to have the residual to go minimized over time, which is usually the standard is 1e- 4. Sometimes you will go 1e-5 or even 1e-6, it depends on your model. So you will trace the residuals, which is the equations solved in the back. You want to have the difference to be minimized.

That means it's converged. Another thing people use to check the convergence of their system is what's called imbalances, which usually depends on the mass flow rate. Also, you can look at quantities of interest.

You define the amount of interest and you can look at the amount of interest that's being used. You define what's called monitor points. You define the monitor points for a particular location, for a particular output. For example, drag or pressure drop, and see how it goes.

When it became constant, that means it's converged, because now the variable itself does not change over time. If it keeps going up and down, then you have a convergence issue. So what affects the solution to convergence or not?

There are different variables that affect that, which is a particular situation.

So, for example, if you have a situation where you have two or three different models, which is the one in blue, which is a proportion of the physical model, which is a proportion of the physical model, which is an accuracy of the physical model. Is your model represented accurately?

Did you have everything that represents the model correct? What type of assumption you are using? Are you assuming it's symmetry, but it's not symmetry? Are you assuming it's one phase, but it's multiple phase or vice versa? What kind of mesh are you using? Do you have an accurate mesh?

accurate mesh and fine enough to capture all the details. If you have a very coarse mesh that can easily lose some details especially at critical points. Otherwise, also another factor is numerical errors. How kind of numerical analysis and models you are using in your system.

The last step after you're done with your simulation, your simulation is done, you have results, is to check the results.

Which is very important to know how you do this because it will tell you if your system is solved correctly, if your system represented correctly, and also to understand the performance of your model and capture all the details that you need to understand from your CFD analysis.

So you can compare it with any experiments results. So when you look at the results, you have to visualize different stuff like some contours, some flow lines, and it will help you to understand the total performance.

It will give an idea if there is any separation, if you have any shear layers, etc. All of this you can understand from the results.

You can look at the forces in the moment, you can look at the average sheet transfer coefficient, surface and volume integrated quantities, there is a lot of tools in both CFX and Fluent you can use to examine your results that help you to understand completely the performance of your model and how your model performs.

Once you are done this, then it's up to you to decide is my model fine enough? If my model performed the way I wanted to do or do I need to do some revisions?

So you can use the tools that you have in both CFX and Fluent to determine if your model is fine enough, if my model performed the way I wanted to do or do I need to do some modifications to your model. or some modification to your model. That's a typical step a lot of people will do.

It depends on your understanding of the system. So here for example, the main question you need to answer to decide if you need to do some revision or not. For example, you decided particular model as a laminar. Do I need to repeat it with a turbulent flow or not?

If I need, what kind of turbulent model do I need to use? When you look at your system, is it stable, is it steady or it's unsteady flow? So do I need to change my boundary condition? Do I notice any compressibility effect or not? If yes, then what do I need to modify?

Is there a 3D effect while I'm using 2D model? All of these questions will help you to decide if your model is good enough or you need to revise it and do some modification. Also, you need to think about the computational domain. Is it large enough? Remember we decide this part will be my domain.

Was that good decision or need I need to extend my model? Sometimes you need to extend your model at the inlet and outlet to avoid any turbulence or any backflow at the at the inlets. The mesh itself, it's another point to judge.

When you solve a CFD model, same like FEA, you need to do some kind of mesh convergence analysis to understand if your model is mesh dependent or not. When you have a mesh and solve and then you have a final mesh, if the result change, then the mesh was not enough at the beginning.

If the change in the result is not significant, that means your model is mesh independent and that the first mesh was good enough. Sometimes you need to check the mesh quality. Do you have any orthogonality? Do you have any skewness in some location?

If you have that, then you might need to modify this because this might have an effect on the result. We're going to talk about that later. Okay, so now we have the model steps defined. We defined the domain, the solid model, we defined the mesh, the setup and the solution and the post-processing.

Now let's go in detail in some of these steps because we have to have some deeper look inside these parts. One of the most important step is meshing. When you do meshing, you have to understand how to do the mesh and how to approach your model.

But to do this, we need to understand what the purpose of our mesh. When you solve CFD, you solve equations and the equations basically solve for cells or nodal location. So each point here, the node, each triangle here, is a cell.

So how many of these you will have, how accurate this mesh represents your model will decide if you have an accurate solution or not. Okay, so when you think about meshing, you think about efficiency, accuracy and quality.

Sometimes you need to find for high resolution gradients and find geometric details and sometimes you have coarse mesh when it's not important.

So for example here, you can see this part here is significant in your model because the flow will go through thick, like wide to thin, so you might need to have a finer mesh here. But because the flow far here has no impact, so you can have coarse mesh here. So that's an important judgment.

It's up to you. And your understanding of the system, where to have a fine mesh and where to have a coarse mesh. Another thing to do is the quality of your mesh. When you do a meshing, you need to look at the quality. Do I have good elements or do I have distorted elements?

You always try to avoid any bad elements, especially at the location of interest. So you try to have the best meshing possible to avoid any troubles. Here's two judgments that can help you to decide if you have a good mesh or not.

You can always look at the matrices of your mesh and you look at the skewness and the quality. If you have skewness, you always want to have the skewness to be as minimum as possible. That means that your elements are in good shapes.

If you have a high skewness, then you have to reconsider your mesh by doing some modification at this location. When you have high skewness. The opposite is orthogonal quality.

If you have orthogonal quality issue, which means very small orthogonality, then this is a bad thing and you need to modify your mesh because you always want to aim high orthogonal quality for your elements.

When you have high low skewness and high orthogonal quality, that means your mesh has a very low skewness. High orthogonal quality means that you have high quality and that means it will be accurate in solving your problem.

If you have this problem in high level or in the location with critical values, then you might need to reconsider your mesh. If you have any question, just write it and we will answer this, either me or Adam. Here is a summary on how do you judge your mesh.

There are three factors, accuracy, efficiency and easiness to generate. Accuracy is basically, you define that based on the skewness and orthogonality. As we mentioned before, you try to avoid, you try to have the best skewness and best orthogonal quality to have the highest level of accuracy.

For efficiency, you try to minimize the number of counts. How many elements you have in your model will make your model slow or fast. But at the same time, you want to make sure it's not very high or it's not very low.

Some people will try to have very low number of counts to have a fast solution, but it might not be very good idea because it doesn't represent your model in the best way and can miss some of the details. Also easiness to generate.

When you generate automatically, you probably will have tilt dominant mesh. Some people try to say, no, I want to have hex or hybrid mesh. This will make you have to spend more time to have the mesh you are interested in unless you do just automatic mesh, which will give you more tilt element.

Do you need to do this or not? That's your call. So any engineer who's doing a CFD analysis have to think about these three points when he's doing his meshing to make sure that you have an accurate, efficient, and easy to generate mesh to save you time without affecting the quality of your model.

Okay, beside the mesh, there is another decision you have to make. First of all, I'm gonna use laminar flow or I'm gonna use turbulent flow. If you are using laminar flow, it's basically if you are saying I'm gonna use linear analysis. Very simple, very easy to solve.

But if your model is not that simple, then you need to go for turbulent flow. If you are, and this is basically you can decide yourself based on your Reynolds number. The higher your Reynolds number, the more turbulent your flow would be. So how do you do this?

This is how you calculate your Reynolds number. It's a factor that's based on the density, viscosity, and the speed of your flow and the length of your system.

When you calculate this, you can decide what is the number and then based on this, do I need to use turbulent flow or laminar flow will be enough to solve my problem.

So it's always recommended before you start your solution to calculate the Reynolds number to help you to decide what kind of flow do I need to use in my analysis. Keep in mind, laminar flow means simple and fast to solve.

Turbulent flow, more complicated but more accurate if you have turbulent flow but longer time to solve. Okay, so how we do this? You have, when you start your model, you need to define the viscosity in your system. So we do the viscous.

You click on the viscous tab here and then you have different turbulent flow. You pick which one you think is the best representation in your model. The most common turbulent flow is the SSTK omega model.

This is usually effective for most of the analysis, especially for predicting detailed, complex, and complex problems. So you can use this to calculate the velocity of the flow. When the flow is high, you can use this to calculate the velocity of the flow.

In your model, you'll see that you can test product flow and flow separation. You can also see that you can test product flow and flow separation. For the SSTK model, you can do the same thing by selecting detailed local heat transfer or flow separation on highly refined measures.

Sometimes you'll use the reliable k-epsilon model as well. Both of these are good for most of the flows. So you can read about these to understand what's the difference between them. Boundary condition. One of the most crucial things is to define your boundary condition in an accurate way.

You need to understand your system so you need to understand your input because if you give wrong or inaccurate boundary conditions, you basically solve different problems than the one you are interested in.

So you need to understand what the value you are using regarding temperature, velocity, mass flow for both the inlets and the outlets. As I said, poor defined boundary condition can have significant impact on your solution.

Other factors you need to consider when you're solving for the boundary condition is where is going to be my inlets, where is going to be my walls, where if I need to apply symmetry, which part I will use that for.

You also need to understand that everything depends on your geometry, the availability of the data, and numerical consideration. Okay. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Okay.

So to make you understand more about this issue, in the CFD, softwares like Fluent or CFX, you need to define boundary condition internally or externally. And sometimes they help you once you defined it, it will be recognized by color to make it easier for you to recognize it in your model.

External flow is basically pressure inlet and outlet, velocity inlet and outlet for incompressible flow. If you have a compressible flow, so you have the mass flow of the inlet, pressure & outlet. Other boundary condition is wall, symmetry, axis, and periodic.

Special flow boundaries is inlet-outlet vent and intake and exhaust fan. Sometimes you will have internal boundaries if you have internal fan inside your system. If you have a porous jump or interior radiator wall.

As you can see there is some stuff or some parameters can be internal and external as well. Please if you have any question at any time just write it. Okay this is a common this is a common point when you do a CFD is how close or far you need to define your boundary condition.

Remember we said that you cut your domain out of the whole system. Sometimes you need to make your inlet far enough from the domain to the boundary condition. So if you have a CFD, you need to make your system itself so it does not affect your solution.

If you look here, here we have a cut just like just very short out of the inlet and this can cause some trouble here what's called backflow condition. So sometimes you need to extend your outlet make sure that you have a stable outlet flow and then you have an accurate solution.

You can see that the representation here is much accurate than here. Just by extending your system especially at the outlet. Especially if you have turbulence conditions. So to generate to think about the boundary condition this is some guidelines that can help you to define it.

So like the inflow and outflow boundary location and shape such that the flow is either goes in or out normal to the boundaries. You always try to avoid the backflow. You try to avoid inclined flow as well.

Okay this will help you to have better conversions because a lot of conversion issues happen at the inlet and the outlet. You should not observe large gradient in the directions normal to the boundary. That means incorrect setup or you just need to move it further upstream or downstream.

Minimize grid skewness near the boundary. Error resulting from high skewness will propagate through the rest of the competition domain. So when you do the mesh make sure that your boundary condition have good skewness level so to avoid the mistake. I think we have a question here. One second.

What is backflow? Okay let me go up here. Backflow is something like this. Sometimes it's like the flow will go supposed to go out but then it start going in again. So at the outlet you have outflow and inflow. Okay or some turbulence here you can see this turbulence here.

So the arrow will go this way and come back again. So for the same outlet you have out and in. This is what we call by. Backflow. All right. Specifying well-boost boundary condition. This here is a standard numbers people usually use for defining your domain or your boundary system.

You try to if you especially if you have an external flow. This is for external flow. Like for example if you're starting an airplane or the flow on a car. So this is your car or your plane and this rectangle here this is your fluid. Okay so this is your solid. This is your flow.

So you have typically 5H of the system and 10W for the weights of the system. This is recommended to have more accurate representation to have in your boundary condition far enough so it will not be affected by any turbulence. Or any cause any backflow.

It's usually better to have bigger than this but that means you will have more you will have more computational time. So you have to think about this when you have an outflow. How far I should go to have accurate representation without increasing my computational time. So solving overview.

So how do you think about your model? It's basically if we look here you just set your solution parameter and then you enable report definition of interest. Basically what I'm going to look at in the results. Then you start initializing your solution.

Usually they start by zero unless you have some previous results. You can use it to initialize your solution and then you start solving your system. And then once you solve you start checking your system conversion. Did my system converge? Then it's good. So let's solve. Is it accurate? Yes.

Then we're done. If you have a conversion issue then you need to modify your solution parameter or your meshing or your model that you are using to make sure that you have and then you resolve to make sure you solve your conversions.

This can this is usually happen during your solution a couple of times. Especially if you are dealing with a new system or a complex system. You want to make sure that you solve the conversions.

Sometimes you solve the conversions so your model is solving fine but when you look at the result it's not accurate enough. It's not really realistic.

So in this case you need to define your system or your model by defining, checking your parameters at the boundary condition or the material property etc. to make sure you have good representation. Alright so if you look here for conversions this is how we study the conversions.

We look at the residuals or we can look at particular monitor. The residual should keep going down.

If you have a good model it will fluctuate a little bit but over time it should be going down below 1e- 4. If you have a monitor keep fluctuating at the beginning and then it should be steady state at the end. This is how you usually look at the residuals or the monitors.

Here is what we look at in the conservation is momentum or energy. This is basically the residuals. If you look for monitors it's probably like a velocity, temperature or pressure drop at particular location of interest. This is automatically defined. This is defined by the user.

Okay monitoring conversion using residual history. This is the first graph. Decrease in the residual by three orders of magnitude can be a sign of conversions. Scaled energy residuals should decrease to 10e-6 for pressure based solver.

When we talk about energy residuals that mean we have temperature included in your system. Like you have a heat transfer in your model.

When you have a heat transfer or energy residual usually go with smaller residual instead of e- 4. 1e-4 you go to 1e-5 or even 1e- 6. For scaled species residual we need to decrease to 1e-5 to achieve species balance. Okay. Best practice is to monitor quantitative variable to decide the conversion.

It's like what we said you define your own monitor to make sure that everything is converged not just the residual but as well as your parameters. So monitor other relevant key variable physical quantities for confirmation.

The report definitions are used for this purpose which could be temperature, pressure drop, velocities, etc. It's strongly recommended to use one or more report definition for all simulation. Okay I see one question. Let's see.

Can you tell me the difference between pressure and density based solvers? So this is probably depends on the system. I don't have this from the top of my mind right now. I can check it on an email back to you this. But basically and Adam please interfere if you have an answer for that.

When you have pressure base is usually defined at the outlet. Sometime when you have as I said you define an outlet pressure value and then you try to make sure this pressure is maintained correctly and sometime your analysis is based on the mass flow.

So the density variation would be the main judgment in the analysis. So typically people use typically people use the pressure base solution but sometime they might need to use a density base as well. Okay I see Adam provided an answer here.

Density based solver are better for flow with high match Mach number. Pressure based solver is better for most common problem. Okay great. Thank you Adam. I appreciate that. All right so this is basically the main presentation.

We will go to the demo now but just before we go to the demo I would like to pay your attention that if you go to our website ozlininc.com you will have more information about the training we provide as well as the webinars. At Ozlin we provide a lot of training in our company in different aspects.

We have CFD training, we have FEA training, we have particular mesh training, we have training for all the products of ANSYS. We also have an event for webinar like the one we have right now.

We try to have weekly webinars so I highly encourage you to look at these webinars because there is a lot of interest topic that might be useful for all of you. Beside the trainings we also provide the software purchase as well as consulting.

If you have any question regarding the purchase of ANSYS software, training or consulting, please feel free to email info at ozlininc.com. We also have discovery live trial. Discovery live is one of the most recent product for ANSYS where it solves problems instantaneously.

It's for linear analysis for FEA and CFD where you create, bring your model, define your boundary condition and it automatically mesh in no time and give you a solution literally live.

So I highly recommend that you look at discovery live on our website and if you have any question or interest on testing it, let us know and we will be more than happy to help you with that. With that said, we go now to a demo. So give me one second please. I need to access my machine here.

Okay perfect. Okay oops. Why I see all that? Okay, okay. Okay let me, let me re-access my machine here. One second please. Please if you have any question, please ask until I access my ANSYS machine.