Lesson finished
Lesson finished
Lesson finished
Lesson finished
Lesson finished
Lesson finished
Lesson finished
Up next: 2D Flow Over a Cylinder: Vortex Shedding, Steady vs Transient
Unlock all 11 lessons of Introduction to GUI based CFD using ANSYS FluentFull course ₹40,000 · paid
Course Roadmap, Simulation Types and a Mesh Refinement Study
A free lesson from Skill-Lync's Introduction to GUI based CFD using ANSYS Fluent course, on Grow by Skill-Lync · Last updated 16 Nov 2026 · Session: “1.0 - Introduction to GUI based CFD using ANSYS Fluent”
What will you learn in this free Introduction to GUI based CFD using ANSYS Fluent lesson?
The instructor describes the course contents, who it suits and the weekly simulation problems, starting with a mixing tee and a cylinder. He overviews topics such as combustion and porous media, then demonstrates a finer mesh in Fluent and a grid dependence test.
- Weekly problems and physics covered in the course
- Combustion and porous media modelling ideas
- Why finer meshes slow down the solver
- Grid dependence test on outlet temperature spread
What comes after lesson 1?
The next free lesson is lesson 2, “2D Flow Over a Cylinder: Vortex Shedding, Steady vs Transient” (48 min). Later in Skill-Lync's Introduction to GUI based CFD using ANSYS Fluent course (paid): lesson 4, Exhaust Port Simulation; lesson 5, Rayleigh Taylor Instability; lesson 6, Dam Break Problem. The full course has 11 lessons (15 h 52 min of video); see the course overview or its price and syllabus.
Lesson transcript
Machine transcript of the lesson audio (lightly cleaned). Technical terms may contain errors; the video is the reference.
Part 1
Hey guys, hope you're all doing good and welcome to the course on Advanced Computational Fluid Dynamics using Ansys Fluent. In this video, I'm going to be primarily talking about what is included in the course. We'll go through the course contents in detail and then I will be explaining who can take this particular course. This is extremely important if you're considering enrolling for this course. You should definitely pay attention to this if you're considering enrolling into this particular program. Once you've talked about that, I will explain what you will be able to gain from this course.
More importantly, what is the type of simulations that you will be able to perform on your own after you finish this particular course. And then towards the end, I will basically talk about the list of available resources that you can basically make use of when you're part of this program. All right. So for those of you who have never heard of Computational Fluid Dynamics, CFD is an amazing technique which helps us simulate how fluids actually behave in real life but using a computer program. All of this is made possible by solving a set of equations which are called as governing equations using a computer.
Now, you can actually write your own code for it to do this. You know, you can use something like C or C++ or even Microsoft Excel to build something like this. But when it comes to industrial applications, majority of the engineers out there use some kind of a CFD software program. Now, this program can be a commercial software or it could be an open source software. Now, in this course, we will be putting our efforts to learn the software called as Ansys Fluent. Ansys Fluent is a very popular Computational Fluid Dynamics package that's used by a lot of companies.
All right. And Computational Fluid Dynamics can be used for a wide range of applications. It can be used for automotive applications, aerospace, energy engineering, defense, manufacturing. So any thermal system or fluid system that you can think of can be simulated using Ansys Fluent. I'll put it in such a general way. Right. Now, all of this is possible through a very fundamental approach. Computational Fluid Dynamics, as I mentioned before, is all about solving governing equations. Correct. If you solve the required governing equations, right, you will be able to capture any physics that you are interested in.
For example, say that you are trying to simulate a problem where two fluids are actually mixing. Right. And if these two fluids are of different phases, then you can basically enable an equation which would basically solve for this two-phase problem. Right. Which is what you would call as a two-phase equation. And you can basically do this by creating something called as a transport equation. Which would basically track one phase from the other phase. Right. Similarly, say that you are simulating flow that is actually in the turbulent flow regime.
Then you can simply do this by enabling the turbulence model. Similarly, if there is a problem where heat transfer is involved, then you need to use the conjugate heat transfer approach to basically say capture, conduction, convection or whatever phenomenon is happening in your particular problem. So, it turns out that Amsis Fluent can be used for capturing different types of physics. And the entire course is centered around this team because the software is so comprehensive and complex that it's going to be very hard for a beginner to understand how to set up each of these problems independently.
Right. And this is where the course is going to add a lot of value. Each and every week, we are going to teach you how to do one such application which involves different physics and how the setup works for the same and how you can actually achieve or get useful results from it. So, in the first week, we take up a very simple problem. This is called as a mixing T problem. But the idea is you are trying to see how two fluids coming in at different velocities and different temperatures mixes, right?
So, here you can see that there are three configurations. You will be simulating all the three configurations using Amsis Fluent. And you will be looking at how well these two fluids are actually mixing when it comes to the end of the pipe, right? And we will be defining mixing efficiency in terms of standard deviation, right? In this particular week, you will be learning how to set this up completely right from geometry prep all the way up to meshing, all the way up to setting up the problem in Amsis Fluent and then finally doing the post-processing, right?
So, if we go to week two, what we are going to do is we are going to shift our focus to capture flow over an external object. So, here we take up a simple cylinder and we are interested in performing steady state and transient simulation to capture the flow field around it, right? So, what we will basically do is we will run this problem at different Reynolds number and we will visualize an important concept called as vortex shedding. Then we will proceed to calculate the frequency of these vortices and the rate at which they shed and then, you know, it turns out there is good amount of experimental data available for this.
We will teach you how to perform validation, right? And validation is one of those areas that's commonly done in computational fluid dynamics because at the end of the day, it's a software tool, right? It's not going to be 100% realistic, right? Though there are certain things that you can do to fine-tune the accuracy. So, an experienced CFD engineer is one who is extremely comfortable and who is extremely confident with the results that he is getting from a particular CFD solver. And they are able to do so by performing validation in a very systematic way, right?
So, in this course, we will be teaching you how to perform validation by taking up this particular problem, right? Though the problem is simple, the steps that you would take in performing any validation study would be covered in this particular example. Then, building on the previous week, we will learn how to actually perform flow simulation over an external object, right? This object looks like a cuboid, but it has a name to it. This is what you call as the Ahamud body, right? The Ahamud body is a made-up geometry, but it's made up with a lot of thought in mind, right?
So, the dimensions of the Ahamud body is set in such a way that it represents a generic car. And the nice thing is, the people who developed this model have run a lot of experiments in a wind tunnel to measure a lot of data points, which you can actually use as part of your CFD validation. Now, in this particular case, we are not going to be doing a validation study, but you will be understanding how to set up an external flow simulation. And then, you will basically learn how to optimize the mesh for an external flow simulation.
Now, the experimental data for this particular model is available online. So, if you are interested, you can take it up one step further and perform the validation study yourself. Then, moving on to week number four, you will be able to simulate conjugate heat transfer on an exhaust port. So, here we basically, you know, have an exhaust port model. This is very important for us, because the shared topology feature helps us create mesh interface between the fluid and the solid. Because, as you know, conjugate heat transfer is said to take place in a system where there are more than two modes of heat transfer.
So, in the exhaust pipe, what's happening? You have, you know, hot fluid that's actually flowing in, right? So, if you assume that this is your engine, your hot exhaust gases are flowing in through various ports, and they all basically meet here, and they went out from this external outlet, correct? So, here, as the hot exhaust gases flow through this pipe, they are going to heat up the walls of the pipe, and it's going to result in increasing temperature of the external solid, right? So, you can see that there is conduction through the solid walls of the pipe, and there is convective heat transfer that's taking place at the interface of the hot exhaust gas and the solid, right?
Since there are two modes of heat transfer involved, this is what you would call as a conjugate heat transfer problem. And in this case, we want to make sure that the mesh interfaces are set up in such a way that the heat transfer information is transferred from the fluid side to the solid side without any additional interpolation. In fact, this type of consistent meshing is what you would call as conformal meshing, and that's extremely important if you want to capture heat transfer in an accurate way.
Moving on, we set up another problem, but this time we take a problem from the electronics industry. So, what you are looking at here is the model of a graphics card, right? Now, this is obviously a made-up graphics card, but it captures all the details of one. So, in this particular problem, we will basically set up a processor, which is basically shown in pink here, and we will provide a volumetric heat generation source to it. And then, we will basically put this entire graphics card in a domain, and we will have air flowing through it, right?
Just to kind of simulate the effect of a fan blowing air over the graphics card. And we will then look at, you know, hotspots, we will look at Nusselt number, and so on, all right? So, then moving on to week 6, we will be looking at a very, very interesting problem. This is what you call as the Rayleigh-Taylor instability analysis. To keep it simple, this is a two-phase simulation, right? So, here you have two fluids, right? One is colored in red, and the other one is colored in blue.
Now, if the density of these two fluids are different, right? If the red fluid is denser than the blue one, then it creates some disturbances to the interface, and the interface becomes unstable, and you have this really cool-looking patterns, which is what you call as the Rayleigh-Taylor waves. Rayleigh-Taylor waves have a lot of applications with respect to Lagrangian spray modeling, and are used for a lot of theoretical calculations as well, right? So, here we will basically be using the volume of fluid method to set this up, and you will be learning the importance of Atwood numbers, and in order to kind of, you know, set the alpha values, which is called as the volume fraction, right?
You will be learning something called as fluid patching. Don't worry if you don't understand what alpha value is. To keep it very simple, you know, red is basically the heavier fluid, and blue is the lighter fluid, right? Alpha essentially tells which is which. You can imagine that alpha equal to 1 is the red fluid, and alpha equal to 0 is the blue fluid, or the other way, right? But if you're tracking where the information changes from 0 to 0, That's the location of your interface. So, in this particular problem, we will enable the volume of fluid method, which in turn solves a transport equation for alpha, which is called as the void fraction.
And that is what helps you in getting this result. All right, moving on to week 7, we take a bit more interesting example. In this case, we look at a gate valve and its parametric, which means you can vary the lift and you can basically, you know, look at the pressure loss that occurs at different valve opening times. This is a very common exercise that is done in valve industry. And, you know, working on problems like this is extremely important for engineering students because it helps you get a good idea about how CFD is used in companies.
Week 8, we look at another industry application, which in this case is cyclone separator. A cyclone separator is usually a filtration device that is used in a lot of industries. So, here the concept is very simple. If you have a vacuum cleaner, right, it basically has a cyclone separator in it. So, essentially what happens is here you have the heavier particles coming in, right? And what happens is the heavier particles tend to rotate and they settle down, whereas the lighter particles, they get bent out, right?
So, there is a pressure difference that's applied at this side and this side. So, usually the pressure at top is lower than the pressure at the inlet, right? So, you can kind of imagine that at the outlet you are trying to suck in the particles, right? Now, the problem is only the lighter particles would be sucked in. Now, if the lighter particles represent the impurities that you are going to be removing, then this works very well, right? In the other way, if heavier particles is what you want to separate, then that's great.
The heavier impurities would get settled down at the bottom of the cyclone separator, which helps you in separating your heavier particles and lighter particles. Now, from a simulation part of view, you need to learn a technique called as discrete phase modeling. Now, in general, with respect to Navier-Stokes equations, when it is basically solved on a discretized domain or a mesh, within each and every mesh point, within each and every mesh volume, within each and every cell, you have a particular value for temperature, pressure or any other transported quantity, right?
In a discrete phase, what you basically do is you inject particles which are tracked, right? So, between two particles, you can have a huge space, right? And there can be discontinuity with respect to the particle distribution itself. You can have particles in one location, but you might not be able to find it in another location. DPM is super important because these discrete phase particles or entities interact with your mesh or interact with what you would call as the Eulerian phase, the continuous phase, right? And this is super important in calculating the efficiency of the cyclone separator.
So, in this particular week, you will be performing a study where you will look at the effect of particle size and injection velocity on separation efficiency. So, week 9, we tackle a bit more challenging problem. So, this is where we introduce you to UDF, right? Again, remember that this is a gentle introduction to UDFs. It's not a detailed presentation on various user-defined functions that ANSYS has to offer. Like I said, UDF stands for user-defined function. So, this is where ANSYS gives you, the user, an opportunity to interact with the back-end code, right?
So, in this particular case, we have a made-up gearbox, right? You have these two gears that are going to spin around. And what we are going to do is we are going to initialize some amount of liquid in it, right? And as the gears are going to spin, they are going to splash around that liquid, right? This is commonly done for gearbox or transmission simulations in the automotive industry to make sure that the system is being lubricated in a sufficient way. Now, here, in order to spin the rotors at a particular speed, we basically will write a UDF to provide the boundary motion for these gears, all right?
And you will be looking at some very intro-level settings of dynamic machine, and you will be understanding the importance of relaxation factors with respect to convergence. Now, remember that this geometry is fairly simple, and we will not be covering the detailed aspects of user-defined functions that are available in AMSYS Fluent. Week 10, we will basically introduce you to combustion simulation, right? So, this is where we will teach you the basics of chemical kinetics and combustion modeling. And we will be taking a very simple system of methane-air system, and we will be performing combustion analysis on it.
And I will also teach you how to basically get the mass fraction plot, right, along the transverse and longitudinal axis of the combustor, right? This is what you call as axis-symmetric geometry. You know, it looks like a rectangle, right? But actually, the geometry would be cylindrical in nature. This would be the axis of the geometry. So, if you rotate the rectangle about the axis of the geometry, you would get a cylinder, right, which is basically the combustion chamber, all right? So, this is essentially what is covered in the course.
The next two weeks of the course would be spent in your projects, right? So, each and every week, you know, you're working on assignments, and during week 6 and, you know, week 10, you know, you have your major projects. And the course essentially continues for another two weeks so that you get enough time to finish all of these 10 modules. All right. So, who can take this course? If you're an undergraduate student, say, pursuing mechanical, aerospace, automotive engineering, or manufacturing, or even electrical engineering, and you're really passionate about computational fluid dynamics, and you're looking for a starting point, this course is going to be extremely great for you.
It's going to be very comprehensive. It's going to be very hands-on so that you're not overwhelmed by the theory of computational fluid dynamics, and you can start seeing simulation results, which will hopefully give you the motivation to dig deep and learn about the back-end of CFD solvers, right? Now, if you're a graduate student who is venturing into the CFD domain, meaning you have no prior experience in CFD, then this course, again, adds a lot of value to you. Now, professionals without experience in CFD, it's a great course.
It's highly industry-relevant, and, you know, it's to the point. So, you will find a lot of value to it, and you can basically take the course on your own time and work on the different applications on your own time. Now, if you're a working professional with experience in CFD, the course will be useful, right? But the thing is, you will find the content to be very basic. Now, let me give an example. Say that you are a graduate student in some lab working on ball valves, or say that you are an engineer working in some valve manufacturing company, and you have been using CFD for a while for that particular application.
And you want to learn about additional simulation methodologies such as volume of fluid, multiphase, conjugate heat transfer, combustion, and so on. Then, this course is going to be super useful for you. But if you are a professional in CFD, and you are already aware of different simulation techniques, then this course might not be that useful for you, right? Please make sure that you read through this course syllabus before you enroll into that course, if that is the case. Similarly, we have designed this course keeping undergraduate student in mind.
So, in other words, this course is at an advanced level for a typical undergraduate student. So, make sure that you go through the course syllabus before you purchase this course. Now, what will you gain from this course? And a thing that we can definitely deliver is give you the ability to perform a wide array of simulations with Amsys Fluent with confidence. Any new simulation that's thrown at you, you will be able to take logical steps towards it, and you will be able to, you know, come up with a decent baseline simulation.
Now, if you are working in a company, getting a baseline simulation is always the first step. And you can always make your simulation accurate with the availability of more information about the system that you are setting up. The second thing that you will gain is the complete understanding of the pre-processing to post-processing pipeline. This is something that a lot of people struggle with because they cannot visualize the entire flow. And for an undergraduate student to gain this skill set is going to be extremely valuable when they are later going to be looking for jobs.
The final thing that the course definitely helps you with, which is very important at an undergraduate level, is to develop the ability to perform validation studies. Like I said before, Computational Fluid Dynamics is a tool. It's going to simulate reality, but the accuracy is always something that will be questioned, right? And the only way you can be sure about the model choices, the mesh choices, and everything else is by performing one or two validation studies, right? This course teaches you how to do that and what are the different measurement tools that you can use in Ansys Fluent so that you can measure data from simulations and compare it against the experimental data that might be available.
Now, in terms of available resources, this course comes with comprehensive technical support, right? So please do make use of this. Now, remember, this course is designed for an undergraduate student. So a lot of dense theory is not covered in this particular course. However, we have detailed articles on the same topic. So if you are a student who wants to learn more about, say, discretization techniques, then do let the support engineering team know and they will basically give you an article about it and they would be happy to provide you more details regarding that over an online session or over a telephonic call, right?
The reason why we are not publishing that information to all students is because the intended audience for this particular course is undergraduate students who might have not had taken a formal course on the theory of CFD and also advanced theories such as turbulence modeling or combustion modeling. Alright, with that, I would like to thank you so much for taking the time to watch this video. Do let us know if you have any questions. Thanks. Bye.
Part 2
Hey guys, hope you're all doing good and welcome to a new video. So today we are going to be discussing the fundamentals or one-on-ones of computational fluid dynamics. Anytime you are trying to learn a simulation tool, it could be SOLIDWORKS flow simulation, CONVERT CFD, Ansys Fluent and whatnot, learning the fundamentals of computational fluid dynamics is extremely important so that you are no longer tool dependent. So today's presentation is going to give you a very good overview of this entire space. We are going to start by discussing what the Navier-Stokes equations are and then we will jump into something called as discretization, also called as mesh building process.
Then we will talk about a major classification of computational fluid dynamics simulations called as the steady state and transient simulations. In general, all the simulations that you typically do with a CFD solver can be categorized into these two fields. Once we have discussed that, we will talk about the general steps in computational fluid dynamics and I will be giving examples along the way. After that, we will talk about the importance of boundary and initial conditions. Once we have discussed that, we will talk about physical models.
Understanding the physical models and their background equations is probably the most important part of being a CFD analyst. One of the reasons why CFD helps us perform multiple variations of simulations is because you can always add a new equation to capture some new physics. And hence, understanding the fundamentals of physical modeling is extremely important. So here we will be going over some very important models. For example, turbulence, conjugate heat transfer volume of fluid method, computational combustion, which is basically chemistry of reacting flows, and then porous media.
Once we have discussed physical models, we will talk about linear system solvers. At the end of the day, your computer is solving a bunch of linear equations that represent the original PDE. For solving linear equations, there are multiple methods. We will discuss the most common categories, which are direct solvers and iterative solvers. And in this, there are multiple methods. So we will be giving you an overview. Now, remember, this lecture is just for someone who is beginning with computational fluid dynamics. And there might be a lot of things in this presentation that might not be super clear.
So as far as a learning path is concerned, if you are a student who is interested in learning a particular tool, this presentation is all you need. Whereas if you are a student who kind of wants to understand the details of each and every step, then just this presentation might not be enough. You might have to take up one to two beginner level textbooks in computational fluid dynamics for further understanding. Or you can also take a look at some of our existing courses on this topic.
All right. So let's start with the Navier-Stokes equations. Navier-Stokes equation basically describes motion of viscous fluids. Now, an interesting thing here is all these mathematical equations basically represent conservation laws. In your undergrad, I'm pretty sure you would have heard the terms conservation of mass, momentum and energy, right? Now, these are written in the English language. Now, when these equations are converted into their mathematical form, they are usually coming into either a partial differential form or an integral form. Historically, the momentum equation, which basically says that F is equal to MA, was called as the Navier-Stokes equations.
But as time passed and as computational fluid dynamics developed a lot, it's very common among engineers these days to refer the continuity, momentum and energy equations together as Navier-Stokes equations. Though that is kind of what is going on today, originally only the momentum equation was referred as Navier-Stokes equation. Or in fact, that was the original Navier-Stokes equations. But that being said, if you take a look at the mass, momentum and energy equations together, it basically describes the motion of fluids and primarily viscous fluids. In other words, the effect of viscosity is taken into account.
If you are not taking account the effect of viscosity, then the equations are called as Euler equations. Now, if this is a bit too complex for you, just remember that the Navier-Stokes equation is nothing but Newton's second law applied for fluids. At the end of the day, you get a set of coupled partial differential equations that can predict the fluid behavior. Now, this is an image from NASA's web page. So credit goes to them for making this image. So here we have listed the XYZ momentum equations.
As you can see, these are partial differential equations because of the presence of the terms dou by dou T, right? Now, in addition to this, these equations are coupled. What does that mean? Well, if you look at the first equation, you can see that there is rho, which is density, U, which is the X component of velocity. Now, the same two variables are found in the other equations as well, right? You can see that rho and U appear in multiple locations in all the equations. Similarly, if you look at V and W, which are the other components of velocity, they appear in other equations as well.
So this means you cannot solve these equations one after the other. You can still do that. However, the solution would be approximate. In order to solve the system perfectly, you will have to solve them together. And that is why these are called as coupled equations. Now, if you are someone who is already familiar with Navier-Stokes equations, remember that this is written in a non-dimensional form. And that is why you get Reynolds number in the right hand side. And also the body force terms have not been included in this equation.
Now, let's talk about discretization. Your computer cannot solve partial differential equations. So if I go to a computer and write a C++ program, I cannot say that, hey, C++, solve this program for me. A computer at this point cannot do this, right? So you will have to take these partial differential equations and you will have to convert it into a form that your computer can understand. And this process is what you call as discretization. Now, remember, these partial differential equations are continuous functions, right? These are continuous functions, meaning at any given XYZ location, you can compute density, velocity and temperature if you solve these equations correctly.
But since that is not going to be possible, as there is no analytical solution for these equations, we will have to solve these equations in discrete points or discrete shapes or discrete volumes. Depending on the methodology that you choose, you can actually get three types of discretization techniques, which are called as finite volume, where you are discretizing the entire domain into volumes. Or finite difference, where you are discretizing the entire space into points and finally finite elements, where you are splitting the entire domain into 2D elements.
So at the end of the day, discretization techniques helps us solve partial differential equations. This is one of the most important things that you need to learn and is a common interview questions. Many times when I talk to students, they tell that finite element method is for structural applications, finite volume is for CFD applications, and most of them don't know where finite difference is used. In general, you can use finite volume, finite difference, finite element to solve any type of partial differential equation. So you can use finite element method for solving a structural analysis problem, you can use finite elements for solving a CFD problem, and you can use finite difference for solving both of them, right?
But in reality, certain methods have certain advantages. You will typically see that finite volume methods are used for computational fluid dynamics, because it helps with something called as conservation. Since you are writing governing equations for a volume, it is inherently more conservative. Finite differences are used whenever you are solving time-dependent problems or hyperbolic problems. It is very common to see finite difference being used in conjunction with finite element or finite volume to solve partial differential equations. Finite element method, again, is very efficient for structural analysis problems, right?
So that being said, I would like to reiterate, you can use any type of discretization technique to solve any type of PDEs, but historically it has been observed that certain types of discretization techniques are working very well for certain types of problems. So let me give you an example for finite volume method. So for example, here, you basically have a diffuser type geometry. Now, each of these blocks is a volume, correct? Now, in finite volume method, we write the governing equations for this entire block, correct?
Whereas, in finite difference method, you write the equations for this particular point. Now, in finite differences, the problem that you usually encounter is computing derivatives. Now, say that you are in this red point and assume that this direction is the x-axis. So if you want to calculate the derivative of temperature at this point, you would need the temperature in this particular node, and you would need the temperature in this particular node. Unfortunately, there is no physical node in this particular location. So the best you can do is interpolate from, say, this particular node and this particular node.
And that is not very efficient and it gives you error. Whereas in finite volume method, you are writing the governing equations for each and every volume. And what you basically do is you receive information from the neighboring volumes, right? And since the volume is always connected to the neighboring volumes, except for the boundaries, this methodology works very well for complex geometries. And since finite volumes are primarily used for computational fluid dynamic purposes, we will not be discussing finite element method anymore. All right. So let's jump on to the next topic.
Here, we are going to be talking about steady state and transient simulations. So a steady state simulation is one where the characteristics of flow field do not change with time, right? And these type of simulations are said to have a larger time frame to capture results when steady state has been achieved. Now, we perform steady state analysis when our interest lies only on the end results rather than the path that's taken to achieve the result. Now, this might be very confusing, but let me explain. Say that you are simulating flow through a pipe.
We know that we are expecting a parabolic profile at the end of it, correct? So you really don't care about what your initial solution is and you don't care about the intermediary solutions. So you're going to be able to see that you are not going to be able to see that you are not going to be able to see that you are not going to be able to see that you are not going to be able to see that you are not going to be able to see that you are not going to be able to see that you are not going to be able to see that you are not going to be able to see that you are not going to be able to see that you are not going to be able to see that you are not going to be able to see that you are not going to be able to see that you are not going to be able to see that you are not going to see that you are not going to see that you are not going to see that you are not going to see that you are not going to see that you are not going to see that you are not going to see that you are not going to see that you are not going to see that you are not going to see that you are not going to see that you are not going The only thing that you care about is your final results and that is why this is called as a steady state simulation.
Now there are simulations where the solution in one time depends on the solution of the previous time step. This is what you call as a transient simulation. A great example for this is simulation of flow through an internal combustion engine. Now think about it. Say that your final goal in this entire process is to get the emission data. Now if you have to get emissions correctly, that means you will have to simulate combustion correctly. If you want to simulate combustion correctly, that means you need to capture fuel vaporization correctly.
If you want to capture fuel vaporization correctly, you need to model the spray breakup correctly and also turbulence correctly. Now in order to model both of this correctly, you need to get the right amount of airflow rate. In other words, if you assume that your engine is a four-stroke engine, you need to simulate all the four-strokes perfectly. And this is a great example of a transient simulation. Now another type of transient simulation that commonly occurs is what you call as pseudo-transient simulations. So a great example for this is flow over a cylinder.
Now it has been found out that when flow goes over a cylinder and reaches a Reynolds number of around 100, you start getting something called as vortex shedding. So that is basically separation zone that's created behind the cylinder, which keeps on fluctuating. Now the flow field is continuously changing with time, correct? So one might say that this is a transient simulation, but one can also argue that the frequency of these fluctuations reaches a steady state. Meaning after some time, the rate at which the separation zone flaps becomes a constant.
And that is typically the output, right? However, since you are dealing with frequency, you need to capture things as a function of time accurately, which means you need to perform a transient simulation. Now all the simulations that you might perform using a computational fluid dynamics package falls either into steady state or transient. Now as a general rule, transient simulations are the superset. Meaning any steady state simulation can technically be run with a transient solver, but any transient case cannot be simulated using a steady state solver.
In other words, if you have to just pick a single solver to solve all computational fluid dynamics problems, then theoretically you would be choosing a transient solver. Hope that makes sense. All right. So now let's discuss the common steps in computational fluid dynamics. Pre-processing, post-processing and solving. Pre-processing is fairly easy to understand. And this is where you prepare your geometry for the required simulation. Now, depending on the geometry in which you are performing the simulation, you know, you might have to clean up the geometry, extract the fluid volume, or, you know, split the geometry into multiple pieces.
This is what you call as pre-processing. Once pre-processing is done, you will have to discretize. And typically this involves taking that entire 3D space and cutting it down into tiny volumes. And this is done using some type of a meshing software. And this is where you would also create named selections so that you can assign initial and boundary conditions. Now, typically softwares like ANSYS, SpaceClaim, and Workbench are used in this step. So here you can actually see an exhaust port. They are actually using the fluid volume extraction utility in ANSYS to get the inner volume.
And here you can see the computational mesh as well. Right. The final step involved in any CFD solver is called a solving. Now, in order to solve the system of equations that we get, we need to provide the right initial boundary conditions, material properties, and physical models. An example of that in this particular case would be the turbulence model. Finally, once the simulation is done, you can analyze the results, which is the most important part of any analysis. Right. Apart from getting colorful images, you need to make sure that your simulation provides the right data.
So this has to be both qualitative and quantitative. For example, here we are showing you results from a combustion simulation. Right. So qualitatively, this looks like a flame. Right. There is high temperature zone in the middle. And as you move away from the flame center, you have low temperature, which is very typical of a combustion simulation. But this is only qualitative, meaning just from visual inspection, you can see that the simulation is correct. But that is not enough for engineering. For engineering, you should also do quantitative analysis where you will have to measure data from the simulation and compare it with, say, experimental data or some other data source that you trust.
So here, for example, in a combustion simulation, what is typically done is engineers or analysts, they create line probes along the transverse direction and they measure the temperature along this line. They measure mass or mole fractions of species like hydroxyl ions, carbon dioxide, the fuel, carbon monoxide and so on. And they compare that result with experimental data or any other data source that they trust. And if the data matches within a particular margin of error, they trust the simulation. Right. So post-processing is very important. Oftentimes, students feel like post-processing is only about creating colorful images, but it's not.
You always need to question the simulation results because when you perform numerical discretization, you are making an approximation to the solution. Right. So you always need to make sure that you check your results. All right. So let's talk about boundary conditions. You would have heard of boundary conditions, you know, while solving simple differential equations in your second or third semester mathematics. You might have seen this in CFD, finite element analysis, multi-body dynamics or any other scenario where an ordinary differential equation or a PDE is solved.
To keep this simple, boundary conditions help us tell the simulation what are the answers at the boundaries. For example, say that you are solving a differential equation and you are solving for fluid velocity. You would provide the values for fluid velocity at the boundaries and that provided value is going to help anchor the solution. And that boundary value is going to govern the solution everywhere else within your computational space. So in other words, it's very important to provide boundary conditions correctly and you cannot just guess them.
So typically in any type of engineering simulation, boundary conditions are obtained either through simple back of the envelope calculations or most commonly they are obtained from experimental data. So let's discuss some popular choices of boundary condition. If you are simulating flow through any type of a device, as long as there are inlets and outlets, this set of boundary conditions should work. You would use a velocity inlet and you would use a pressure outlet or you would use pressure inlet and pressure outlet. Or you would basically provide mass flow rate at the inlet and pressure at the outlet.
Now, the reason why these are popular choices. Now, the reason why these choices are popular is because they help you get a converged solution. That is one thing. And secondly, it's very easy to measure pressure and velocity information. Now, in addition to this, you have other type of boundary conditions. For example, symmetry boundary conditions and moving boundary conditions. Now, these conditions are really problem specific. And if you do not understand what these mean, it's completely fine. Usually, symmetry boundary conditions are used to simplify the problem.
For example, if you are simulating flow over a car, remember that the car is symmetrical, meaning say that you cut the car exactly in the middle. From the outside, the left side of the car and the right side of the car are going to look exactly the same. And this is the concept of symmetry. Since both sides are identical, you can just pick one side and run the simulation on that side. And then you can take the data and you can mirror that information. Correct. This is the essence of symmetry boundary conditions.
Now, there is something called as axis symmetrical boundary condition where you know that the geometry and the flow field are symmetrical about a particular axis. So, any derivatives along the radial direction can be ignored. Moving boundary conditions are again fairly simple. A great example for this is your internal combustion engine. In your internal combustion engine, your pistons and valves are actually moving, right? Now, as they move, they are going to push the fluid that is attached to them. So, you need to provide the velocity with which these boundaries move.
That is what you would call as moving boundary conditions. As far as classes of boundary conditions, you might hear the term Dirichlet, Neumann and Mixed Boundary Condition or Robin Boundary Condition. Dirichlet and Neumann are the most commonly used boundary conditions. It is very simple. In Dirichlet Boundary Condition, you are directly specifying the value of some property at the boundary. It could be velocity, pressure, species information or temperature. In Neumann Boundary Condition, you are specifying the derivative. So, let us discuss a very common Neumann Boundary Condition.
Say that you are running a heat transfer simulation, okay? And there is a wall that is completely insulated. So, which means there is no heat transfer across that wall, correct? So, if there is no heat transfer, that means there is not going to be any gradients in the temperature across the wall. So, if the wall is along the x-axis, then dT by dx or rate of change of temperature with respect to x is going to be equal to 0. This is what you call as the Neumann Boundary Condition or the 0 gradient boundary condition.
We call it 0 gradient because the gradient of temperature is 0. Now, remember for each equation that you solve, suitable boundary conditions are required. For example, in your simulation, you have equations for turbulence. You need boundary conditions for the turbulence parameters. And if you are solving for species diffusion, then you need boundary conditions for the species. So, in general, any equation that you solve in computational fluid dynamics, in addition to mass momentum and Navier-Stokes equations, can be called as scalar transport equations. So, the scalar can be energy, it could be species or one of the turbulence parameters.
Now, why can't you pick any BC? I hope at this point you can kind of figure out why this is. But basically, boundary condition is the reason why we are expecting a unique solution for our problem, right? So, at the end of the day, think about it. You are solving the same set of Navier-Stokes equations. And there are only two things that are different. One is the geometry and the other is boundary condition. If you change any of them, you are totally simulating a different problem.
You cannot expect correct results. Now, most softwares offer protection against picking boundary condition, but the only thing that they do is they help you prevent providing wrong choices of boundary condition, but they will not protect you if you enter the boundary condition incorrectly. A great example for this is say that you're simulating flow through a pipe. Softwares like Ansys Fluent will tell you that provide the inlet velocity, right? But that's what it can do. It cannot tell you if the velocity should be 5 or 10 meter per second because that is a user input. Similarly, there are something called as initial conditions.
Now, I like to say that initial conditions are nothing but boundary conditions for time, meaning when you're starting the simulation at each and every volume, you know, in the context of finite volume needs to have values for temperature, pressure, and so on, correct? This is what initial conditions are. Now, let's talk about something extremely important, physical models. Now, in computational fluid dynamics, in order to get more information from the simulation, you typically have to solve more equations. If you want to capture turbulence, you need to have equations for turbulence.
If you want to capture, say, combustion, you need to have special equations that can solve for combustion. Now, these are what you call as physical models, okay? So, if you want to run a simulation, so you cannot expect that the same set of equations will give you all the information that you need. In general, depending upon the type of simulation that you have, you might have to pick and choose additional equations. Now, this is where a lot of challenge comes. Usually, I see that students have a particular simulation idea in their mind, but they are unable to break it down into different physics.
If this is a problem that you are facing, that typically means that you are not understanding the physics of the actual problem. Now, if you take turbulence models, there are a lot of classifications. So, two classifications that are really common are RANS, which are in Reynolds' average Nabea-Stokes, and LES, large eddy simulation. And in RANS, we have a couple of famous categories, which are called as K-omega turbulence model and K-epsilon turbulence model. Now, as of today, majority of your engineering simulations are RANS-based. And in general, it is accepted that RANS-based simulations are less accurate than LES-based simulations, and LES-based simulations are less accurate than direct numerical simulations.
Direct numerical simulation, or DNS, is the highest quality simulation that you can ever achieve, but 99% of the engineering problems are not solved using DNS. They are solved mainly through RANS and then LES. That is because in terms of speed, RANS-based simulations are much quicker. Now, if you are interested in learning more about turbulence model, you ideally need to take a course on turbulence modeling. The second type of physical modeling is called as conjugate heat transfer. CHT refers to simulations where there are multiple modes of heat transfer.
A great example for this is an heat exchanger. In an heat exchanger, you actually have both conduction and convection modes of heat transfer, right? And if you are interested in capturing this, you need to perform conjugate heat transfer. In conjugate heat transfer, there is no limit on the number of heat transfer mechanisms. For example, you can also model radiation to the same problem to increase the accuracy. Now, here is a great example for that. It is the same exhaust port. Now, in this exhaust port, what we are doing is we are coloring the external surface by the temperature.
Now, remember there is also fluid flowing inside, which also has its own temperature field. The next type of physical modeling that we are going to be talking about is for multi-phase simulations. That is, you have more than one phase, right? Example for this is boiling type simulations where you have vapor inside water or cavitation-based simulations. So, in this particular technique, you have two immiscible fluids, meaning fluids that don't mix with each other. And the idea here is very simple. So, this is your finite volume, right?
You basically have a chessboard type layout, and you are solving for something called as void fraction. So, wherever void fraction is zero, that typically means that you have some type of a gas phase. Wherever the void fraction is one, you have liquid. So, at the interface, you will be switching from liquid to gas, right? So, there, the values will be between zero and one. And this type of simulations are called as volume of fluid method. Now, VOF methods can be used for performing a lot of cool simulations.
For example, gearbox animation, you know, like the ones that we are showing here. In addition to this, you can also, you know, perform simulations of solidification and melting. However, in order to do that, you need to basically use VOF plus some type of porous modeling approach as well. Now, there are several types of reacting flow simulations that are possible, but here are some popular categories. First is surface reactions. If you are interested in capturing reactions at surfaces, as the fluid is flowing over them, then that is what you call as surface reactions or surface chemistry problems.
An example for this would be your flow through a catalytic converter. Now, as the emissions are flowing through the catalytic converter, reactions are taking place on the surface to convert pollutants into neutral products. Similarly, your simulations are classified based on how the fuel and air are mixed. For example, say that you have an engine that has a carburetor, then you would have pre-mixed simulations. If you have a diesel engine where fuel is injected just before combustion takes place, that's what you call as non-premixed. And anything in between is called as partially pre-mixed.
Now, in terms of combustion modeling, there are primary classifications, which is called as finite rate chemistry and eddy dissipation models or turbulent chemistry interaction models. So, finite rate chemistry is where you are assembling a system of governing equations, primarily ordinary differential equations, to capture how each and every species evolves over time. This is probably the most accurate way of doing combustion simulations, but it's extremely time-consuming. A commonly used alternative is called as turbulent chemistry interaction models, where you use turbulence to kind of estimate the time scales associated with combustion.
An example for this is eddy dissipation models. In general, computational combustion is a hardcore subject and is hard to understand. If you are interested in learning more about this, ideally, you should read one to two fundamental combustion book or consider taking one of our existing courses. So, here is an example of a combustion simulation of a methane-air mixture. So, here you are looking at the volume rendering of temperature. And these type of simulations are actually done on burners and gas turbines. And the end goal is fairly simple.
You want to look at the temperature uniformity at the outlet. And you would also perform a conjugate heat transfer simulation to make sure that all mechanical parts are maintained at a safe temperature. Porous media modeling is extremely important because without it, you cannot do certain types of simulations. For example, assume that you have an air filter and you want to simulate the flow through the air filter. Now, these type of simulations are technically not possible because the pore sizes are in the order of micrometers, whereas the lens scale of a filter is on the order of 10 centimeters or more.
Correct? Now, because of this discrepancy in lens scale, your simulation is going to be running forever. And because of this, we have something called as porous media modeling. The idea is very simple. You are not interested in how the flow takes place inside the porous media. You are just interested in capturing the pressure drop that occurs across a porous media. If you think about it, that's exactly what a porous media does. It creates pressure drop. Now, if this is something that's hard to imagine, I can give you a very simple exercise that you can do at home.
So, try inhaling as much air as possible and try to blow them out of your mouth immediately. Right? You can easily do this without any resistance. Now, what I want you to do is take a thick blanket or sponge and try to do the same thing. When you're trying to blow out, you get a lot of back pressure. That's because there's a lot of pressure drop across either the thick blanket or sponge, which basically makes you do more work in order to achieve the same flow rate.
And this is exactly what is required by engineering companies. If you are putting a new air filter, they want to know what's going to be the pressure drop across it. The tricky thing while setting up simulations with porous media is in understanding the coefficients. Typically, porous media models rely upon a lot of correlations and constants which depend upon the material, which is basically the filter that you use. Usually, this data is available from the filter manufacturer. But if you're an independent student, then it's very hard for you to set up these coefficients to get the right pressure drop.
At a bare minimum, you need to know the relationship between velocity and pressure drop. All right. So the final thing that I want to talk about is linear system solver. At the end of the day, discretization basically yields in a set of linear equations that you need to solve. Now, here you typically have two methods, direct methods and iterative solvers. Direct methods, you might recall from undergrad education, Gauss elimination, matrix inverse and Gauss-Jordan. Now, these methods are not used for solving large CFD problems. They can be used for really, really small problems.
But for any type of CFD, FE or MBD simulation, you generally use iterative methods. Because the direct methods do not scale very well as you increase the number of equations. As far as iterative methods, you know, some popular methods include Jacobi, SOR, Gauss-Seidel and conjugate gradient methods and its variants. Now, to keep it simple, here is how an iterative method works. You have a linear equation or a system of linear equation. And what you do is you take a guess value. Now, based on this guess value, you get a residual.
And from that residual or error, you compute a better guess. You keep on doing this process. And at the end, you get a converged solution, meaning you get a result for your linear systems. All right. With that, I would like to conclude the first lecture on the fundamentals of computational fluid dynamics. I hope with this introduction, you will be able to use software tools a bit more confidently. Thank you. Bye.
Part 4
T-Joints are important flow control devices. You can find them in a lot of places including your home. It's called a T-Joint because it splits the flow and the angle between them is 90 degrees. Here is a fun example. Say that I'm in a room that needs to be air-conditioned and my refrigeration system can provide chilled air at 10 degree Celsius because that is where its efficiency is optimum. But 10 degree Celsius is too cold for me, right? I would prefer something like 25 degree Celsius. So what is the solution? This is where a mixing tea is useful.
I can have the outside air come through one of the inlets and I can have my chilled air come through the other inlet. Now by adjusting the flow rate of the chilled air, I can get the desired temperature at the outlet of the mixing tea which would be connected to the air-conditioned space. Seeing this in action is going to be the objective of our simulation and we are going to be simulating this entire process using Ansys. By using Ansys workbench, I can establish a workflow to work on this project.
We are going to start from the left-hand side and work towards the right. The first thing that I need to do is create a computational model and I'm going to be using Ansys SpaceClaim. You can see that I'm creating a solid model here. Now remember for CFD, you don't need the solid model but I'm still doing it because I might need it for a future analysis. To capture the fluid volume, I'm going to be using the fluid volume extraction utility in SpaceClaim by selecting the respective edges.
This is the wet volume and this is the solid volume. Wet volume is basically the volume that's occupied by your fluid. Then we are ready to create a computational mesh. We are going to be using Ansys mesher for this and I'm primarily going to be creating a tetrahedral mesh with inflation layers to capture the boundary layers. Now we can use hexahedral mesh because they have better quality but tetrahedral should do the job. After checking the simulation quality and measuring important metrics such as orthogonality and skewness, we are ready to move to Ansys Fluent.
Ansys will be used to set up the initial conditions, boundary conditions and the required numerics to solve this particular problem. We are employing a steady state solver in a pseudo-transient mode in this particular case. And as far as the turbulence model goes, we are using realizable k-epsilon and after this, we set up our convergence criteria. Now since we are simulating flow through a mixing T, we need to measure the effectiveness of mixing. And this is where elementary statistics comes to our rescue. We can calculate the standard deviation of temperature at the outlet to measure the effectiveness of mixing.
A high standard deviation means that the mixing is bad and a low standard deviation means that we are getting pretty good mixing. And in order to make the study fun, I have set up three configurations. One is the baseline configuration where the coolant comes in at a velocity of 3 meter per second. Second, we are using the same configuration but the velocity is increased to 6 meter per second. And for our final configuration, we use a mixing T that is slightly longer. Here are the results.
Now in order to get good mixing, we see that the velocity of the chilled air is very very important. By increasing the velocity of the chilled air by a factor of 2, we are able to reduce the outlet temperature quite nicely. The longer configuration is not at all helpful because it gives very similar results to our baseline configuration. When we actually increase the velocity, we get better turbulent mixing. In this particular figure, you can see that the majority of the mixing occurs because of the higher velocity.
And as we go along the length of the pipe, there is additional mixing that occurs because of diffusion, but it is not that dominant. Alright, I hope you found this video interesting. In the next video, we will show you how we set this simulation up. Thank you. Bye.
Part 5
Hey guys, hope you're all doing good. In this video, we are going to be showing you how you can actually simulate flow through a mixing tea. Alright, I hope in the previous video you got some background on the uses of mixing tea. So let's get started. Make sure that you start Ansys Workbench first. And then what we are going to do is we are going to take in the Fluent module and we are going to drag and drop it here. So this is basically the list of steps that we are going to be working on. We are going to be starting with a geometry preparation where we will load the geometries and you also have access to these geometries, so you can load them as well.
And we will be performing a series of operations in the geometry part, the meshing part, the setup part, and finally post-processing. So let's get started. So what I want you to do is double click on the geometry part. And on the bottom left corner, you will be seeing a message called as Starting SpaceClaim. So SpaceClaim is basically a very nice CAD cleanup software that can be used to prepare CAD models for simulation setup. It could be for FEA or it could be for CFD. Now, once this is loaded, what I want you to do is click on File, Open, and you are going to navigate to the location where the geometry files have been stored.
So we are going to be selecting the mixing tea shot geometry and we are going to click Open. All right. So now the geometry has loaded. So this is what you call as a mixing tea. So a couple of things that you need to get familiarized. So you can use your middle mouse button to basically, you know, rotate the geometry. Use the scroll wheel to go in and out and use control plus scroll wheel to actually pan the geometry. Right. And all right. So it's very useful to learn these navigation tools because as you work with complex geometries, you need to be able to navigate to the finer regions of the geometry and select the quantities that you want.
So here you can see that this is basically a solid model. How do I know that? Well, if you actually zoom in, you can see that the thickness is provided and you really don't want this to be part of your CFD simulation unless, you know, you are basically going to be doing a solid simulation as well. What do I mean by that? For example, in the same problem, if I select this outer phase and if I apply a particular value of heat flux and I'm interested in seeing how this heat travels through the solid and how it heats up the fluid and how this heat flux affects the outlet temperature.
Right. If I want to perform such a study, then retaining the solid geometry makes sense. However, in this particular case, we are interested only in performing the flow simulation. So we do not need, you know, for example, this particular phase, this particular phase and this particular phase. Right. So I can go ahead and delete all those things. But, you know, that's going to be a time consuming process. And that is where SpaceClaim plays a huge role. It provides you a lot of options that you can use to extract only the fluid volume.
Right. So what does fluid volume mean? When you're running a simulation, your fluid is going to interact with particular parts of your geometry. In other words, in this particular case, the fluid is not going to be involved with the outer surfaces of this particular model. Right. So the fluid volume in this particular case involves this particular guy and this particular guy. Correct. So this right here is your fluid volume. Now, how do I extract that? Well, if you go to prepare, there is an option called volume extraction.
Right. So what we are going to do is we are going to first use the select edges method. And we are going to basically select the following three edges. Right. So you don't have to hold on control key for multiple selection. Once you make a selection, it's going to show up. And if you hit escape, all of your selections would be lost. So let me select the edges again. Select this guy, this guy and this guy. All right. So once I do this, I can just click on this green mark.
And if I wait, you can see that a new object has been added to my tree. So what I'm going to do is I'm just going to hide the solid component. And whatever I have left is basically my fluid volume. I don't have to delete my solid component. I can just hide it and to make sure that the solid component is not loaded for the mesh generation process. I can right click and use the suppress for physics option. So this way, the solid model is not going to be loaded when we are doing the meshing process or when we are setting up the simulation is influent.
And you can see that there is a small symbol on the solid. Right. There is a cross symbol which basically says that it has been suppressed. So make sure that you do this before you proceed on to the meshing stage. All right. So now what I can do is I can just close this particular model. Now things will get saved automatically. And if I double click on the meshing module, it's going to say that it's starting the meshing module. All right. So you can see that the geometry has been loaded.
And this is basically the meshing interface of ANSYS. And again, from a navigation point of view, you know, you can basically use the middle mouse button to rotate the geometry, use the same thing to zoom in and out and use control plus the middle mouse to basically pan. Right. So pretty standard options. And one thing that I forgot to say in SpaceClaim is you have this anti tools here. So just click on Z and it's going to align the view accordingly. Right. So this is very convenient in my opinion.
Right. So make sure that you play with these things. And they have a very nice way to basically look at the dimension. So, for example, if I click on a particular face, you can basically measure the area. But say that you want to know the diameter of it, you know, approximately. So you have this black bar here or at least black and white bar here, which you can basically use to basically estimate the length of a particular component. So, for example, this distance right here is zero point zero two meters.
And I would like to say that, you know, at this level of zoom, my diameter is maybe, you know, half of point zero two five. Right. I can kind of make that estimation. Similarly, you can view the model like this and, you know, go really small and you can estimate the model is approximately point two meters long. Right. So this is very convenient. So what we are going to do now is we are going to basically name the boundaries that we have so that we can add the appropriate meshing.
And also when we are applying the boundary conditions, you know, we can select the boundaries that we have created. So first, I want you to familiarize yourselves with the entity selection here. So you can basically select a single vertex by clicking on this button. You can use the edge selection and you can see that the shortcut is control E and use control F for face selection and control B for body selection. Right. So I'm going to hit control F and I'll be able to select this particular face.
And then if I right click and if I select create named selection, I can provide a name for this particular face. All right. So what I'm going to do is so I'm going to provide the following name inlet dash Y. Why am I giving this name? So first of all, it's an inlet boundary. Right. I want one fluid to enter from there and dash Y because it's aligned along the Y direction. Right. So let's click OK. Similarly, you can select this guy. Right. Right. Click. And you can use the shortcut key N, which is very convenient.
And this is going to be another inlet. And this time this is going to be inlet X. Right. Because it's along the normal of the inlet phase is along the X axis. Similarly, we can select this guy and I'm going to use the shortcut key N and I'm just going to call this outlet. Right. So in addition to this, we have to select our walls. So let's select this guy and this guy and use the shortcut key N. And I'm just going to call this as a wall or walls.
OK. Doesn't matter. So then we are going to go ahead and do the meshing process. So first for the machine, I'm just going to use the very basic options. Right. So first click on meshing and you can immediately see that you have details of the mesh. Now, in general, if you have never done CFD before, don't try to play with too many parameters. And especially in cases where the meshing is being done manually, try to stick to a very basic mesh first and make sure that you understand the process.
Once you are able to get one type of solution, slowly change one parameter at a time to see what happens. At the end of the day, you need to gain experience with machine to basically understand, you know, what each option does. So you have to be a bit patient. Right. So in this particular case, what I'm going to do is let me just open up the sizing option here. And you can basically see that most of the properties have already been set. How does this happen?
Well, Fluent basically uses a bounding box. What's a bounding box? Now, say that you have this particular geometry. What is the smallest size box that I can use to fit all of these components? That would be a bounding box. Now, a box has a particular dimension. Right. LX, LY and LZ. Based upon that box, Ansys Fluent would basically create a mesh automatically. All right. So I will be explaining those details later. So for now, just trust me and whatever options you have here, just leave it as it is.
And that's basically what I'm going to do. So right click on mesh and click on generate. That's it. So this is basically computational mesh. Right. So what I can actually do is let me put this along the y direction. And I'm going to basically use this new section plane button, click and drag to create a line. And what the mesh is going to do is it's going to cut the geometry so that you can see what's actually happening. Now, in this particular case, the geometry is not that complex.
But if you have a lot of parts, then you might want to create multiple section planes like this to investigate various parts of the geometry. Right. So in this particular case, what I'm doing. So what I'm going to introduce, I'm going to basically click on this quality and for mesh metric, I'm going to basically select element quality. And when I do this, I immediately get an image which looks more like an histogram. On the x-axis, we have the element metric, which is basically the element quality.
And on the y-axis, we have number of elements. So you can basically see that the minimum quality is 40% and the maximum quality is 100%, right? So that's a scale of 0 to 1, 1 being the maximum you can get. Now, ideally, you want all of these cells to have a very good quality, but I still haven't talked about what quality is. Now, we will not be discussing how quality is being calculated exactly in this particular video. In fact, that's not going to happen for a while because in order to understand the term quality, you need to basically understand what finite volume discretization is.
And you need to basically understand how fluxes are computed and what type of interpolation schemes that are being used. In order to understand how the quality is being computed, you need to have a very good handle on topics like finite volume method, interpolation schemes, aspect ratio, skewness, and orthogonality, all right? These topics will be covered in detail in this course at a later point of time. So for now, what we are going to be looking at is we are going to be looking at our element quality and we want to make sure that it's not lower than 5%.
Now, in this case, it's 40%. So 0.4 is 40. So we are happy with this. So with this, you can actually run your simulation successfully. So what I'm going to do is I'm just going to close this guy and I'm going to open up setup. This is going to launch Ansys Fluent. So in my case, I'm going to be doing a parallel simulation. Your license might not allow you to do that. So if it is serial, meaning if you can run it only on one processor, that's fine.
These simulations would not take that much time anyway. I'm going to be doing a parallel simulation here with four processors. If you have more, for example, if you have six or eight, you can put in that many. So let's click on OK. All right. So now that we have opened Ansys Fluent, let's learn a little bit about Fluent's environment. So here you basically have a console. So this is where important information regarding your simulation, geometry, or mesh is presented. If there are any errors, it's going to show up here.
Some errors show up in dialog boxes. But if it's a runtime error, meaning if the error is generated as your simulation runs, then that's going to appear here. And you have a graphics plot here. And this is where you're basically going to see your simulation. So in this particular case, it looks like the geometry is not loaded. So what I'm going to do is I'm just going to click on right click, click on update. And let's just click on the show progress bar to see the progress.
And what I'm going to do is I'm just going to close Ansys Fluent and I'm just going to reopen it again. So now let's try opening up Ansys Fluent one more time. And this time, one small change, I'm going to be using the double precision option. I missed that earlier. So now this time you can see that more information is being printed out automatically. Right. And now if I go to the graphics window, you can see that the geometry is already there. Now for Ansys Fluent, the UI navigation is a bit different by default.
So you have to use the left mouse button to rotate. You have to use the scroll wheel to scroll. And I believe it's control plus scroll wheel to pan. Yeah, that option is already there. And you have the same type of views. Right. So which is very handy. Right. You can just click on this to rotate this. You can do it like this. It rotates by 90 degrees. Right. This is also a very convenient feature to have. Right. So when you're starting with any type of CAD software or CAE software, you know, navigation is something that can drive people crazy.
So make sure that, you know, these options are familiar to you. So what do we do here? You can see that there is options on the left hand side sidebar and then there's options on the top menu. Right. So you can use both the approaches. Typically, what I like to do is I use the top menu bar options to basically set up the case because it actually has a very nice flow. Right. You set up the simulation, you set up the physics. If you have any user defined parameters or function, you set that up and then you talk about solving like initializing your model.
And then, you know, if you want to do hybrid initialization, you would do that. And then you would finally do the post processing. And I personally feel that the top menu is less cluttered than this guy. All right. So what we are first going to do is we are just going to, for example, you know, check the mesh quality. So we have clicked on it. And if you go to console, you can basically see that just prints everything. And it says checking mesh and it says done.
You don't see any type of warning messages or error. That typically means your mesh is OK. So what Fluent is doing is checking for the mesh quality. And if it passes its minimum standards, it's just going to move on. All right. So in this particular case, I'm directly going to jump to setting up physics. Now, in this case, we are going to be using a steady state solver. And for really small Mach number flows, it's good to use pressure based solver. Now, what is pressure based and what is density based solver?
These things would be covered when we're talking about the required theory for this particular course. Right. So for now, just use the pressure based solver. And for velocity formulation, we are going to be using the absolute option. Now, we are interested in solving for temperature. So we will be using the energy equation as well. And then in this particular case, the flow can be turbulent. Right. So I'm going to select a viscous model and I'm going to be using the K epsilon turbulence model. And we are going to be using an upgraded version of it, which is called as the realizable K epsilon model with standard wall functions.
So all of these options have to do with how turbulence is being captured in your case. So in this case, at least the K omega K epsilon models is what you call as a RANS based turbulence model or Reynolds average Navier Stokes based turbulence model. So in this case, let's just click OK. And then what we are going to do is we are going to add materials. So let's just click on material. And what we want to do is we want to basically say that air is the working fluid.
Now, if you want to use a different fluid. So, for example, let us say that we want to select water so we can actually go to the fluid database. Here you have a list of species that you can select. So in this case, let's just go with airflow and let's just close this guy. Right. So then what we are going to do is we are going to assign the boundary conditions. So go to zones and first select cell zones. So here what we are going to do is we are going to select our volume region, which is basically the internal volume, right, which is basically your mesh.
Right. And we are saying that it's of type fluid, which is correct. And if you click on edit, you can see that the material name is air. So if you change the material name to water, you better change it here or else you are not going to see any difference. So let's click on OK. OK, the next thing I'm going to do is I'm going to select boundaries and let's start with Inlet X. So you can see that the type is automatically set to velocity inlet.
How does Fluent know that? Well, based upon your boundary name, whenever it sees a name, inlet, outlet or wall, it's going to select the boundary conditions or apply the boundary conditions automatically, which is a very nice feature. So, for example, in this case here, I have named my boundary wall. You can see that the type is automatically set to wall. So that's pretty cool. So let's select Inlet X and we're going to click on edit. And we're basically going to say that along my inlet face, my velocity is going to be three meter per second.
Right. And this is a velocity magnitude which is normal to the boundary. Right. So the normal for that particular face is going to be pointing inside the model. Right. And which means along the positive X, the flow is going to be three meter per second. And what we are going to do is we are going to say that the temperature of this incoming inlet is going to be 300 Kelvin. So let's click on OK. Now, say that you are used to a different unit system. What would you do?
So that's where you go to setting up domain and you have units. So what you can do is you can basically say select temperature and you can say that that units is going to be Celsius. Let's close this guy. Go to setting up physics. Click on boundaries. And let's click on inlet. And if you click on edit, go to thermal. You can see that the temperature has automatically been converted to degrees Celsius. So let's just round this off to 25. All right. And let's click OK.
Inlet Y, that is also going to be a velocity inlet. But this time what I'm going to do is I'm going to say that, you know, the velocity is 10 meter per second. Feel free to play with these values. You know, once you have done your first simulation, right. So for the first time, I want you to just use my exact values. But once that simulation has run, feel free to change it to basically see how the results change. For thermal specifications, I'm going to say that the incoming fluid or incoming air is at a much lower temperature, 10 degree Celsius.
Let's click OK. All right. So that basically concludes the setup. I mean, we are almost there. Now, this is going to basically let the air flow and it's a steady state simulation. And hopefully, you know, we can expect to see some temperature information at the outlet. Now, what else do I need to do? So if I go to solving options, right. As the solution is being solved, you know, I basically want to monitor some physical quantities, for example. So let us say at the outlet, I want to see what's the temperature.
Right. I can do those things very, very easily. So what I'm going to do is I'm going to click on definitions and I'm going to say that I'm interested in computing the area weighted average of, you know, for example, the variable that I'm looking at is my temperature. Right. So the phases where I actually want this averaging to be done is the outlet. All right. So I'm going to select this. I'm going to say, OK, create a report file and a report plot. And this is going to be done.
And let's just click on OK. So this is going to add one type of monitor. So let me do one more thing. In this particular case, we are doing a mixing problem, right? So in order to understand the effectiveness of mixing, one thing that engineers look at is surface is standard deviation. This time, what I'm going to do is use surface report, but I'm going to be looking at standard deviation. And I'm going to be looking at the standard deviation of the temperature at the outlet. And let's just create another report for that.
And let's click OK. All right. So anytime you want to manage the things that you have created, you can go to definitions and you can go to edit and you can basically see the reports, right? So for example, report one, it's a standard deviation on a surface and the surface that you're looking at is the outlet, right? And it's going to be plotted and it's also going to be saved in a particular text file. So let's just close this guy. So once you have set up your initial conditions, boundary conditions, material properties, the next thing that you need to do is initialize the solution. So when you're performing CFD analysis, you're solving for velocity, pressure, temperature, and other variables possibly, right?
So what you need to do is you need to provide a starting value for all these variables. For velocity at the inlet, you have provided a value, but for pressure, we have not provided a value. And in fact, that actually brings me to an important point. I actually forgot to set up the outlet boundary condition. So let's click on outlet. You can see that it's already set to pressure outlet. And if you click on edit, you can see that at the outlet, the pressure is zero, right?
Now, remember, this is gauge pressure, which means the absolute pressure or the actual pressure is one atmosphere, right? Only then the gauge would say that the pressure is zero. And this has happened automatically, which is good for us, because we have named that particular phase as outlet, right? And this is very, very useful. So in this case, let me click on OK. And we're going to go to Sol and we're going to initialize. When we click on initialize, you can see that we are using the hybrid initialization process.
The hybrid initialization process uses a potential flow solver to compute an approximate value for velocity and pressure. So what you can do is you can click on initialize. And if you go to the console, you can see that after 10 iterations, the initialization is done. So now we are ready to run the solver. So I'm just going to run it for, say, 150 iterations and I'm going to hit calculate. So when you do this, you automatically see a lot of plots. So the first plot is you're looking at the area weighted average of temperature and you can see that it's converging around 18 degrees Celsius and the calculation is almost complete.
You know, it went up to 150 iterations. So what I'm going to do is I'm going to look at a bunch of parameters here. You can see that, first of all, the standard deviation is maybe close to one degree Celsius, right? And the average is 18. So you can say that our temperature is quite uniform because our standard deviation is quite low. So what I can do is I can increase this 250 and click on calculate. And I'm just going to say that use settings for future and current calculations.
And if I go to console, we are going to be starting from 150. All right. So we'll let it run and it's going to run for another 250 iterations. And if you can go to the graphics and you can basically see that both standard deviation and the temperature have reached a steady state. And here you're looking at something called as residuals. What do these mean? When you're doing CFD, you're solving a bunch of equations, governing equations. And these are partial differential equations. So we'll be looking at these equations when we talk about the theory.
But any partial differential equation or governing equation is going to be of the form, you know, the sum function which is equal to zero. Right. Now, when you are solving these functions, these governing equations on a computer, you employ some kind of an approximation. And that is what you call as numerical discretization, which in this case is called as finite volume method. So we use the FVM approach to discretize our partial differential equations and we solve it. When you actually solve the function, it might not evaluate to zero.
Right. It might actually evaluate to something really, really small, something like one minus five. So in this particular case, that is what you call as a residual. Now, the residual definition changes. But whatever description that I gave should be good enough for now. So as the number of iterations continue, what you are seeing here is that the functions are being calculated and they are very, very close to zero. Except in certain cases, for example, here you have the continuity equation, which, you know, evaluates to something like one e minus something like five e minus three.
Right. Which is a small number. And you can see that it is keeping on decreasing. So with that, our simulation is complete. So now we need to post process the results. So I'm going to go back to Workbench and I'm just going to double click on results. Make sure that you close Ansys Fluent before you do this, because you might have licensing issues. So, for example, currently you can see that I have this problem. So let me just click on request license. And there you go.
So you can see that my model shows up. I can enable all the components by, you know, showing them. So I'm just going to hide all my inlets and outlets. And for my wall, if I double click on this and if I click on rendering, I'm going to apply some transparency to it so that I can see through it. So now what I want to do is I want to look at the velocity distribution and the pressure distribution to see if it makes sense. So what we are going to do is we are going to do insert location plane and let's just call it as plane one.
That's fine. And I'm just going to be creating this in the X, Y plane. And we are going to say that the Z value is going to be zero. So let's click on apply. And that should create a plane right in the middle. Oh, and by the way, in CFD post to rotate the geometry, you have to use a left mouse button to zoom in and out. You can use a scroll wheel and to pan use control plus the scroll wheel. And I might actually go back to my wall, render and, you know, make it a bit more transparent.
And I'm just going to change the color to white. So that way, I think it's going to look very clean. So now what I can do is I can go back to my plane, double click on it. I can color this guy by a variable and let's say temperature and let's just click apply. So now you can see that the temperature changes from 9.976 all the way up to 25 degrees. So what you can do is for the range, you can say that it's user specified and you can type in a value of 10 and you can go up to, say, 25.
Click apply. And you can see that that has changed the legend. And then you can actually click on default legend view, go to appearance. And for the text parameters, we are going to use a fixed point notation, maybe up to two decimal places. Click apply. And that's going to give us a very nice looking legend. So a lot of interesting things can be observed, right? So what's happening here is the mixing process is primarily determined by the velocity of the incoming jet. And you can basically, you know, use different values of momentum ratio, which basically means the ratios of velocity at inlet X to inlet Y.
So currently you can see that the velocity here is 10 and here it's 3. So depending upon how you define the momentum ratio, it's either one, you know, almost one third or it's 3.3. Correct. So depending upon how you define, you can basically examine the effectiveness of the mixing process, which is part of your homework. Now, here you can see that the solution here looks quite diffuse and that's because of a mesh resolution. So what I can do here is I can click on my plane, actually double click on it.
And when I render it, I can show the mesh lines, right? So you can see that there are really large cells at this interface. And because of that, you are getting a very diffused looking profile. This is very common. And this is what you call as a course simulation. And as I said, during the start of the video, whenever you begin with CFD, run these type of simulations to first understand how a CFD result looks like. And then slowly refine your setup. So in the next video, I'm going to be showing you how to add inflation layers to the mesh and how to make the mesh more refined.
I'll see you guys in the next video. Bye.
Part 6
Alright, so with that, our simulation is complete. So now we need to post-process the results. So I'm going to go back to Workbench and I'm just going to double-click on Results. Make sure that you close Ansys Fluent before you do this because you might have licensing issues. So for example, currently you can see that I have this problem. So let me just click on Request License and there you go. So you can see that my model shows up. I can enable all the components by, you know, showing them.
So I'm just going to hide all my inlets and outlets and for my wall, if I double-click on this and if I click on Rendering, I'm going to apply some transparency to it so that, you know, I can see through it. So now what I want to do is I want to look at the velocity distribution and the pressure distribution to see if it makes sense. So what we are going to do is we are going to do Insert Location Plane and let's just call it as Plane 1. That's fine.
And I'm just going to be creating this in the X-Y Plane and we are going to say that the Z value is going to be 0. So let's click on Apply and that should create a plane right in the middle. Oh, and by the way, in CFD post, to rotate the geometry, you have to use left mouse button. To zoom in and out, you can use a scroll wheel and to pan, use Ctrl plus the scroll wheel. And I might actually go back to my wall, render and, you know, make it a bit more transparent and I'm just going to change the color to white.
So that way, I think it's going to look very clean. So now what I can do is I can go back to my plane, double-click on it. I can color this guy by a variable and let's say temperature and let's just click Apply. So now you can see that the temperature changes from 9.976 all the way up to 25 degrees. So what you can do is for the range, you can say that it's user-specified and you can type in a value of 10 and you can go up to, say, 25.
Click Apply and you can see that that has changed the legend. And then you can actually click on Default Legend View, go to Appearance and for the text parameters, we're going to use a fixed point notation, maybe up to two decimal places. Click Apply and that's going to give us a very nice looking legend. So a lot of interesting things can be observed, right? So what's happening here is the mixing process is primarily determined by the velocity of the incoming jet. And you can basically, you know, use different values of momentum ratio, which basically means the ratios of velocity at inlet X to inlet Y.
So currently you can see that the velocity here is 10 and here it's 3. So depending upon how you define the momentum ratio, it's either one, you know, almost one third or it's 3.3, correct? So depending upon how you define, you can basically examine the effectiveness of the mixing process, which is part of your homework. Now here you can see that the solution here looks quite diffuse and that's because of your mesh resolution. So what I can do here is I can click on the plane, actually double click on it and when I render it, I can show the mesh lines, right?
So you can see that there are really large cells at this interface and because of that, you are getting a very diffused looking profile. This is very common and this is what you call as a coarse simulation. And as I said during the start of the video, whenever you begin with CFD, run these type of simulations to first understand how a CFD result looks like and then slowly refine your setup. So in the next video, I'm going to be showing you how to add inflation layers to the mesh and how to make the mesh more refined.
I'll see you guys in the next video. Bye.
Part 7
Hey guys, hope you're all doing good. In the last video, I showed you how to set up and run a very simple simulation of flow through a mixing tea. So now we are just going to improve the setup. So we'll start with the meshing, okay? I'll open up the meshing module and we are just going to click on our mesh, take a look at it. So this time what I'm going to do is I'm going to go to sizing. You can see that there is a parameter called element size which is set to 1.02E-2.
We'll be taking a look at how these calculations are made by the meshing solver. For now, what I'm going to do is I'm just going to type this as 5E-3. All right, so let's do that. And what we're going to do is we're going to right click and we're going to clear generated data and then we are going to redo the mesh generation process. You can see that it's a bit more finer. So what I can do is I can make this 0.001 and generate the mesh one more time.
Now you can see that it's taking a little bit more time. And now you can see that our mesh is quite fine. In fact, this is probably too fine, but that's fine. You can see that our mesh is quite fine, but this should be okay. Usually you don't have to use this many mesh points, but for demonstration purposes, I think this is okay. So what you can do is you can just close this guy and we are going to move to ANSYS Fluid. Now, one thing that you need to keep in mind is that if you are following along these lectures using an academic license of ANSYS, you might not be able to run this particular case.
So you make sure that you use an appropriate grid size so that you are still able to practice these exercises. So it looks like we have a license error. That's because CFD post is still open. So we are just going to hit enter to exit and we are going to try opening it again. All right. So let's click OK. So what's going to happen is if everything is set up correctly, the mesh should load again. So we go to console and looks like the mesh is being loaded and let's go to graphics.
So looks like this is not updated. All right. So now what we are going to do is we are going to update our mesh so that this mesh is also available in our Fluent calculation. OK, so now we are going to right click and before we start the Fluent setup and then what we are going to do is we are going to double click and open the Fluent setup. It says the upstream mesh has been changed. Do you want to load the new mesh? Click on yes.
So this is the cell count. You can see that it's actually quite huge. We have about 5 lakh. So you can see that we have about 0.5 million cells. That's a lot, especially for this type of a problem. Let's just make sure that our settings are still preserved. So let's go to setting up physics, go to boundaries, click on Inlet X. And if you edit this, you can see that the settings are still there. So now what we are going to do is we are going to go to solving.
We are going to initialize this guy. Now, since you have a lot of cells, everything is going to take a little bit of time. We are going to be using 300 iterations. Now, remember, I have parallel license of answers, so I can run it in four processes. If you have a serial license, then this process will take a decent amount of time. So let's click on calculate. And if I go to graphics, in a short amount of time, the residuals and the other reports should open up.
So I'm looking at standard deviation and scaled residuals. You can see that each iteration is quite slow. That's because you have more cells to process. Now, remember, your governing fluid dynamic equations are being solved in each and every cell. So the more the number of cells, the more equations you have to solve. And that is why your simulation time is increasing. In addition to that, there is something called stability. Meaning, if you increase the number of cells, the compute time does not increase linearly. It can increase in a quadratic way as well.
Right. So the bottom line is that you need to be very careful when you're playing with the mesh settings. You cannot use a lot of cells without the proper justification. Here you can see that the residuals are dropping, but they are dropping very, very slowly. This is very common, especially when you're refining your mesh. Because as you add more cells, your solver is going to take more iterations to converge to a particular value. Whereas if the mesh is coarse, meaning if you have few number of cells, then your linear solver can actually converge pretty fast.
All right. So it looks like the calculation is complete. And if I'm looking at the standard deviation of temperature, you can see that it is around one point. You can see that it's around one degree Celsius, kind of the same value. So let's just quit this guy. You can see that Fluent is writing the case file so that it's updating the current solution. So now we can actually right click and update our results section. So this way, the results from Fluent is going to get transferred to CFD post.
And now if you double click CFD post, it's going to open it up. And most likely the post processing that we already did is still going to be there. Now you can see that our temperature distribution is much more sharp. So I can click on my plane and I can just remove the mesh lines, click apply. And you can see that this looks a bit more finer. It is still a bit smoothed out, but it looks much better. Right. So this is what you call as mesh refinement.
And as I said, you really don't have to use this fine of a mesh for this particular problem. And that is why you perform something called as a grid dependence test. You start with the base mesh and you make it finer in a systematic way. And with each and every increment, you should get a converged solution, meaning your outlet temperature in this case or the standard deviation of the outlet temperature in this case should reach a steady value as you are progressively refining the mesh. Obviously, this has to occur at a particular tolerance.
And when this happens, and when this happens, your solution is set to have grid converged. All right. So with that, I would like to conclude this video. And with that, I would like to conclude this video. Please make sure that you attempt the challenge on our challenges interface before you move on to the next video. Thanks. Bye.