 Yes, so I would request Professor Harikrishnan to begin the session. So Professor Harikrishnan is a professor in the division of mechanical engineering from coaching a university of science and technology. So over to you, Professor Harikrishnan. Hi, thank you very much, Sahil. Very good afternoon to one and all. So myself, Dr. Harikrishnan, I am an assistant professor from division of mechanical engineering, full of engineering, coaching university of science and technology, Kochi Kerala. So I welcome all of you to the open form community. I would say it's a community because it's an open source software and we need to help each other in order to gain knowledge. And regarding my background, I did my PhD as well as postdoc from IIT Madras. I completed my PhD from IIT Madras in 2019 and I was working as a postdoc in IIT Madras for almost two and a half years. So I received an institute research award for my PhD thesis. And I have almost eight to nine years of experience in CMD by using various commercial software as well as open source software. So I will quickly introduce about my research topics before I begin, especially related to CFD. So this is the single slide I would just like to explain or briefly summarize my experience in CMD for last eight years or nine years. So I worked in various problems starting from single phase problems and multi-phase problems. I hope you have been knowing the difference between single phase and multi-phase. In multi-phase problem we have more than one phase. So it can be multiple liquids, it can be air, water, or it can be air, solid particles, etc. The first slide or first part of the slide indicates single phase flow in a corrugated channel. Corrugated channel is nothing but a channel in which we have some sort of corrugations. And what happens when we have corrugations? We know that there can be some circulations present inside the chamber. And I was mostly interested in the flow transition phenomenon happening inside various corrugated channels. This was part of my PhD work. So this part actually I was using open source for doing the simulation. I was using direct numerical simulations for doing this part of the study. And multi-phase flow, I did two different approaches. The first one is Eulerian Lagrangian approach. So I hope you may be knowing the difference between Eulerian approach and Lagrangian approach. Lagrangian approach, we are practicing these particles individually. So this is actually a two-dimensional simulation for a makeshift isolation enclosure. What is a makeshift isolation enclosure? So we know that during COVID, so the hospital, we know that there was a constraint about the availability of hospitals rooms. Government tried to convert these rooms, available pooled rooms as well as some auditoriums, some small isolation enclosures. And those rooms are known as makeshift isolation enclosures. And in this particular simulation, what we did, we did considered droplets spread in a hospital room. So we know that when we talk, why we are wearing masks, when we are using, when we talk or when we cough, or when we sneeze, thousands of particles are injected into the room, right? We are trying to prevent it. We are trying to screen it with the help of a mask, right? So in this particular simulation, what we considered is somebody is lying. So this is actually a bed, a two-dimensional simulation. And this is actually a bed and somebody is sneezing or somebody is coughing. So thousands of particles are injected and this is actually a closed room in which we have an inlet here and outlet here. So some sort of ventilation is provided. So what happened is, when somebody sneezes, these particles enter into the room and due to this ventilation or circulation problem inside the room, so this is actually spreading within the room. So in this part of the work, what we considered is how to prevent or how to reduce the risk of contamination or how to reduce or how can we efficiently fresh these particles away by changing different parameters. And the last part is a part of the work I did as part of my post-do. So here we just consider particle cluster falling in a pool of liquid. So we know that when we have a cluster of particles, it behaves like fluid, right? So when we have a bucket of sand and if you just pour it, it may look like a fluid flow rather than a centrifugal solid particle flow rate. So we use oilier in oilier in approach in which we have some container is filled with air in which filled with water and air. We are dropping a bucket of sand particles into it and its distribution is analyzed with the help of commutation fluid dynamics. So the left hand side and right hand side, left hand side figure is an animation I made in paraform. So this part of the work I did using open form, using multi-phase oilier in form. And the right hand side is a particle cloud shape evolution with respect to time. And left hand side is the two-dimensional view corresponding to that. So moving on to the content, so this is a brief overview of my research topic. So I hope at least some of you are new to this open form. I would like to know how many of you are doing open form for the first time? Like learning open form or not at all familiar with open form? Yes, I can see there are more number of hands. So don't worry, even if you don't, even if you are not getting the first simulation, I will tell my experience when I started learning open form in 2014, it took one week for me to install it. At least you were able to complete the installation within a day or two, right? So it was almost one week. So if you are facing some challenges, so be ready to face it, there are a lot of learnings. That's what I would say at this point of time. Yes, you can lower your hands. So regarding the content, as I said, there are multiple sessions. And I'll be mainly talking about introduction to machine. And Jenny madam already gave a very good introduction to the computational fluid dynamics and brief overview about it. So I will start with that, then I will move on to machine. So basically I will be covering what is machine, why it is important. And there is one more session in the afternoon by Dheera sir. And sir will be taking the advanced portion for the machine. And I will also touch upon some basic things like machine generation in open form. What way we are making machine, we can do machine in open form, etc. So let me move on. So as I said, I will just quickly in two to three slides, I will quickly explain what is computational fluid dynamics once again. So we know that any physical phenomena with certain assumptions can be modeled with the help of some mathematical equations. So if it is a fluid flow, the mathematical equation is known as Navier-Stokes equations. So these mathematical models can be linear or polynomial relationship differential equations and partial differential equations, etc. So for a fluid mechanics problem, we know that the governing equations for the mathematical model that fits is partial differential equation. And the basic equations are known as mass momentum conservation equations. So the main important parameter is with certain assumptions. Assumptions are very, very important. So with certain assumptions only we can convert that into mathematical models. If our assumptions are wrong, we will get wrong results. So in computational fluid dynamics, what we do is we know that the basic governing equation that is Navier-Stokes equation or mass and momentum conservation equations. So they are partial differential equations. So as I said in the previous session, we are going to convert these partial differential equations into set of algebraic equations. So what do we mean by algebraic equations? Algebraic equations are nothing but x is equal to b form. Or we have certain unknowns and we need to find unknowns by analyzing or by solving a set of equations. So the first part is when we convert these partial differential equations, so let us consider a simple case. Let's say I have a simple case that is close to a channel. So these are the walls and there is some fluid entering, some fluid leaving. So this is a physical problem. So physical problem, let's say somebody asked me if I give the velocity of one meter per second, what will be the pressure at the outlet? Somebody asked this kind of question. So the first thing is we will just try to make a geometry and we will also learn what are the conditions, what are the fluids used, if it is water, if it is an ketone fluid and if it is a pain or if it is blood, we should consider non-rotant fluids, etc. So there are certain assumptions that is involved in this part. And the second part is we are going to divide this domain into certain number of cells. So that part is known as meshing. So when we have domain, we are just dividing this domain into certain number of cells as in the previous session also madam already introduced this thing. So then what happens? We are going to apply these equations here. So in each cell we are going to apply this governing equation as well as we know that when we have a partial differential equation, or normal differential equations, we should also need a boundary conditions to solve this. So along with the boundary conditions, as well as initial conditions, we will be solving this. And finally, okay, when we solve this, so when we apply this partial differential equation in each cell, finally we will be getting an equation in the form of algebraic equation. So that means, so in a flow problem, you know that the variables are u, v, w that is x component, y component and z component of the velocity and pressure. So these are the unknowns. So we are essentially calculating or identifying the x velocity, y velocity, z velocity as well as pressure in each and every point within the domain. So that one we are actually identifying the help of solving x is equal to b. So when we come, when we discrete, when we divide this domain and apply these governing equations and discretize the equations, we would say we finally we will get an equation in the form of x is equal to b. So ax is equal to b as I said earlier is an algebraic equation and we need to solve this with the help of computers. So the computational fluid dynamics indicate we are basically using computers to solve fluid dynamics problems. So if the number of cells, so to say the level of difficulty about CFD, if the number of cell is 10 or let's say 30, so we need to find 30 unknowns, that is you not all not just 30 unknowns, we have 30 equations. And we need to find unknowns 30 into 4 because u, v, w and p we need to identify in each and every point. See if we have 30 cells, these 30, these 30 unknowns, 30 into 4 unknowns we need to find by solving ax is equal to b equation. So manually calculating it is not possible. So we have with the help of computers, we are solving this. So there are different approaches, there are different methods for doing this. So those things you will learn in the upcoming classes, upcoming session. So what are the applications? So I think you may be already familiar, some of the applications are listed here. The first one is flow pass, bluff body, the aerodynamics. The second one is more relevant now. So the COVID once again there is another wave is expecting. So in this particular animations what they considered is somebody sitting in a restaurant and he's sneezing of his talking. So these particles are injected into the room and deposited in the table. So this is a computational fluid dynamics simulation based animation. And nowadays it is widely used to understand the spread of these contaminant particles or this virus laden droplet. And the third one is another interesting one. This is battery thermal management seminal electric. We know that nowadays there are a lot of accidents happening in EVs. So some fires it was reported by multiple manufacturers that fires are occurring in different battery systems. That is mainly due to poor thermal management. So if we know that there are multiple lithium ion batteries are tagged into a battery system and we know that during the charging and discharging. There will be heat generation and if the heat generated heat is not dissipated in a proper way what happens? There will be a temperature gradient within the temperature rise obviously and also the temperature gradient will also exist within the battery or within the system. So when there is a temperature gradient there is a good chance of short substituting and that gets to the fire. So battery thermal management system is one of the leading topics in which the computer fluid dynamics people are working on. And some of the other interesting part is this one is a floating object. So this one I downloaded from YouTube. The link is also provided. So floating body in a container and the last one is a propeller in an open water. Let's again open from base simulation. So these are some of the applications of computational fluid dynamics. Let me move on. So I just covered what do we mean by computational fluid dynamics. So there are different packages involved or there are different packages are available in the market and for doing these kind of simulations. So some of the most popular CFD packages are in stone here. First one is ANSYS. I hope you may be knowing ANSYS. I have a Fluent as well as CFX. Then Star CCM, Comsol, Autodesk CFD, Simulia, DSS Simulia, etc. And open form and juries. The main difference between the packages shown in the left hand side and the right hand side is the packages shown in the left hand side is purely commercial software. So if you want to use it, you may need to pay some amount. We know that for some software like ANSYS2 and student versions are available but there are very limited features available for students. So that is available for one year and I think cell size up to 5 lakh will be provided. And if you want to use above that, it is not possible. You may need to pay lakhs of rupees to get a license. On the other hand, the right hand side software like juries or open form are open source software. So anyone can download, anyone can install in any system and you can use it. No one need to pay any single amount of money to anyone. So open form as well as there is some of the open source CFD packages widely used in the community. So before we move on to open form, we should also know why what are the advantages of open form or what are the advantages of open source software. The basic thing as you know already is free of cost. As I said earlier, it is available and you can install, you can download, anyone can do it. There is no licensing issue. So you don't need to pay any amount for license. And the second one is source modification. That means these software, especially these proprietary software or commercial software, they may have some unique method for computing some certain things. So they will, they always keep their board at a black hole. So if you want to do some modification in the source code, it is not possible. On the other hand, open form or any open source software, it is very easy. The code itself is available and some modifications. If you want to do it is possible. And the third part is collaborative development as I showed in this image. So there are, let's say you are, you are working in India and you have a collaborator in US, Japan, et cetera. And they are also working with certain companies. So if you want to work on a similar problem using information software, they should also have access to the software, right? They should also purchase the software. On the other hand, if you're using an open source software, wherever maybe you can actually work together, you can put your code in a GitHub or something. They can download it. They can make use of it. They can actually improve the code. They can identify if there are any mistakes or if you can improve the code. So this collaborative development is possible when we have open source software. So even though we have certain advantages, there are disadvantages also. The first part is lack of documentation. So we know that when we want to learn a software, we should have some, they should provide some documents to begin with if you are not familiar. So the lack of documentation is one of the part. And nowadays we have actually documentation part is really improved day by day. And the second part is GUI. What is GUI? Graphical user interface. So right now I am using, let's say PowerPoint. And if I want to change the font size, I can just double click and change the font size, right? So basically we have a graphical user interface for doing all these things. On the other hand, if I use latex or some other software without GUI, I need to type some command to reduce the font size or increase the font size. Or if you want to add some effect, I may need to use some CLI command line interface to modify that. So similarly in open form or mainly in open form GUI is not available for doing the simulations. For post processing the results, GUI is available, that is not a SPARA view. And the open form version of SPARA view is not a SPARA form. Okay. And the last part is user friendliness. So as compared to commercial software, it is not that attractive. You may not get the results within a day. Or if you are doing the simulation for the first time, I'm pretty sure that you won't be able to get the results in a day or two. It may take some time for you to. So user friendliness is one of the disadvantages of open source software. Briefly, I will just quickly go through this slide. So open form is a full form of open source field operation and manipulation. It's a finite volume method and it's based on C++ language. And it started in the early 90s from Imperial College London by Henry Weller and Jasak. So when we specifically talk about documentation. So the interesting part is the open form user guide available in the open form directory is only around 200 pages. On the other hand, if you just consider a commercial software like ANSI's fluent, the user guide itself is 5000 pages and paid contents as well as other tutorials are also available along with this 5000 pages. So the support for the commercial software is huge. So where to begin or how to begin. So open form in the olden days also they used to conduct workshop yearly and workshop materials will be available free of free to all. And if you can refer this workshop materials to begin with and Chama University CFD with open form is a course by Professor Hakan Ilsan. So the course assignments are also available online. And from India we have a very good group from IIT Bombay that is 4C project. I hope we already pile matter give introduction about 4C initiatives and along with that training companies like some of the websites from Wolf Dynamics CFD Naina they also provided. They are also providing you tutorials to own open form and some YouTube channels like Joseph Nagy and other contributors. So I would like to appreciate the effort by 4C team that is mainly the attempts from India. India we know that there is a group for 4C it's supported by National Mission of Education through ICT. So basically they are actually promoting the open source software for last 10 to 12 years. So there are multiple opportunities for students those who are learning such as case study projects, research migration projects, etc. So through these materials you can you will also get an opportunity to work with open form as well as you will also get a certificate from 4C SLS IIT Bombay. So those things are already explained along with that explained by a pile matter. So along with that one interesting part is along with the case studies. So if you are working on some research problem, okay, let us say you are working on a flow plus blood body. So it is very difficult to find the resource material in all the days. Like in earlier days it is very difficult to find the sample case file or sample document or documentation for a flow plus blood body or some any fundamental problem. But the interesting thing is how in 4C case study initiative you have around I think 150 or 100 plus sample cases along with the documentation. So if you wanted to explore some let's say EV thermal management system some sample cases are available in their repository. You can just download it and install it along with they are also neatly documented it. So as a beginner it will be very, very helpful for you to refer the case study projects available in the 4C website. Okay, so moving on to the next topic. So we use command line interface mostly in open form rather than a GUI. Okay, so command line interface as you know already so probably you might also not familiar with open form or not familiar with open form. It will be very, very irritating to type all these commands each and every time, right? So if you type dog mesh, if your M is not capital it will give some error. So what is our advantage or why we are still working on this ordinary idea of command line interface rather than GUI? Any thoughts? Sir, customized modification is possible sir. Yes, so let me show you a simple example. So when I attended the previous session, so there was questions from students where is this P file? Where is this U file or where I should mention this NU that is Dynamics Excosity or Kinematic Excosity values etc, right? So basically, I hope you might be knowing open form is certain open form sample case. It contains certain number of files, right? So there is a difference in the open form version I use. I am using a software known as Blue CFD to install open form in my PC. I felt this is very convenient. You can try this after the workshop because workshop everything they plan based on WSL or Ubuntu virtual box. Blue CFD is a third party software which is free to all and very convenient to work with Windows system. You don't need to install separate WSL box or anything for using open form. So and also you don't need to type CPCD all those commands every time. So let me just show you the application of command line. So I just right click and open Blue CFD terminal. I will get this terminal. You can explore this after the workshop because in the workshop everything we plan based on WSL or virtual box. Okay. So let us say you already run this cavity tutorial and you don't know there are multiple folders, zero folders, constant folders, system folders etc, right? So zero folder there is some more files P, U, etc. And constant folder you have Polymesh, transport properties etc. And system folder there are multiple folders. So it is very, very difficult for us to identify. In which file if you are, let's say in which file you need to give the kinematic exclusive. So if you do with GUI what we need to do we just need to open each and every file and just type control and find NU, right? So instead of that I will just demonstrate the application of terminal commands or command line interface. I will just type a command GREP-R then type NU. So GREP-R basically search the commands, search a keyword followed by this GREP that is NU in the files. So right now my folder is a posi and cavity. So when I just type GREP-R then NU you can clearly mention I should go to console then transport properties. There I will find the NU line. Okay. So this is just an example. So for time being you can explore this. I will just copy paste this in the chat box. You can also GREP if you want to just type zero boundary condition or yes zero gradient. So GREP-R then zero gradient. So these are the two files, two situations in which we use zero gradient. And first one is P itself. P itself there are two lines in which we use zero gradient. Okay. So this is very, very convenient. So right now I just showed a simple example. So let us say you don't know some of the special boundary conditions available in open form and you don't know which tutorial they considered. So you just go to the open form tutorial folder. Just type zero gradient. So I will just type example. Yes. So I just opened the tutorial folder as such. So if tutorial folder we have multiple solvers basic to all stress analysis. I just wanted to know in any case they use a boundary condition called cyclic boundary condition. So I just type GREP space hyphen R just type cyclic. So I will get all those corresponding files. I can just go to these. These stories if I want to explore more. So let us say I just wanted to know only incompressible then CD incompressible then type GREP. So incompressible itself I can find there are multiple folders or multiple tutorials they explore the cyclic boundary condition. So this I just wanted to demonstrate the application of command line interface. Initially it may be irritating and later stage you will start loving it. Okay. So before we move on so when we learn some new software you should also know why we need to spend or why we need to invest this much of time and time for learning a software and whether it will make us an M whether this adding or learning this open form till whether it is beneficial for job or industry. So these are some of the companies multibillion multinational companies in which they started using open form as their CFD packages along with the commercial software like Boeing, Pfizer, GM, AMD, 3M, etc. and there are also multiple startups along with this multibillion multibillion multinational companies there are certain startups also like ESA, Total Sim, SimScale, Pandav, CFD, etc. They are also working on open form based simulation. So soon you can start seeing that industry slowly moving towards open source software. So learning this software is actually an additional skill and that will make you employable once you complete your course. And another thing I said is lack of GUI. So there are certain startup companies they started working on developing dedicated GUI for open form. Some of them are Simscape, then Bluecape, Simflow, Helix, Simples, etc. and I also learned that OC team is also developing a GUI and there is a session by OC team I think tomorrow I am also eager to attend that session. So these are some of the difficulties faced by the open form community and these companies are actually helping us to resolve those difficulties and promoting the open form software. Yes. So let me move on to the technical part. So we use computational fluid dynamics for doing or for analyzing something. So there are certain things we should clearly follow when we start our computation. First one is experimental validation. So we know that we are actually using some sort of governing equations and we are applying it to physical problems and we are solving it using computers. So it is quite obvious that we may get incorrect results. So if you are doing some mistakes in the boundary condition, so some mistakes in the solver setting we will get incorrect answers. So the first part when we do our CFD simulation is we should compare our results with the experimental results. If you have our own experimental facility, it is fine. We can compare the results with us or we can just compare the results with the available results in the literature. So this particular figure shows the comparison of CFD code application nuclear safety. So experiment versus simulation. You can find the liquid air water interface for the simulation from the experiment is in this way. And you can also note that for simulation it is more or less safe. So this will make sure that our computations are correct. So our results are correct. So if you are not doing the experimental validation, so if you are submitting some paper somewhere or if you are presenting somewhere, so the first question they will ask is how can we make sure that our results are correct? So that is first part. And second part is grid independence. So as I said earlier, we are dividing the domain into certain number of cells, right? So we know that if we increase the number of cells it will take more time for completing the results. So we need to make sure that our results are independent of the grid. So there is some critical number in which our results will be independent of grid. So I will explain that in the subsequent slide. And the third case or third part is time step independence. So when we do unsteady simulation, those who already did unsteady simulation, they may be knowing we will be giving some DT value or time step value for completing the simulation. And the DT value, selection of DT value is also very very important. We may get the results if we take DT value that is time step as 0.1 and we may get a different result if we take time step as 0.01. So we need to make sure that the time step independent results are obtained when we do the simulation. These are some of the resources available for experimental validation. Moving experiment is very very closely assessed. So there are certain websites initiated by different communities. Like NASA, they actually shared some of the interesting experimental results and anyone can download and make use of it. So for overview validation and verification, they have certain websites. As well as some test cases as well as benchmark cases are available for basic flow as well as heat transfer problems and multi-phase problems. So these are some of the repositories available to all and we can make use of these repositories for validation. Some of them are like AirCoffee database, NASA, the new nursery manager, etc. So let me move on to grid independence. So what do we mean by grid independence? I said earlier that we are dividing the domain into certain number of cells, right? So we need to make sure that these number of cells are independent of, the results are independent of number of cells. So I would like to explain this with an analogy. So this is not related to CFD but numerical method. So let us say I want to draw a circuit, okay? So right now I will just start with only 4 points and I am just going to connect it with a straight line, okay? So the second case is I will just increase the number of points. Instead of 4, I will keep 6, 8. And again I will connect this with straight line. And third case, it will be obvious to you now. I will increase it. So from 8, I will just increase it to 16. So 4, 8, 16. So I think picture is more, you know, it is more clear to you what is going to happen. So we are closely moving towards the circle for rectangular, right? So when we increase it to 32, I hope it is clear to you what is going to happen. So I am just, I am not repeating it. So in this particular example, what we can see is when we increase the number of cells, we can actually move towards our required geometry, right? So I will just go into tabulate it. So the number of points is just 4. And I am also calculating the area. So we know that area 4, let us say it is a one unit circle. So area for that circular circle is non. So A minus A dash. A dash is a required area. So basically it gives the error between the obtained value and the actual value. So 4, 8, 16, 32, etc. So we will calculate A by A value and we will find the error. So it is very obvious that what happened. If you just plot this in the n versus error or percentage error. Percentage error means you will just divide it with the reference into 100. So what happens? It is slowly moving towards 0, right? So there is a point in which if you take n is equal to say 64 and 124, the percentage error will be less than 1 percentage or less than 2 percentage. Okay. So what we can say? The geometry or the area we are going to calculate is independent of the number of points. We cannot completely say independent. It is well below 1 percentage or 0.5 percentage. We can say in that way, right? So this is the brief overview of the grid independent solution. So when we do numerical simulation. So this in this particular case, we clearly know that the final value, that is A zero value is very like non true. But in actual flow problems, we may not be knowing the final solution. So in that scenario what we do? We will just start with an arbitrary grid. Let's say G1. So let's say this is 100, then 200 or 400 or 1000. Then what we do? Let's say if you are considering a flow problem, we will just explore or we will just extract the velocity profile at certain plane. And we will just compare for different grid. And what we can say is once we increase the grid, the error, percentage error between G1, G2, G3 reduces. And if it is less than 1 percentage, we can say that since the percentage error between G3 and G4 is less than 1 percentage, and we are selecting this particular grid for doing subsequent simulation. So this is the first part. The first part is validation as I said earlier. The second part is we need to make sure that our results are independent of reference. So grid independence is the second step we should follow when we do a systematic computer or scientific computing. This is just another example. So here they made a rabbit in some element. So you can see the number of elements. When we increase the number of elements, the rabbit will become more and clear and clear. So this is just an example, not from CFD, the finite element or that. So, yes. So this is an example from our cavity problem. So when we do a coarse mesh, we may not get an accurate result, but when we have a medium mesh or a fine mesh, you can see the medium mesh and fine mesh. There is some similarity. So if you go with a very fine mesh, you will see the difference between the fine mesh and very fine mesh will be very less. So in computer fluid dynamics, if you want to create some simulations, we should follow some steps. The first one is let's say a physical problem, right? So we need to conduct, we need to make a geometry. So if it is a lead-driven cavity problem, we need to make a regular job. Then after that, what we need to do is we should generate mesh. So once we generate mesh, what we need to do? We need to prepare our case. We should clearly mention the package or see if it's open for more answers, what are the boundary conditions, whether it is laminar flow or turbulent flow, or if it is a heat transfer problem or mass transfer is also there or not. So we need to prepare the cases. Then we will be doing the simulation. That is the third part, that is running cases. And finally, the post-processing of the results. So we use a paraform for post-processing in the results. So the right hand side is an animation made using a blender. Blender is a software, open source software widely used for animation software. So nowadays people are using paraform results along with the blender to get the more realistic animation. So you can clearly see it is far better than the paraform results, right? So these are additional things in which you can explore or visualise your results. So this particular figure just show a mesh corresponding to ship. Okay, then we are moving towards mesh generation. So as I said earlier, basically we are dividing the domain into certain number of cells. Like that particular process is known as machine or mesh generation. So this is a simple case rectangular block. We are just dividing it into certain number of cells. Next direction, y direction, z direction. Just dividing into let's say 20, 20, 20. So we are generating the mesh. But in actual case, our geometry will not be very smooth or very regular like this. So we may have some carvel in your faces. We may have some protrusion somewhere. So the machine, you cannot always get this rectangular mesh or rectangular mesh always. So there are multiple types of mesh elements. So some of them are tetrahedron, hexagon, prism, pyramid, bold, et cetera. So these are some of the mesh cell elements or mesh elements we usually use in our simulations. So in order to do meshing, what are the things required? So you know that when we make a geometry, we should know first thing is we should know the vertices. And these vertices will be connected by lines, right? And these lines will be connected by surface. And these surfaces will be finally, it will make a cell volume, right? So this is the hierarchy we use for mesh generation. So we can classify generally, we can classify different measures for mesh generation method. The first one is structured grid. The second one is unstructured grid. So structured grid as the name indicates, it is well structured grid. So for this figure shows a typical example for a structured grid. The grid points are placed at the intersection of coordinate lines. And indirect grid points are fixed number of neighboring grid points. What do you mean by that? I will explain in the later stage. So, and grid points can be mapped in the matrix. They are located in the grid structure in the matrix. Metrics are integrated by IJ. So those are some indices. So let us say I wanted to focus on this particular grid. And I can easily find who are the neighbors, right? So if I just consider that, I don't know whether you are familiar with this notation. If it is a two-dimensional one, the neighbor is going to be I plus 1J, I minus 1J. And the top and bottom IJ plus 1 and IJ minus 1, right? So it is very easy for us to identify the neighbors for this particular cell. So this is very, very important when we discretize the equation and apply it. And some structural grid itself, we can have a uniform mesh and non-uniform mesh. So non-uniform mesh, there are some cases we may need very fine, refined mesh near the wall or near some interfaces. So we may use different cell height and those kind of cells are known as non-uniform mesh and the equally spaced cells are known as uniform mesh. And it is not necessary that we can, we should have a rectangular or orthogonal purely orthogonal grid. You can also have a slightly modified non orthogonal grid also. Okay. So in the subsequent slide, I will explain some parameters to check the quality of the mesh. And regarding the unstructured grid, as compared to structured grid, unstructured grid, it can have under volume with any shape. And there are no restrictions to number of adjacent cells. So in earlier case, we saw that for a two-dimensional case, we have only left-hand side boundary, right-hand side boundary, top boundary, neighbor and bottom. Only four cells are surrounded. On the other hand, for an unstructured grid, we can have multiple, and there is multiple cells surrounded by the cell and there is no restriction on the number of adjacent cells meeting the points or along the lines for a 3D case. So in practical CFD, typically triangles or quadrilaterals are mostly used for 2D problems and heterohydrants or hexahedron elements are used for 3D ones. So if we ask why we have different measures, we should have only unstructured or structured. So there are some advantages and disadvantages. First one is if you have a complex geometry, typically unstructured grid generation is very easy as well as faster as compared to structured grid generation. And in terms of accuracy, simple geometries, structured grids are more accurate compared to unstructured in general. And complex geometries, unstructured grids are more accurate because for structured grids, the cell quality will not be that good. And if you see in terms of computational time, so convergence, CPU time, structured grid actually takes less time rather than the unstructured grid. So what might be the reason? So the main reason is for a structured grid, if I say this particular grid, I don't need to clearly explain who are the neighbors. So it's obvious that it is i minus 1 j, i j plus 1, plus 1 j and i j minus 1. On the other hand, so if you just control an unstructured grid, there is no particular pattern. So the grid neighbors are 9, 1, 13 for a cell 0. For 6, it is going to be 7, 5, 1. So there is no pattern. So in fact, if you are using unstructured grids, we should have an additional data storage for identifying the neighbors. So for solving, the neighbor states must be known. And structured grids, these neighbors found by adding or subtracting 1 from the skill indices. So you don't need to have a separate memory for that. But for an unstructured grid, it requires storage of cells to cells, pointers, etc. So more storage that leads to lower execution of the project. So that is one of the disadvantage of the unstructured grid. So the next part is quality of grid. So let's say you are making some grids. So how do we make sure that I was really sufficient or whether that will give a correct result or not? So mesh quality decides the solution as well as accuracy. Solution accuracy as well as convergence of the simulation. And basically we use two parameters. First one is skewness and second one is aspect ratio. So skewness is nothing but the angle between two edges. So you can see in this particular case. So we have a line here, we have a line here. The angle between grid lines x and y is indicated by skewness. So here it is indicated by theta. And if you have unstructured grid, unstructured grid, so what we do is we just compare what if we have a base geometry that is equivalent to triangle and we have skewed triangle. So some ratio between these. So minimal cell skewness for better solutions. So there are certain conditions. So minimal cell skewness or minimal changes in the cell between the adjacent cells and aspect ratio. So aspect ratio is nothing but the ratio between the edges. You can see rectangular geometry, triangular geometry and external geometry. These are the regular shape with aspect ratio 1. On the other hand, the second case is a high aspect ratio cell. So it is skewed and the aspect ratio is slightly different or entirely different. So suitable aspect ratio for cell selection is 0.2 to 5. And if it is more than that, so that may lead to error or that may lead to some convergence issue, etc. And sufficient resolution in the region of rapid flow changes in regions with sharp gradients or boundaries. So as I said earlier, when we do refinement, it is also possible we can give fine mesh not only near the wall but also in the areas in which we are expecting sharp gradients. For example, between liquid-liquid interfaces or liquid-air interfaces we can give refinement. So moving on to the open form part. For mesh generation in open form, we basically have two utilities. So first one is block mesh utility. The session followed by this session, by my session, there is a separate tutorial for block mesh. So I am not going to detail for the block mesh as the snappy hex mesh part. I will just quickly give you an overview. The block mesh we generally use for machine symbol geometries like we show some examples in the upcoming slides. Snappy hex mesh is an advanced feature available in open form for doing slightly advanced or complicated geometries. So the mesh generation using block mesh is very simple. When we have a geometry first thing is we should know the points. Points will be connected by lines, it will be connected by blocks and we need to identify the boundaries. So more details regarding the block mesh will be explained in the upcoming slides. I am not going into detail. So to some extent we can also use block mesh for making complex geometries such as if you want to mesh a cylinder or if you want to mesh a section in this way. This is also possible with the help of multiple blocks. So instead of using a single block we use multiple blocks. As you can see these particular grids are known as L grid. So we are just dividing the node semicircle one by fourth of the circle into multiple blocks and then we are applying the block mesh. So there are some examples. So another example is using block mesh. So this is a cylinder in which we have an O grid available here. So this is an enlarged view. This is an isometric view. This is a top view and this is enlarged view corresponding to the geometry. And snappy hex mesh, there is an adhesion. The session followed by me, there is a session by the editor. He will be talking more about the snappy hex mesh standard. So when we have a complex geometry, it is also possible to do meshing in open performance using the snappy hex mesh utility. So this is an example. We need to model a car in a chamber or something. So this is a process in which we use snappy hex mesh for meshing. So is that all? No. So it is also possible to use third party software for meshing in an open form. So some of the open source software for meshing and salo, G mesh, etc. So these are the software which is having GUI. So you can also make geometry as well as mesh using these open source software or commercial software such as ANSI's ICM CFD or hyper mesh, etc. And we can save that meshes into some open form readable format. Let's say for an example, you can save those meshes as .msh file. So there is an option to export the mesh. So right, there is an option to export the mesh to .msh file or some other format in which we can easily convert that mesh file which is made by some other software like third party software and use it in open form. I will just show you an example. So how is it possible? You just type it through on command. So I will just show you an example. Let's say I made a geometry and mesh by using a software ICM. Or I exported a mesh into .msh file. And I want to use it for open form. So you don't need to actually use block mesh. You already made the mesh file in .msh file. You just need to convert that into open form, open form readable format. Then you just type ICO form. If it is a cavity problem, just type ICO form. That is solved only. So we can just convert that by using a simple command flow and mesh to form followed by the file name. I will show you a demo also in the subsequent slide. So if you just type flow and mesh to form followed by .msh file, it will convert that mesh file into open form readable format. So I would prefer showing a demo now just a second. So I just took a tutorial on Elbow. You will find this tutorial in the open form. So you can see I just copy pasted a tutorial and it's now Elbow. You can find it later. So the mesh file is provided here. Elbow.msh is already provided. So I don't need to make the mesh again using block mesh. So what I will do is I will just open my terminal. Okay. I am just going to type flow and mesh to form followed by name Elbow.msh. We just press enter. Some commands are displayed and finally there are some error which will show some error. Otherwise it will show n. Okay. So I just converted .msh file into open form readable format and I am just going to visualize it using para for para view. If I take some time in my PC. Yes, I got the mesh and you can see the mesh. So these are not done by block mesh or snappy mesh. I just imported the .msh file and I just opened it in para for para view. Right. So it is also possible to do that party software generate mesh if you are comfortable with any software let's say GUI if you are comfortable with any software GUI then it is possible to make mesh there and import it here. Okay. And the last part is very, very interesting part. How far we just considered only those cases or measures with fixed sub boundaries. So in actual cases you will also experience measures that are moving with respect to type. There is a simple example is turbine blade. We know that the turbine blade rotates right. The solid body or solid boundary is rotating with respect to type. So how to tackle those kind of situations. We need to do re-meshing every time. The normal CMD simulations with solid or constant ball we just need to do machine only once in a time once in a while before the simulation. But when we have moving mesh there are different approaches. When we have moving boundary so we need to do the machine in a very time step. So simple example some examples are shown here. So this is some like turbine blade kind of thing. There are four objects and it is rotating. So it is rotating collectively also it is rotating. You can see the shape is this part. If you just follow one of them you can see it is also rotating with its own axis and it is also rotating with the common axis. Right. So we need to do machine in time step if you want to see the simulation or if you want to see every simulation. So these kinds of measures are known as dynamic machine which is sometimes giving a brief overview about dynamic machine upcoming session. And the second case this is another interesting part here we have a body solid body is moving left to right or it is oscillating left to right. Right. So some when it is moving towards right the measures in this side is compressing and it is enlarging and vice versa. So these are some of the challenging problems or problems we can expect when we do moving boundaries or moving body. Okay. So with this I would like to end my session. Thank you. Any doubt? Sir only one doubt it is very simple sir. Sir how can do time step independent steps sir. Yes. So we need to start with an arbitrary time value, dt value. So generally what we do is we will refer similar problems similar problems done in the literature or we will refer some standard literature and we will find the approximate range of dt value. So let us say somebody took for similar problem like us somebody took 0.001 as their time step value. We will just take 0.001 as their time step value then 0.4001 and 0.01 etc. So this is a trial and error method and we may need to compare one of the parameters like sprawl numbers or some frequencies in order to know whether the parameters are independent of time steps. Is that clear? Sir for that first of all we do the grid independence then time step right sir? Yes. So for time yes grid if it is an uncertainty problem what we do is we will just take a very arbitrary dt value and do the grid independence first then with that grid we can do time independent. So if we have an experimental result with you to compare always better to compare that result then we can say we can identify which one is more accurate right? Hello sir? Sir I want to ask one question regarding like if there is a complex problem for say like there is an artery which is moving and blood is flowing inside that so which software will be preferred like commercial like ANSYS fluent or open form like which one will be having edge or which can do better like in terms of accuracy and for deformable bodies like if artery is deformable. So it depends upon us actually so people are actually using NAPI-X1 also or some open source software like dialog and some of them are actually using ICM so it's all about how you make the measure. If you are able to make measure with good quality there are second of the software you will be getting that. Ok so it's all about machine accuracy efficient. Ok thank you. Sir? Sir can I do sir coupling type of problem in open form sir? Yes it's possible. Yes I Yes fluid structure interacting is possible. Ok. Actually yes one point I forward to mention so open form is an open form software. There are different groups they modify, they tweak the open form version according to their requirement. We know that in Android it's the companies Google provide Android but for different manufacturers they fine tune their android according to them like Nokia or Xiaomi similarly some companies fine tune this open form version and they will slightly improve it. So FSA I think form extend there is a version of open form and it's form extend I think they are having a dedicated software FSA please do just find it. I'm not sure about form extend there is a dedicated piece of open form version for this particular FSA. You can just find it out. Ok sir thank you. Yes it is possible machine learning is also possible now people are exploring machine learning for peer-to-peer analysis. Yes. I have one question and Dr. Hari Krishna. Yes. So this unstructured mesh like is it determined by the complexity of the geometry or the complexity of the flow? Normally it is complexity of the geometry and the effort required for machine. It is also possible for the same problem we can do with structured mesh and unstructured mesh. But if the geometry is very very complicated let's say car there are multiple objects inside it so making structured mesh will be very very difficult but it may take maybe 10 to 15 days for only machine for structured mesh on the other hand it may take only one day for making unstructured mesh we just say that these are the boundaries in which we need refinement we can just keep working so within less time we can make unstructured mesh. Okay. Thank you. Yes. For geometry we can define the geometry and also we have to choose the geometry for about 14 thickness for some micrometer thickness thickness. We have to define the surface of the geometry so we have to define the 14 thickness of 5 micrometer. Can we define it in the open room? Can you repeat it again what thickness I will engage you? 5 micrometer. Yes, what geometry is used? For choosing an air pipe. Yes, it is possible Yes. So we can get the solution of some nozzle number for the thickness and also for the surface. This was I didn't include airfoil but for a simple rectangular geometry it is possible you can just give an only uniform mesh with a given height. You can define. Yes. You can get the heat transfer equation and you can get the heat transfer results of both the surface. Yes. Okay. Let me check. Okay. Sir, one question sir. Sir, in case of a structured grid suppose we have complex geometry like bifurcation region or any other where the flow of physics changes very much sir. So at that time sir in case of a structure we have to refine that area or it is good sir only for a structured grid. Non-uniform you can give in a structured mesh also. We generally what we do is wherever we can expect deeper gradient near the boundaries or some interfaces we can refine that. Yes. We actually refine in case of a structured grid but in restructure Yes, it is possible. One figure I showed non-uniform mesh grid figure I showed is actually for structure. Yes. And one more doubt sir. What is the quality? Sometimes we use orthogonality or like that. When orthogonality is near to one it is considered to be good. Yes. So what type of orthogonality we use sir? Sometimes it shows a minimum average maximum three types of result. Yes. Which one we consider sir? It should be in a range. For orthogonality range of 0.2 to 5. Someday when we have a very poor mesh we will have minimum value of 10 raised to minus 3 and maximum value of 10 raised to 5. So that we do not prefer. So we consider only minimum range. No, I am saying it should be in the same order of magnitude minimum maximum average or almost in the same order of magnitude. Oh. You understood? So 0.2, 1 and 5 in the same order of magnitude, right? But 10 raised to minus 3, 1 and 10 raised to 5. There is a order of magnitude difference of 10 raised to 8 from minimum to maximum. So that may give error when we do the thing. Okay sir. Okay sir, thank you. Yeah. Okay, if there are no doubt then I will end. Okay, thank you very much.