How to tinker and verify convectional models

StudentinANSYS · November 7, 2017, 07:12

Hello! Here follows a brief project explanation.

We are supposed to develop a methodology in how you can control cooling rates in a nickel based AM alloy. Depending on different cooling rates in the range of 1000 °C and about 700 °C, the material structure in terms of certain particle sizes are changed (this is outside my area). We want to map this relation to the cooling rates and our plan is:

Heat the sample to desired temperature and use a big volume of a copper block, a cylinder about 100 time bigger volume than the nickel sample.

With a IR-camera provided by a company related to our project we will measure one side of the sample as it cools down. We will make this side sooty to be able to estimate the emissivity (which is another great uncertainty).

The goal is to verify our simulations and export the simulated results in terms of cooling rates over the sample volume to try to relate it to particle size later on.

This picture illustrates the temperature difference between two time steps in sims. It is somewhat a bit sketchy to us.
https://imgur.com/nNVjw9V

We need to measure the entire side of the sample to know different temperatures over the sample side to simulate a cooling rate gradient.

Now to my question. We are using a number of standard convectional models taken from J.P Holman literature provided to us in another course. How are we going to be able to verify the convectional models and how should we be tinkering them to aquire reliable simulations. I could imagine verifying convectional heat transfer as whole but when we have different models to vertical and horizontal planes and cylindric surfaces, how do we allocate the convectional heat transfer and where do we tinker the models?

What is a good approach to this problem?

I'm happy to explain more if that is required to answer!

Thankful for everything!

/Jakob

ghorrocks · November 7, 2017, 17:57

I have done lots of work in a past life looking at the heat treatment of steels. When steel cools down it goes through a number of phase transformations, and each of these phase transformations result in heat being released and absorbed. This means that you cannot model the cool down of steel as a simple thermal mass + heat transfer model, the phase transformation heat is significant enough that it needs to be included as well.

I do not know whether that is important in your case, but you mention what appears to be grain size and that suggests you are looking at phase transformations in your alloy as well. In that case you better look into the heat released and absorbed as well, as it is likely these are very important.

Another point: Your simulation results look strange. Have you verified that your simulation is accurate? Your results are what I would expect from too large a time step. You need to confirm your mesh, time step and convergence criteria are suitable for what you want to do if you want reliable results.

And finally to answer your question: I do not understand your question. Are you asking how do you implement a custom convection model, or are you asking how do you adjust a model you have already applied?

November 7, 2017, 07:12	How to tinker and verify convectional models	#1
StudentinANSYS New Member Jakob Lindqvist Join Date: Dec 2016 Posts: 9 Rep Power: 9	Hello! Here follows a brief project explanation. We are supposed to develop a methodology in how you can control cooling rates in a nickel based AM alloy. Depending on different cooling rates in the range of 1000 °C and about 700 °C, the material structure in terms of certain particle sizes are changed (this is outside my area). We want to map this relation to the cooling rates and our plan is: Heat the sample to desired temperature and use a big volume of a copper block, a cylinder about 100 time bigger volume than the nickel sample. With a IR-camera provided by a company related to our project we will measure one side of the sample as it cools down. We will make this side sooty to be able to estimate the emissivity (which is another great uncertainty). The goal is to verify our simulations and export the simulated results in terms of cooling rates over the sample volume to try to relate it to particle size later on. This picture illustrates the temperature difference between two time steps in sims. It is somewhat a bit sketchy to us. https://imgur.com/nNVjw9V We need to measure the entire side of the sample to know different temperatures over the sample side to simulate a cooling rate gradient. Now to my question. We are using a number of standard convectional models taken from J.P Holman literature provided to us in another course. How are we going to be able to verify the convectional models and how should we be tinkering them to aquire reliable simulations. I could imagine verifying convectional heat transfer as whole but when we have different models to vertical and horizontal planes and cylindric surfaces, how do we allocate the convectional heat transfer and where do we tinker the models? What is a good approach to this problem? I'm happy to explain more if that is required to answer! Thankful for everything! /Jakob

November 7, 2017, 17:57		#2
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,857 Rep Power: 144	I have done lots of work in a past life looking at the heat treatment of steels. When steel cools down it goes through a number of phase transformations, and each of these phase transformations result in heat being released and absorbed. This means that you cannot model the cool down of steel as a simple thermal mass + heat transfer model, the phase transformation heat is significant enough that it needs to be included as well. I do not know whether that is important in your case, but you mention what appears to be grain size and that suggests you are looking at phase transformations in your alloy as well. In that case you better look into the heat released and absorbed as well, as it is likely these are very important. Another point: Your simulation results look strange. Have you verified that your simulation is accurate? Your results are what I would expect from too large a time step. You need to confirm your mesh, time step and convergence criteria are suitable for what you want to do if you want reliable results. And finally to answer your question: I do not understand your question. Are you asking how do you implement a custom convection model, or are you asking how do you adjust a model you have already applied?