I really don't believe that ChatGPT can be of some help in producing efficient answers about turbulence model.

I tried to ask some math and physics question and often received wrong answers.

It is based on your particular requirements sometimes we also understand our turbulence model requirements after some time of doing numerical experimentations. So, it can tell you some general models but not optimized for your concerned problem. This kind of decision sometimes leads to excessive use of computational power, which is complete waste.

Dear Güven,

I perceive that your simulation might possess different states, including heat fluxes, temperatures, Reynolds Numbers, fluid types, etc. I strongly recommend you find related papers which are similar to your work and choose the turbulence model based on them. You might ask what if I find three different turbulence models? My answer is you can test which one is near to your experimental outcome or analytical approach.

Best,

If ChatGPT answer is based on real experimental results, it can be trusted.

I think the quality of the ChatGPT answer depends more on the question being asked, as chat is a linguistic model. It works exactly with the statements that are loaded into it. It cannot think in the sense we are accustomed to, it only looks for matches with the question in base.

I tried to ask the GPT chat implemented in ora.ai about which turbulence model to use for the example https://www.comsol.com/model/airflow-over-an-ahmed-body-8565. I described the dimensions of the body, the features of the geometry, the material of the medium and the speed value. The example suggests using k-epsilon model. But GPT chat suggested using k-omega SST model, beacuase "This model combines the advantages of the k-epsilon and k-omega models and is well suited for a wide range of applications. SST k-omega provides more accurate prediction of boundary layer behavior and flow separation by taking into account stress transfer effects. This is especially useful if you have geometric features like your object has: legs and a beveled top."

Since the model in the example is simple, the answer, in my opinion, is correct. Perhaps the GPT chat will be able to suggest models for complex cases, for example, for turbine machines. But it depends on how you write the query and on the database used by one or another GPT model. But mine is easier to find articles with similar tasks or reviews.

ChatGPT shouldn't be used to select turbulence, in my opinion. For the selection of the model, there is additional crucial information.

First question: where could have ChatGPT taken the relative information for training? Anyone who is going to use that might want to know about it.

Second question: why ChatGPT? This is not some complex information we can't extract from complicated data. It's just exactly there for anyone to learn or to search trough. Also, it could be better done by making a new net from scratch just about that, than using a LLM that really has nothing to do with this.

My general comment is that engineers and scientists should not look for answers in a tool that is not even tought to be what they want. ChatGPT just spits out words following a prompt and a statistical model. That is not different from a bad student that just goes by memory without understanding. ChatGPT is actually worst, as it never assumes it could not know something. I don't honestly know who could want to ask ChatGPT about turbulence models here.