[Weiqiang Liu]

hello all,

this is not a very relevant question to fluent. Instead it's a question about diesel fuel mechanism. I am doing a diesel fuel combustion simulation with fluent. I tried to import detailed mechanism of diesel into fluent.

however I am hesitating in choosing what kind of diesel mechanism before the simulation. I read some literature and usually diesel surrogate is used to represent real diesel fuel. The diesel surrogate would include normally n-dodecane, cycloalkane and aromatic. The mechanism of this kind of surrogate would be very complicated which includes hundreds of species and thousands of reactions.

there are some other diesel mechanism which includes only one species like n-dodecane or decane. These mechanisms are much simplified with dozens of species and hundreds of reactions.

all the authors in the literature claimed that their mechanisms were well-validated mechanisms no matter they were surrogate or single species.

I am wondering what kind of mechanism I should choose or can anybody explain why there would exist so many kinds of diesel mechanisms. which one is more accurate? what contributes to those mechanisms' accuracy?

thanks very much

weiqiang

[LuckyTran]

Welcome to the world of chemical kinetics modeling.

So of course there can be thousands of reactions in a any particular detailed chemical process. Diesel combustion is quite complicated due to various crap compounds that can be found in a diesel mixture (it is the bottom of the barrel after all). For any particular chemical process, there are yet thousands upon thousands more of simplified surrogate models.

Everyone (self-advertises) and always claims their model performs better than the full mechanism. What better means is conveniently subjective.

There is no right answer. From the start, how do you define accuracy? You can't just look it up in a dictionary what accurate means.

Which one should you choose? Whichever one you like. You are the modeler. It's like asking what two numbers should I multiply together to find a product? Well, any two that you like. Unless there is someone telling you to do a particular thing (i.e. figure out what two positive integers yields a product of 6, then you'd take 3 and 2).

[Weiqiang Liu]

Originally Posted by LuckyTran

Welcome to the world of chemical kinetics modeling.

So of course there can be thousands of reactions in a any particular detailed chemical process. Diesel combustion is quite complicated due to various crap compounds that can be found in a diesel mixture (it is the bottom of the barrel after all). For any particular chemical process, there are yet thousands upon thousands more of simplified surrogate models.

Everyone (self-advertises) and always claims their model performs better than the full mechanism. What better means is conveniently subjective.

There is no right answer. From the start, how do you define accuracy? You can't just look it up in a dictionary what accurate means.

Which one should you choose? Whichever one you like. You are the modeler. It's like asking what two numbers should I multiply together to find a product? Well, any two that you like. Unless there is someone telling you to do a particular thing (i.e. figure out what two positive integers yields a product of 6, then you'd take 3 and 2).

ok I get it. some authors use one species to represent diesel and some other use surrogate to represent diesel. In general, which way is more accurate? In my opinion, the surrogate is more accurate which is also more complicated, right?

[LuckyTran]

Surrogate models are simplified models of the complete reaction. I completely disagree that they are more complicated or more accurate. If you want generality, then use the best known reaction mechanism or something like it and not a surrogate model.


But surrogate models are a victim of their own creation. They are tuned to be more accurate artificially by producing a specific result a result that not even the best known kinetic mechanism can produce. But again, this is all due to tuning. Sometimes surrogate models involve species that don't even exist in the original mixture, sometimes surrogate models contain species that don't actually exist at all (hypothetical compounds). If you define more accurate as producing a result that you desire, then it is a self-fulfilling prophecy and no one can ever say you are wrong. And that's how surrogate models survive.

[Weiqiang Liu]

Originally Posted by LuckyTran

Surrogate models are simplified models of the complete reaction. I completely disagree that they are more complicated or more accurate. If you want generality, then use the best known reaction mechanism or something like it and not a surrogate model.


But surrogate models are a victim of their own creation. They are tuned to be more accurate artificially by producing a specific result a result that not even the best known kinetic mechanism can produce. But again, this is all due to tuning. Sometimes surrogate models involve species that don't even exist in the original mixture, sometimes surrogate models contain species that don't actually exist at all (hypothetical compounds). If you define more accurate as producing a result that you desire, then it is a self-fulfilling prophecy and no one can ever say you are wrong. And that's how surrogate models survive.

so you mean surrogate mechanisms are tuned to predict parameters in specific situations. When situations change, these surrogate mechanisms might not be accurate or even right any more.

What I need to do is to choose the similar situation the mechanism is tuned to my own case.

is it right?

[LuckyTran]

Originally Posted by Weiqiang Liu

What I need to do is to choose the similar situation the mechanism is tuned to my own case.

It has less to do with the situation than what the model is actually trying to predict. You need to understand what objectives were selected for each particular mechanism. For example, one mechanism might be tuned to give super accurate bulk properties but wrong in everything else. Another mechanism might be tuned to give super accurate predictions of concentration of bananas and not care at all what the concentration of elephants are in the room. What does each model do and what are you trying to get out of it?

[Weiqiang Liu]

Originally Posted by LuckyTran

It has less to do with the situation than what the model is actually trying to predict. You need to understand what objectives were selected for each particular mechanism. For example, one mechanism might be tuned to give super accurate bulk properties but wrong in everything else. Another mechanism might be tuned to give super accurate predictions of concentration of bananas and not care at all what the concentration of elephants are in the room. What does each model do and what are you trying to get out of it?

for my case, I just need to know the outlet temperature of a flow reactor in which diesel is oxidized. so outlet temperature belongs to bulk properties, right?
what I need to do is to choose the diesel mechanism that predicts flow reactor temperature well?