Temperature-dependent sensitivity analysis

[jeffrey...@gmail.com]

Does cantera have a function that can find the sensitivity of an observable (e.g. flame speed) to reaction rates as a function of temperature? In other words, can cantera perturb reaction rates in narrow temperature windows to find the effect on a global observable? Normally, the A-factor is perturbed in a sensitivity analysis, but it is useful to determine the temperature window where a measurement is sensitive to a reaction. For more details on this idea, you can find a publication here: https://onlinelibrary.wiley.com/doi/abs/10.1002/kin.20080 

In the past I had access to a modified version of chemkin (from the authors of the paper above) that could accomplish this, and I was hoping someone had made a similar function in cantera.

[Steven DeCaluwe]

Hi Jeffrey,

From the ReactorBase header file, it looks like only two types of sensitivity analyses are enabled:

 

- k_fwd of the reaction, for each reaction,

- molar enthalpy of each species.

That said, could you back-convert the former to suit your purposes, by using the enthalpy of reaction?

Best,

Steven

——————————————————

Steven C. DeCaluwe, Ph.D | Associate Professor of Mechanical Engineering

COLORADOSCHOOLOFMINES

Brown Building W410B

Golden, CO 80401

Email: deca...@mines.edu

Phone: (303) 273-3666

Twitter: @CORESresearch

https://people.mines.edu/decaluwe/

https://cores-research.mines.edu

He / Him / His

On Jun 21, 2021, at 12:17 PM, jeffrey...@gmail.com <jeffrey...@gmail.com> wrote:

Does cantera have a function that can find the sensitivity of an observable (e.g. flame speed) to reaction rates as a function of temperature? In other words, can cantera perturb reaction rates in narrow temperature windows to find the effect on a global observable? Normally, the A-factor is perturbed in a sensitivity analysis, but it is useful to determine the temperature window where a measurement is sensitive to a reaction. For more details on this idea, you can find a publication here: https://onlinelibrary.wiley.com/doi/abs/10.1002/kin.20080 

In the past I had access to a modified version of chemkin (from the authors of the paper above) that could accomplish this, and I was hoping someone had made a similar function in cantera.

--
You received this message because you are subscribed to the Google Groups "Cantera Users' Group" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cantera-user...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cantera-users/53957b46-4b9d-4117-bcac-236a69baf817n%40googlegroups.com.

[jeffrey...@gmail.com]

Thanks, but I don't think I was clear enough.

I think the sensitivity in ReactorBase works by perturbing the A-factors of reaction rates. But, in a system where the temperature varies (like a flame or ignition problem), I want to perturb reaction rates by a Gaussian or Delta function in temperature-space. This can answer the question: "At what temperature does the flame speed give me information on the rate of reaction i?"

It seems like this doesn't exist, but I could try to make a python script to brute-force this calculation. But, I need a way to multiply a reaction rate by a Gaussian function in T-space. Do you know if there is a function similar to Kinetics.set_multiplier, where I could multiply a reaction rate by a temperature-dependent function instead of a constant? If not, I may have to do some odd curve fitting to use a set of modified Arrhenius reactions (or Chebyshev reactions) that sum to create something that looks like a gaussian perturbation to a reaction rate.

[Ray Speth]

Hi Jeff,

You're right that there isn't anything like this available in Cantera currently. I think it would be an interesting capability, especially if it were reasonably generalized, and would invite you to create an Enhancement Proposal in our tracker with some of this information.

Regards,

Ray

[jeffrey...@gmail.com]

Thanks Ray! I submitted a request here. For anyone who is interested, I wrote a solution to this problem, but it is slow and cannot work with three-body-reactions or falloff reactions. Feel free to use it, it's on my github here.