The Fresh Loaf

News & Information for Amateur Bakers and Artisan Bread Enthusiasts

What is the magic formula?

Anonymous's picture
Anonymous (not verified)

What is the magic formula?

My efforts at finding The Magic Formula which predicts sourdough fermentation time if (Temperature in Farenheit and Percent Prefermented Flour) are known is not going so well.

The most precise fit of the data so far (after 3 hours of computation) is:

Fermentation Time in minutes = -5.66e10*log(Percent Prefermented Flour)/(Temperature in Farenheit + cos(Temperature in Farenheit))^4.616

I suspect my data source could be more sour than my starter.

TOO MANY VARIABLES! (or not enough maths).

(Reminds me of that scene from the movie Amadeus: TOO MANY NOTES, YOUR MAJESTY!: http://www.youtube.com/watch?v=IoY0Qa0zU0A)

Doc.Dough's picture
Doc.Dough

APPLIED AND ENVIRONMENTAL MICROBIOLOGY,
July 1998, p. 2616–2623 Vol. 64, No. 7
Modeling of Growth of Lactobacillus sanfranciscensis and
Candida milleri in Response to Process Parameters of
Sourdough Fermentation
MICHAEL G. GANZLE, MICHAELA EHMANN, AND WALTER P. HAMMES

http://aem.asm.org/content/64/7/2616.full

thomaschacon's picture
thomaschacon (not verified)

I think my problem is one of source data. If the data is wrong (or just plain random), predicitive capacity = 0.

Will have a look your source.

Thanks.

thomaschacon's picture
thomaschacon (not verified)

I let the computation run over the weekend. I'm rather surprised it didn't melt my PC, using all 4 cores at near 100% capacity.

Did it come up with a formula that explained the dataset mentioned below?

Yup, down to the thousandths of a percentage point with a goodness-of-fit near perfect. The formula was about two pages, works for that set of data, and even models the exponential. But did it have any predictive capacity beyond that limited dataset (i.e. could I enter arbitrary values for salt, etc. and have it predict a reasonable fermentation time?)

Not at all. Not one, single iota.

After reading (shall I say 'attempting to read' rather) the paper you linked to, I'm not surprised it failed.

Thank you for the reference.

ehanner's picture
ehanner

You will find the answer to your question HERE. A fellow who hasn't posted here lately and is an expert in the science behind natural yeasts as they reflect to time. He did make a spread sheet that I found workable. I didn't search through the archives but I am confident the answer to your question can be found in the tombs. Bill Wraith who goes by bwraith made many great posts we enjoyed over the years.

Hope this helps.

Eric

gary.turner's picture
gary.turner
thomaschacon's picture
thomaschacon (not verified)

I'm actually using his data, reverse-engineering it if you will (because I don't (yet) understand his methods and want more freedom to vary % salt, % prefermented flour, doubling time, etc.).

thomaschacon's picture
thomaschacon (not verified)

I'm looking for an equation based on his data.

[If it's buried in those rather daunting spreadsheets, I can't (yet) see/understand it.]

Why not just use the table?

I want to vary the % salt and % prefermented flour (not stick with [0 or 2% salt]; [max 25% prefermented flour] presented in the table.). The doubling time is also variable, but the data are static. I'd like to be able to vary that too, as it changes a lot. (Doubles today in x, doubles yesterday x - 11 minutes).

I plugged the data into http://creativemachines.cornell.edu/eureqa and asked it to model an equation.

The best it could come up with is the one I posted above (after 5 hours of computation). It's not good enough, although it does get close. I can't (yet) say "my doubling time is x" and "I want %y prefermented flour" and "%x salt", how long will bulk fermentation take in minutes? How long for proofing? How long for total?

I'll keep futzing with it.

I have the time.

dabrownman's picture
dabrownman

I'm assuming you would have to ferment in a proofing box to make sure the temperature did not vary.  I can see where this could be a very complicated exercise.  Aren't there other things besides temperature and the % of preferment that effect the ferment time required?

I'm not sure how or if you need to account for all the other variable that could potentially effect fermentation time - also in some variable way.  Wouldn't difference in SD starter type including; flours, strength, build time and hydration effect fermentation time?   Or the different hydration levels of different fermenting dough types, the different fermenting qualities of different flour types, the differing volumes of dough also could effect the fermenting time required - No? I'm not sure how many other things could be involved or not either.

Maybe I'm not getting or understanding why only temperature and % of preferment can predict an accurate time  for fermation or don't understand what you are trying to do exactly.  I think you are right - too many variables may be at work.

Still, a very interesting idea if these two things can predict ferment time, if that is your aim and worth you time and efforts.

thomaschacon's picture
thomaschacon (not verified)

With more than a few variables, the results are not significant. (Fit would be high error, so essentially useless).

If I had more computing power, I could model more variables.

Eureqa can connect to Amazon's Cloud. I could rent up to a maximum of 160 of Amazon's CPU cores @ $0.31/hour/core ($49.60 per hour), but I don't trust the data (or my maths) enough yet to spend any money on this.

I originally tried to model a formula based on all of the data in the spreadsheet (except the total time, which was just additive), but that got me nowhere. Well, it got me somewhere for a few of them (i.e. the exact time predicted for two of three data points), but the remainder were magnitudes off.