Forbes: Distilling Liquor With Machine Learning And Big Data

[Foreshot]

https://www.forbes.com/sites/jenniferhicks/2020/07/29/distilling-liquor-with-machine-learning-and-big-data/

 

Another fast aging technology:

Quote

"We use automation and predictive analytics to reduce the chaos, control flavor profiles, and scale precision quality control in the finishing process," said Nick Scarff, Master Blender and vice president of business development at Next Century Spirits. "The systems we use are fully programmable, designed to optimize certain temperatures, alcohol concentrations, and other conditions that form and maintain desired flavors."

 

[Jedd Haas]

Has anyone tried actual samples of spirits that have gone through any of the "fast aging" systems?

I spoke with a different company recently and requested samples of both "before" and "after" versions.

We'll see if they come through.

[Foreshot]

Personally I've not tried one but all the reviews I've heard is that the spirit displays some positive characteristics of aging but some are missing leading to an odd flavor.

Some reviews:

TerraPURE: https://reviews.whiskeyfellow.net/2020/04/oz-tyler-kentucky-rye-review-tasting.html

Cleveland Whiskey's proprietary process: https://distiller.com/spirits/cleveland-bourbon

THEA Reactor: https://thewhiskeywash.com/whiskey-styles/american-whiskey/whiskey-review-round-lost-spirits-old-traditional-new-thea-whiskeys/

 

[Silk City Distillers]

We played around quite a bit with ultrasound.  Did some double blind taste tests that saw a statistically significant difference (3 sample triangle test), however the panelists were somewhat split on what was "better".  So while treatment made a statistically significant difference, it was not better.  That said, with further testing, it was clear that the impact was temporary.  After a few months in the bottle, there was no longer any difference between treated and untreated.  We don't use it in production, but I still occasionally play around with it.  My last working theory was that the difference was caused by ultrasound forcing out dissolved gasses, which was temporary until the liquid could equilibrate with air again.

 

 

 

[et1883]

On 8/2/2020 at 11:17 AM, Silk City Distillers said:

We played around quite a bit with ultrasound.  Did some double blind taste tests that saw a statistically significant difference (3 sample triangle test), however the panelists were somewhat split on what was "better".  So while treatment made a statistically significant difference, it was not better.  That said, with further testing, it was clear that the impact was temporary.  After a few months in the bottle, there was no longer any difference between treated and untreated.  We don't use it in production, but I still occasionally play around with it.  My last working theory was that the difference was caused by ultrasound forcing out dissolved gasses, which was temporary until the liquid could equilibrate with air again.

 

 

 

Hi Silk City, curious about what ultrasound equipment you used in these trials, and what you refer to as a triangle test?   Seems odd that changes were not permanent structural or (molecular?) changes, but taste is always the final consumer measure.  Tks in advance.

[Silk City Distillers]

I used a large high power ultrasonic bath.

3 way triangle test works like this - you put three samples in front of the subject.  2 of the samples are Liquid A, 1 of the samples is Liquid B (and vice versa, at random).  You ask the subject to pick out the sample that's different.  You compare that against what you might expect if someone was just picking at random.  Statistically, it becomes very easy to tell if the subject can actually distinguish any difference at all in a repeatable/reliable way (better than chance).  You then ask the simple question, do you like it, better or worse?

The problem with simple A/B tasting, is that people are coming in expecting a difference, they might have a preference for one or the other, or just be guessing.  Thus, the triangle test, is the ultimate tool to judge if people can really tell a difference, and what that difference is.  If they can't pick out the one that's different, it's not really so different?

For the sonicated vs unsonicated samples, the subjects were able to select the outlier in a way that was better than pure chance, BUT, people were mixed as to better or worse.  Interesting that there is a difference, disappointing that it's not necessarily beneficial.  Maybe a far larger sample size and better procedure control is necessary (We're a distillery, not a lab or product testing institute).

My hypothesis for the temporary nature of this has more to do with ultrasonic degassing of the spirit.  It's a pretty obvious phenomenon when you open a bottle, drop it in the bath, and watch it "fizz" for a few minutes before going still.  Over time, the bottle is going to absorb gas to equilibrate with the atmosphere again, it's a temporary phenomenon.

There is a tremendous amount of "distillers bias" in everything we do.  There is such an overwhelming desire to find evidence that justifies our own hypotheses, that we'll usually find it "Oh, yeah, see that's much better".  But is it?

[et1883]

tks, after reading your comments i just looked at ultrasound gear and it's not cheap and seems mostly lab size vs mfg scale.  "different" is not always "better," that's good to remember.  IIRC there was anoter group in Colorado (?) who featured ultrasound as part of their spirits production, but I haven't tasted any of their products. 

[Silk City Distillers]

Hielscher makes larger scale sanitary processors, they are very expensive. 

https://www.hielscher.com/wp-content/uploads/2013/06/uip16000_0980x12001.png