determining alkalinity via pH and co2*

[karimwassef]

My alkalinity is a function of pH and CO2.. if they're stable, my calculation would be stable.

In general, both are smooth over time, but not necessarily constant.



The two points where there's a disconnect was due to the probe being too close downstream to the kalk drip during times when the drip was very high (continuous).

[orcafood]

Thanks so much! The kalk spike really sk(r)ewed up those results haha .

To properly start testing our standard error we are going to need multipe data points at every time stamp, or just more timestamps in general. Api kits are only accurate to 1 dkH if memory serves?

[karimwassef]

I usually read to 0.5 since it goes blue-green-yellow so I can usually gauge the difference between 8 and 8.5.

There was no kalk spike ... this is normal operation a few months ago. I had just turned off my kalk addition for a day so I could get a clear baseline. That's why it starts st 8.05.

My pH feedback loop had a target that was also stairstepped in time to introduce some pH variance during the day.

[orcafood]

With a burette we could get a much more accurate alkalinity reading, definitely worth the price. Then we could really see how close the alkalinity approximation truly is I am super curious.

To get a good idea of how accurate the alk reading is we would also need data from a full day where the kalk reactor and entire system is humming along in perfect equilibrium.

[karimwassef]

This doesn't take into account nonlinear effects

Case in point - when my kalk reactor ran all night and didn't impact my Alk by more than 2dkH.

So this is only usable in the linear range. Once we get into saturation, it breaks down.

[orcafood]

We could easily use svm or some other nonlinear modeling technique, no problem. It definitely breaks down during a surge of hydroxide, but that should be a very rare occurrence.

[karimwassef]

True

[kenneth wolfe]

Originally Posted by karimwassef

I think it works for me because of my air injector. I push gallons of air into my water constantly through a very fine bubble foamer. It's a dual penductor fed by outside fresh air through a high pressure pump. This means that the time variable for the diffusion of gas into water is reduced significantly. Basically, I force the assumption that I'm at steady state to be more true.

Air CO2 sensors are only ~ $100. You could, in theory, drive it from an Apex and have all the data you need.

Unfortunately, Apex won't let you enter a math formula that returns a function of known values but this app works for me: iFxCalc iFxCalc:

https://itunes.apple.com/us/app/ifxc...786018193?mt=8

It's clunky to figure out but easy once you do

The formula is
Alk=(2*D3*0.0334211*C3/29.41*0.000001*10^(-(-8.712-0.00946*F3+0.0000856*F3^2+1355.1/(((E3-32)*5/9)+273)+1.7976*ln((((E3-32)*5/9)+273))))/10^(-G3)*10^(-(17.0001-0.01259*F3-0.000079334*F3^2+936.291/(((E3-32)*5/9)+273)-1.87354*ln((((E3-32)*5/9)+273))-2.61471*F3/(((E3-32)*5/9)+273)+0.07479*F3^2/(((E3-32)*5/9)+273)))/10^(-G3)*44.01+D3*0.0334211*C3/29.41*0.000001*10^(-(-8.712-0.00946*F3+0.0000856*F3^2+1355.1/(((E3-32)*5/9)+273)+1.7976*ln((((E3-32)*5/9)+273))))/10^(-G3)*61.02)/2*1000/17.9

Where Alk(D3, C3, F3, E3, G3) has
D3 = barometric pressure (mmHg) = 30
C3 = CO2 (ppm) = 450
F3 = salinity (ppt) = 34.5
E3 = temperature (F) = 77.5
G3 = pH = 8.3

Returns 9.8 dKH

Can machine learning reduce the impact of measurement errors- probably but you'll need an accurate and frequent Alk measurement to feed into it.

Man you must be my brother, we need to get a beer together..

Sent from my SAMSUNG-SM-G870A using Tapatalk

[karimwassef]

Thanks. Come on down to Texas and it's on me!

[kenneth wolfe]

Originally Posted by karimwassef

Thanks. Come on down to Texas and it's on me!

I'm in Texas, I live in Texas city

Sent from my SM-G892A using Tapatalk

[karimwassef]

I'm in Plano... north Dallas.

You're in Galveston - by the Texas coast... seems so far away