Traceabillty of mixed raw material batches

In our factory we roast and ground nuts to make a mass that is transferred to a silo/tank where it is constantly mixed. The tank is cleaned maybe once per year and the problem is with traceabillity . I have searched this forum about how one batch of materials is not significant in xx batches later, but do not understand how to calculate it and how much product has the initial material batch in case of a recall. Our product has 6 month shelf life and we produce it every day. We have a paper traceabillity where employees write the time and quantity of nuts that were ground and put into the tank. Can somebody please explain how to calculate the fraction remaining of an old lot that is too small to be of interest?

I worked with a couple of florida oj producers in desinging a recall program - this is very similar in that many times for have oj coming in from multiple countries and mixing into hugh tank farm vessels - we finally were able to separate by type and actually take it all the way down to a section xontaininf 20 trees - oh bot, that was fun!

If your tank is constantly mixed, you'd at best have to calculate input v output daily as well as be able to calculate when (if?) the tank is ever actually homogeneous throughout.

 

E.G. you add 1000 kg of "new" ground nut, that takes ~5 minutes to pump into the tank, at the same time you pump out 2000 kg of product from the tank, did any of the new product go directly into the 2000 kg of output? is your pump out time the same as pump in time seconds/kg? 

At what speed are the tank paddles capable of mixing the tank completely?

 

The manufacturer of the tank should be able to provide you with those specifications and your maintenance department should know the flow rate of your product

 

Once you've figured out all of that, you should be able to calculate an acceptable position +/- 10%, then you can confidently show your regulator/auditor how you got to that point

The solution typically devolves to defining a "Lot". And the associated Recall criteria.

 

This is also analogous to the well-known, continuously fed, silo scenario  which preciously discussed on this Forum. With added complications.

 

A large number of theoretical analyses exist for silo systems  based on Scampi-type logics  and often, from memory necessitating a specific periodic "reset" moment.

 

My guess is some empirical time-based "Lots" are going to have to be introduced/experimented.