I thought you meant microbiological testing? With temperature it depends whether you are cooling something down or keeping it cold.
For cooling down, I would get a datalogger measuring product and air temperature in a stack in several different parts of the chiller, at 'typical' temperatures. It should then give you an idea of how long it takes to cool, you then need to think 'is that cooling time short enough?' and work out what the longest acceptable cooling time is (your critical limit - you might have to do some micro testing or literature searches for that). Then 'in production', I would measure top, middle and bottom at frequencies well within the expected cooling time of each batch. If something wasn't cooling at the required rate, you could have a tolerance limit and corrective actions such as destacking to maximise surface area.
If you are just keeping things cool which are already cool, I don't think it's a statistical problem. I would measure the air temperature, (probably every 2 hours). Only if the air temperature changes would the product temperature change (and there would be a significant time lag). If the air temp is out of spec, then I would start checking product temps and moving product if possible. Again, I think the only way to check product temperature in that kind of scenario is checking each stack; top, middle, bottom because if you did have a chiller failure, product warms up at different rates according to what it is and where it is (airflow etc).
Edited by GMO, 15 May 2008 - 06:30 AM.