CMP Toronto
Hockey Hall of Fame
Sept. 10, 2015.
The problem of using non-probabilistic in-stream probes is that they can be biased negatively one day, be accurate the next, and positively biased a week later. This means an assumption that the process stream is homogeneous must be made, which is difficult. The amount of sampling error that is introduced depends on varying characteristics of the stream; the flow rate, the pressure, velocity, SG of slurry, viscosity and piping layout.
Why does bias happen with non-probabilistic samplers? If there are the slightest differences in the slurry entering the sample volume due to stream characteristics, then the sample volume will be different than the average of the whole stream. Additionally, even if a totally turbulent state could be artificially created in the flow just before the sampling point, the introduction of an obstacle would re-structure the flow in an unpredictable manner.
Top-Mounted Cutter Sampler with Mixing Flume ahead of fixed cutter. Encourage the slurry to be well mixed before a cross sectional cut is taken.
In conclusion, Heath & Sherwood has experience with customer applications where non-probabilistic sampling stations, such as pressure pipe probes and even in-line gravity cutters, have to be replaced with a probabilistic linear moving cross-stream cutter station due to the lack of sampling accuracy.
Reference
1) Pitard F. Francis - Pierre Gy’s Sampling
Theory and Sampling Practice – 2nd Edition -
The Delimitation Error page 243 14.5.2.
Taking Part of the Stream All of the Time