Checking Precision and Bias Before and After Construction of Sampling Systems

Sampling systems are never perfect. They all have some bias and give variable sampling errors. The ways of thinking about bias and precision should be different before and after construction (or reconstruction) of a sampling system. Before a sampling system is constructed, the fundamental error variance can be estimated using GyÆs formula or by doing test work on the material to be sampled. The component of sampling variance which varies with the number of sample increments can be computed for the expected patterns of grade variation. The smallness of the bias can be checked using discrete element modelling, by checking that the proposed design complies with clauses in standards, or by copying an existing sampling system that performs well for similar material. After a system has been constructed, sampling precision can and should be monitored directly. Anyone claiming to take responsibility for a sampling system should know its precision. Bias could be checked by comparison to a more accurate sampling system, such as stopped-belt sampling, under conditions where bias seems most likely. However, it is more practical to conduct regular checks that the system is still operating as designed, that the design is still appropriate despite any changes to material flows, and that sample masses are consistent with the correct operation of the sampling system. If precision is not regularly assessed then it should be assumed to be very poor. If a sampling system is not regularly inspected then it should be assumed to be biased.