In practice a sampling solution is a compromise of many requirements: the sample should be representative; the sampling system should also be reliable, less than 95 per cent availability restricts the use of the assayed sample for process control purposes; no sample is the worst sample, and investment costs and installation space/layout are frequently a restriction._x000D_
For instance, online X-ray analysers for process control introduce a delimitation error by the nature of X-ray physics. Thus only in special cases a relative accuracy better than three per cent relative, as one standard deviation, is achieved. A sampling system for process control needs not to be, as representative, for so many sample properties, as one for composite laboratory analysis samples. Sampling system maintenance is also an important issue. There are many sampling systems in the industry which have been difficult to maintain and are in disuse for lack of resources or incentives to keep them running._x000D_
Pierre Gy developed the most generally accepted sampling theory. One of the main principles is that all particles in a one-dimensional process stream should have equal chances of being included into the sample. A correctly designed and operated transversing cutter comes close to this goal but has limitations in temporal representativity. For slurries an acceptable solution in many cases is to use a fixed vertical cutter(s) in a position where the slurry is well-mixed by turbulence and other means._x000D_
For online analysis, reliability problems are mostly related to faulty design or installation of the piping system to the analyser. Properties of material processed, process flow sheet and piping details before and after the sampler have to be taken into account when deciding on which sampling arrangement is appropriate. Guidelines and some solutions to these issues are presented in this paper.