ABSTRACT
Researchers often have some expectations about the experimental results, but only in rare cases, all calculated quantities perfectly match the theory. There are deviations from theory, and the researcher has to judge whether there is reason for the deviation in the phenomena they are investigating or whether it is just a lack of sensitivity or accuracy of the equipment at the individual measurement points.
It is very common that researchers think about this question after their experimental setup is done and the experimental work is ongoing or even finished. Error analysis is often done after all data reduction and result discussion. This chapter will show that it has to be the other way around. The error analysis has to be done as early as possible and must be a part of the experimental planning.
Consider an experiment where inlet and outlet temperatures of a fluid stream are measured. If the temperature change is only 1 or 2 K and the heat flow is calculated from that temperature change, even a measurement uncertainty of 0.2 K can lead to a deviation of calculated heat flow that may question the scientific outcome of the experiment. An early analysis of measurement errors accompanied by an analysis of error propagation would have shown how the uncertainty of the individual temperature measurements influences the calculated heat flow.
This chapter shows common ways of error analysis, error propagation analysis, and its importance in experimental planning. It outlines the methods to decide which measurement device should be as accurate as possible and which one could be less accurate and helps to make reasonable decisions on how to spend the budget.
Measurements are nowadays often used to validate calculations done in computational fluid dynamics (CFD). While error analysis also plays a major role during the planning of the experiments, calculation results must be compared with experimental data in a way that the validation can be justified. As deviations are always present, the question is where they come from and how they could be avoided. Again, these questions should be addressed in advance because that makes result discussion easier after the experiments are done. CFD models are based on assumptions concerning the boundary conditions. If the experiments do not reflect these conditions, this could be a source of deviation. The design of experiments for validation of CFD calculations is addressed in the second part of this chapter.
