In the paper Numerical Study of Large-Amplitude Free-Surface Motions (1966) states:
There are several purposes which computer studies such as these fulfill. First, they serve to test directly the applicability of the equations on which the computations are based, by direct comparisons of the results with physical experiments. Second, they give information about the processes for which laboratory experiments are too difficult or costly to perform. Third, and perhaps most important, they resemble accurately controlled and recorded experiments, furnishing data upon which analytical models can be based and tested. Thus, computer studies complement the work of the laboratory experimenter and of the analytical theoretician, but certainly do not replace either.
The paper describes how calculations which include the full nonlinear features of the Navier-Stokes equations in several space dimensions serve as controlled experiments, whose detailed output can guide the development of analysis as well as the planning of experimental programs.