In this discussion of numerical and physical modeling, I would like to mention one advantage of model testing that you do not find to the same extent in numerical analysis: the possibility of detecting unexpected or unknown phenomena.
I would like to illustrate this aspect by referring to a couple of examples of test results from my years of experience in model testing of offshore structures. The first example is from the testing of an articulated tower—a type of structure that was extensively used for mooring and transfer of oil from production platforms to shuttle tankers some years back. It consists of a slender tower, fixed to a foundation plate by a universal joint, and the foundation plate finally piled to the seafloor. Our tests included towing the tower horizontally in waves to verify the predicted wave loads during tow-out.
When we hoisted the model into the basin, something unexpected happened. The base plate tilted out to an angle way beyond the design capacity of the articulation. If this had happened in prototype, it would have destroyed the articulation, the jumper hoses, etc. The designers had done all kinds of extensive finite element structural and hydrodynamic analysis of the tower and base plate to determine stress levels, but they had forgotten to do a simple check of the hydrostatic stability of the baseplate. Of course, once revealed by the model test, the problem was easily rectified by proper ballasting of the base plate during tow-out. The tower is still operating satisfactorily today.
Another example: Many years ago we tested a rather unusual vessel designed for a very specific offshore application. It was more than one km long and consisted of two steel pipes of a couple of meters diameter connected by a narrow deck structure. The main purpose of the tests was to verify the section forces and bending moments in waves. All went well as long as we tested the vessel in waves and current in the longitudinal direction of the vessel. However, as soon as we applied some current in the transverse direction, the vessel flipped over and capsized. The immediate reaction of the designers was surprise, since the vessel had very large metacentric height. (It had been carefully calculated in advance!) However, having seen the vessel capsize in the basin a few times, it was easy to understand the physics of it. The structural design of the vessel resulted in a relatively large bending stiffness in the horizontal plane and a much lower bending stiffness in the vertical plane. Combined with a large length-to-width ratio, this led to instability and capsize even for relatively small lateral forces on the structure. Due to this instability revealed in the model test, this vessel never went further than the model test stage.
I could mention many examples where unexpected, unknown, or forgotten phenomena have been discovered in model testing of offshore structures. Sometimes the problem has been easy to rectify, other times it has led to discarding the whole idea of the structure or concept. Safeguarding against such surprises can hardly be done by numerical analysis, since numerical analysis can only deal with those phenomena that are included in the underlying theory of the analysis program.
So, my message may sound like advertising for the model basins. My recommendation is to take care when designing offshore structures and to perform careful model testing of the complete structure, system, or marine operation. The model testing will verify that no significant phenomenon has been forgotten in the numerical analysis.