6. DON'T USE FINITE ELEMENT AS A SUBSTITUTE FOR GOOD STRUCTURAL DESIGN

There is an even more basic and pernicious problem with our current use of Finite Element Analysis. Before FEA came along, naval architects had to stick to structural concepts that were easy to analyze with the limited tools available. These structures tended to be simple and elegant. At their best, the structure flowed in a regular series of continuous rings -- transverse, horizontal and longitudinal. Discontinuities, sharp corners, small radii were studiously avoided for everybody knew they generated high stress concentrations but nobody pretended to know how high.

With the advent of finite element, a new philosophy emerged. At some yards, it's called "design for producability". The idea is to stick together flat plates of steel in whatever way best suits the production process, and then use FEA (aggressively interpreted, of course) to beef up the scantlings in the high stress areas that result. The result is complex and hard to follow stress patterns, small inserts of very thick plate, myriad discontinuities and sharp corners, plate thicknesses bouncing all over the place -- plug ugly structure which violates the fundamental rules of good design. As the Class rules become more and more FEA based, surveyors become more and more helpless to resist this development.

It's far from obvious that the yards gain much from this development. Elegant, simple structures tend to be the most efficient. But it's clear that owners, underwriters and eventually society as a whole lose big time. One possible solution is to separate structural approval into two stages. In the first stage the structural concept would be approved. This would require a series of simple drawings showing only the major structural members for the entire ship (not just the parallel mid-body) but no scantlings. Once the structural concept has been approved, then FEA can be applied (conservatively interpreted, we'd hope) to determine the scantlings.

If we can't enforce decent structural design, then we must limit the size and frequency of scantling changes. The yards go to ridiculous lengths to save a gram of steel. In areas, where the stress is changing rapidly, we have seen 0.5 m2 or smaller with the thickness changing by 5 mm or more between adjacent plates. It is far from clear that the yard gains anything given all the extra cutting and welding. But what is clear is that this nonsense makes a mockery of the FEM mesh size and generates multitudinous stress concentrations and additional opportunities for corrosion and welding screw ups. The Hellespont spec is that no insert can be smaller than the FEM mesh size and the change in thickness between adjoining plates can be no more than 3 mm. Class should enforce at least as stringent a requirement.

PREVIOUS NEXT