There is discussion on Twitter and elsewhere about the nature and value of engineering. I think it is time to set out what I see as a major problem. It may be limited to structural engineering but I suspect not.
I see many examples of what Ben Goldacre (@bengoldacre) rightly calls bad science in the modern practice of engineering. This begins with "believing the model". That is building a model (eg in FE) which is founded on a series of preconceptions and them believeing that the model represents reality.
Worse than that, though, is "believing the test" by which I mean, building a test that is predicated on a series f conceptions of behaviour and then believing the test result represents the whole of behaviour.
An example is the series of tests carried out on masonry bridges in the 1980s. They are widely believed to provide definitive results on the ultimate capacity of such bridges. However, the bridges were tested using a line load across the full width. This was an attempt to create a 2D test to correspond to the 2D analyses that were available.
But the bridges are NOT 2D. The stiffness varies hugely across the width and that variation is an important part of the drivers of behaviour. The side walls (called spandrel walls) form extremely stiff and strong elements compared with the arch plus fill which is assumed to occupy much of the width. Running the loading beam close against the parapets meant that it was sitting on those walls, which are noticeably thicker below the surface. The result actually tells us nothing useful.
These results are written into a code of practice for bridge assessment and it is explicitly stated that the results are correct and that analyses must deliver these results if they are correct.
John Broughton (@isambardslad) commented this morning that Engineering is hard - you don't get cut much slack in engineering. In my experience, engineers are often cut too much slack until they create a failure and then they (or often their minions who cannot honestly be held responsible) are castigated by colleagues.