We Built the Model. Then the Asset Told Us Something Different.

Structural engineers are taught to design for loads. You establish the load cases, you apply the factors, you size the members accordingly. The process is well-defined and the codes are thorough. What the process does not prepare you for particularly well is the moment you instrument a structure that has already been designed and built, and discover that the loads it is actually experiencing are not the ones the model assumed.

This happens more often than the profession discusses openly.

The gap between the model and the asset#

A design model is a representation of reality. It captures the geometry, the material properties, the connections, and the applied loads as understood at the time of design. It does not capture what the asset actually does under operating conditions — the dynamic amplification from nearby machinery, the load redistribution that occurs as connections loosen over years of service, the temperature gradients that induce stresses the static analysis did not account for.

For most assets, in most conditions, the gap between the model and reality is within the factors the designer applied. The structure performs as intended. Nothing goes wrong.

For some assets — particularly those subject to dynamic loading, those that have been modified or repaired over their life, those approaching the end of their original design life — the gap is larger than the factors cover. The structure is carrying loads the model does not know about, and the inspection programme is not set up to find them because no one is looking for something the model says is not there.

What the data says#

We instrumented a set of vibrating screens at a CHPP to assess their operative parameters — orbit shape, amplitude, operating speed. The purpose was to check whether the machines were running within specification.

What the data also showed was the load path through the support structure under live operating conditions. The dynamic forces being transmitted from the screens to the steelwork were measurable, real, and in some locations different from what a standard analysis would have predicted. Not catastrophically different. But different in ways that were relevant to the fatigue assessment of the connections.

That kind of information does not come from a shutdown inspection. It only comes from the asset running.

What this means for how we assess things#

The engineering profession has always known, in principle, that operating loads can differ from design loads. The practical question is what to do about it when you are assessing an ageing structure and you do not have instrumentation data.

The honest answer is that you make conservative assumptions, apply appropriate factors, and acknowledge the uncertainty. That is what the codes are for. It is not a failure of the methodology.

But there is a category of asset where the uncertainty is large enough, and the consequence of being wrong is significant enough, that conservative assumptions are not a satisfying answer. For those assets, instrumentation is not a luxury. It is the thing that turns an assessment from a defensible estimate into an informed one.

The model is still useful. It will always be useful. But it is a starting point, not a conclusion — and the assets that have been running for thirty or forty years have had time to develop opinions of their own about the loads they carry.

Listening to them is worth doing.