A second opinion is not an accusation

One of the strange things about engineering is that we spend our entire professional lives checking things, yet we can still become uncomfortable when somebody checks us.

A calculation is checked. A drawing is checked. Materials are tested. Structures are inspected. Loads are combined according to rules developed precisely because relying on intuition is not enough. We accept all of this as normal. Then, on certain projects, somebody proposes an independent review of the structural design and suddenly the atmosphere changes. It can feel as if the second engineer has been called because the first one is not trusted.

I have never liked that interpretation.

Years ago I worked on the independent structural validation of a long span pedestrian bridge in Perth, Australia. The assignment involved building an independent finite element model and studying the behaviour of the structure under conditions that included wind and pedestrian induced dynamic effects. The objective was not to redesign the bridge, and it certainly was not to search for mistakes made by the original designers. The purpose was much simpler and, in my opinion, much more valuable: to approach the same physical problem independently and see whether another engineering path led to the same conclusion.

That experience stayed with me because it reinforced something I had already begun to understand. In structural engineering, the calculation is rarely the most interesting part of the calculation.

Before a computer solves anything, an engineer has already made a long sequence of decisions. We decide how to represent the structure. We define restraints. We assign stiffness. We idealise joints that in reality are never perfectly pinned or perfectly fixed. We decide which effects deserve to be modelled explicitly and which ones can reasonably be neglected. We choose material properties, damping assumptions, imperfections, load distributions and combinations. On complex structures, these decisions can influence the result as much as the mathematics that comes afterwards.

The software calculates the model we give it. It does not tell us whether we have understood the structure correctly.

This is why two competent engineers can analyse the same structure and produce models that are not identical. That does not necessarily mean that one of them is wrong. Often it simply means that they have translated reality into mathematics in different ways.

The interesting part begins when the results are compared.

If two independent models, built from different assumptions and perhaps with different software, describe essentially the same structural behaviour, the value of that agreement is considerable. Confidence increases not because the same calculation has been repeated, but because the same physical conclusion has survived two separate ways of thinking.

If the results do not agree, the situation becomes even more useful. You begin asking why.

Was stiffness represented differently? Was a connection idealised in another way? Was the mass distribution different? Was one model more sensitive to second order effects? Did someone interpret the boundary conditions differently? Is the disagreement numerical, or is it revealing something about the structure that deserves to be understood better?

Those conversations are some of the most valuable conversations engineers can have. They are also exactly the conversations that become difficult when verification is treated as a judgement on the person rather than an examination of the problem.

There is another reason why independent review matters, and it has very little to do with technical ability. It has to do with familiarity.

When you have worked on a project for months, you know it too well.

That sounds like an advantage, and most of the time it is. You remember why a certain beam was moved. You know the architectural constraint behind an unusual geometry. You remember the meeting where a connection was changed. You understand why a structural system evolved in a particular direction.

But familiarity also has a cost. You gradually stop seeing some assumptions as assumptions. They become part of the project. The model begins to feel like the structure itself.

An engineer coming to the same problem for the first time has none of that history. They do not know what the structure is supposed to do. They only see what has been given to them and ask whether it makes physical sense.

That distance is extremely valuable. It is very difficult to create it artificially when you are the person who developed the original design.

This is not unique to structural engineering. A writer can read the same page twenty times and stop seeing an obvious mistake. A programmer can spend days looking for a problem in code that another developer notices in five minutes. A company can become so accustomed to one way of working that an outsider immediately sees something nobody inside questions anymore.

Expertise helps us see more, but familiarity sometimes makes us see less.

Independent validation gives a project another pair of eyes before concrete is poured, steel is fabricated or a problem becomes physically expensive.

There is something else I learned over the years. The best reviewers are not the engineers who approach another person’s work trying to demonstrate that they would have done it differently.

Doing something differently is easy. Engineering contains judgement, and judgement naturally produces different solutions.

The useful reviewer asks a different question: is the reasoning defensible? That distinction matters enormously.

If I would have chosen another structural system, that does not automatically make the chosen one wrong. If I would have used a different modelling strategy, it does not mean the existing model is invalid. If my calculation produces a reinforcement ratio that differs by ten percent, the difference itself tells us very little until we understand its cause.

A second opinion should not be an exercise in professional ego. It should be an attempt to reduce uncertainty.

This becomes more important as structures become less intuitive.

A simple beam gives us a great deal of physical intuition. We can often anticipate its deformation, its critical sections and the direction in which forces will travel before opening any software.

A long span bridge, a cable supported roof, a geometrically complex shell or a structure sensitive to dynamic excitation is different. Behaviour emerges from interactions that are much harder to understand by looking at a single element. Small modelling choices can influence global response. Modes interact. Serviceability can become as important as strength. Human comfort may govern a design that would otherwise be structurally safe.

In these situations, complexity creates another problem: very sophisticated calculations can produce very convincing images.

Modern structural software is extraordinary. We can create beautiful three dimensional models, colour maps, mode shapes, stress plots and animations that make a calculation look almost unquestionably precise.

But visual precision is not the same thing as physical truth. A wrong assumption presented in twenty colours is still a wrong assumption.

Sometimes the more sophisticated the model appears, the more disciplined we need to be in questioning what went into it.

This is one of the reasons I have always valued independent modelling. When possible, I prefer not to begin by opening somebody else’s model and checking the settings one by one. There are situations where that is necessary, of course, but there is a different value in starting from the drawings, the geometry, the loads and the engineering problem itself.

You build your own representation. Only afterwards do you compare.

If the objective is true independence, the second engineer should have enough freedom to make their own mistakes.

That sentence may sound strange, but it is important. If I reproduce every assumption made by the first designer, I may simply reproduce the same blind spot. Independence only becomes useful when two engineers are allowed to think independently enough that their errors are unlikely to be identical.

Then agreement means something.

This way of thinking has influenced much more than the individual projects I have validated. It has become part of how I approach structural engineering in general.

I am not particularly interested in being right quickly. I am much more interested in understanding why something is right. There is a difference.

Being right can occasionally happen because an assumption was conservative, because two errors compensated each other, or simply because the problem was forgiving. Understanding requires more work. It means testing the structure mentally and numerically, changing assumptions, looking at deformation before stress, checking whether reactions make sense, comparing simplified calculations with sophisticated ones and asking whether the behaviour remains believable when the model becomes more complex.

Sometimes the most useful check on a finite element model is still a calculation made on a piece of paper. Not because paper is more accurate than a computer, but because a simple calculation forces us to state clearly what we think is happening.

If the sophisticated model says one thing and a rough physical estimate says something completely different, I do not immediately trust the sophisticated model because it took longer to build. I investigate the difference.

That habit becomes even more important when another engineer is reviewing your work.

The natural reaction to a disagreement is often to defend your model. The better reaction is curiosity.

Why did they get a different result?

There may be a mistake in their analysis. There may be a mistake in yours. More interestingly, both analyses may be reasonable and the difference may expose a sensitivity in the structure that neither team fully appreciated before.

In that case, the disagreement has improved the project.

This is why I think the language we use around verification matters.

A second opinion is not an accusation.

It does not say that the first engineer is incompetent. It does not diminish authorship. It does not transfer responsibility. It simply recognises something that engineering should be mature enough to admit: serious technical decisions deserve more than one opportunity to be questioned.

We accept redundancy as a positive quality in structures. If the failure of one component does not lead immediately to disproportionate collapse, we consider the system more robust.

There is a kind of intellectual redundancy in engineering too. Different engineers. Different models. Different assumptions. Different ways of looking at the same physical reality.

When they converge, confidence grows. When they do not, we learn something before the structure has to teach us itself.

There is also a professional lesson in this.

The longer I work as an engineer, the less interested I am in proving that I know everything about a project. Complex work has a way of curing that ambition.

Some of the best engineers I have met are very confident in what they know and equally comfortable identifying what they want someone else to look at. They understand that asking for another opinion does not weaken their judgement. It strengthens the decision that eventually has to be made.

This is particularly important because structural engineering has an unusual relationship with success.

When we do our job well, very little happens. The building stands. The bridge carries people. The roof remains where it was designed to remain. Years pass. Nobody notices.

There is no dramatic evidence of the thousands of decisions that produced that normality. When something goes wrong, however, the consequences can be immediate and physical.

That asymmetry should make us humble.

The cost of an independent review is usually tiny compared with the value of the structure being designed, and almost irrelevant compared with the cost of discovering a serious problem after construction has started.

Yet the true value is not only economic. It is the opportunity to be wrong while being wrong is still inexpensive. On drawings. In calculations. In a meeting. Between two engineers. Before reality becomes the final reviewer.

I have been on both sides of that table. I have reviewed the work of others, and I have had my own work questioned. Neither position should be comfortable all the time. If a review never creates a difficult question, it probably has not gone very deep.

But discomfort is not the same thing as conflict.

A good technical discussion should be capable of separating the idea from the person who proposed it. The connection works, or it does not. The deformation is acceptable, or it is not. The assumptions represent the physical system adequately, or they need to be reconsidered.

The structure has no interest in who wins the argument.

That may be one of the healthiest things about engineering. Reality eventually removes the ego from the calculation. We can defend our models, our experience and our decisions as much as we want. Once the structure exists, it responds only to physics.

For that reason, I would rather have someone challenge an assumption while there is still time to change it. I would rather answer a difficult question during design than explain later why nobody asked it.

And if another engineer builds a different model, approaches the problem from another direction and arrives at the same conclusion, I do not see that as duplication.

I see it as evidence.

A second opinion is not an accusation.

On the projects where the consequences matter most, it is simply another way of respecting the fact that what we design eventually has to exist without us.