8 min read
Konxis Engineering
2026
Why FEA results can be misleading
A converged solution is not automatically a correct engineering answer.
Practical articles on analysis, failure investigation and verification — including when advanced simulation is necessary, when it is not, and how to tell whether an engineering result can actually be trusted.
8 min read
Konxis Engineering
2026
A converged solution is not automatically a correct engineering answer.
6 min read
Konxis Engineering
2026
What verification and validation can establish — and what they cannot.
7 min read
Konxis Engineering
2026
How to decide when the physics requires nonlinear behaviour and when a simpler model is better.
A converged solution is not automatically a correct engineering answer.
Convergence is a numerical statement. It says the solver found a solution that satisfies the discretised equations to a tolerance. It says nothing about whether those equations describe the physical problem in front of you.
The common failure pattern is familiar: a model runs, the contour plot is smooth, stresses look plausible, and the result is accepted because nothing obviously went wrong. But a model can converge perfectly and still represent the wrong load path, the wrong constraint, or the wrong failure mechanism.
Each of these can be wrong while the solution still converges cleanly. The plot looks the same either way.
A result becomes evidence when it has been challenged: mesh and convergence checks, sensitivity to boundary conditions and material inputs, equilibrium and hand-calculation sanity checks, and — where it exists — comparison against measurement.
The question is never “does it converge?” It is “what would have to be true for this answer to be wrong — and have we checked?”
That is the difference between running a simulation and doing engineering with it. The contour plot is the beginning of the argument, not the end of it.
What verification and validation can establish — and what they cannot.
Trust is not a property of a model. It is a relationship between a model, a specific use, and the evidence available to support that use.
Two distinct questions are usually mixed together. Verification asks whether the analysis has been set up and solved correctly — geometry, loads, boundary conditions, material inputs, mesh, convergence, numerical behaviour. Validation asks whether the model represents the real system adequately for its intended use.
Validation compares prediction against something independent: a component test, field measurements, strain data, vibration response, operating behaviour. The comparison is only meaningful if the intended use is stated first. A model validated for trend prediction has not thereby been validated for absolute life prediction.
Our evidence ladder makes this explicit: assumed input, calculated result, simulation result, measured result, validated result, released design. Confidence increases as independent evidence increases — and not every project needs to reach the last rung. Stating where a result actually sits is part of the engineering.
A model is trustworthy when its limits are as well documented as its results.
How to decide when the physics requires nonlinear behaviour and when a simpler model is better.
Nonlinear analysis is not a higher level of engineering. It is a specific response to a specific physical nonlinearity. The method should follow from the behaviour the answer depends on.
If the answer to all four is no, a linear model is not a compromise. It is the correct, proportionate tool — and it is easier to verify.
A nonlinear model multiplies the ways to be wrong: contact definitions, convergence parameters, load stepping, material characterisation. Every one of those needs evidence. A complex model that cannot be verified does not outperform a simple model that can — it merely moves the uncertainty somewhere less visible.
Often the right sequence is staged: an independent hand calculation to establish scale, a linear model to find the critical region, then nonlinearity introduced only where the physics demands it.
The objective is not the most complicated model. It is a model and evidence base that are appropriate for the decision.
Sometimes the right conclusion is that a more complicated analysis is not necessary. Saying so is part of the job.
You do not need to know which analysis method you need. Tell us what is happening, what you already know, and what decision you need to make.