The purpose of computer simulations is to obtain the value of a quantity of interest on which decisions can be based. The question is whether there is the courage to take responsibility for the computed numbers and "sign the blueprints". Verification is the process leading to the confidence that the numerical approximation of the exact solution of the mathematical problem (which exists and is unique) is sufficiently accurate.
The talk addresses verification in context of a straightforward engineering problem involving a shell structure supported by a stiffening ring. We will present solutions to the problem obtained by various analysts using commercial programs and show that many of these results are significantly or completely wrong. We will discuss general principles how to obtain confidence in the computed numbers and apply these principles to the shell problem computation. A modern hp-adaptive finite element solver as well as the classical h-version of FEM are considered.