Computational Engineering / What computational engineering is
Responsibility for the output
Who owns a computational result, what a review is for, and why 'the software gave me that' is not a defence.
Lesson review: expert verification remains open
This lesson has received a full-context editorial review and a check of its technical explanations. Selected arithmetic has been checked separately. This is not independent engineering sign-off.
Checks needed for project use
- A computational engineering reviewer must check algorithms, convergence claims, statistical assumptions and numerical examples beyond the cases explicitly recorded.
These are educational models and exercises. Check their assumptions before using the results. The full technical content and every diagram have not yet been independently verified.
Concrete and steel calculations retain their stated first-generation Eurocode teaching basis. Use the standards and amendments specified for your project; do not mix generations without an appropriate design basis.
Read the lesson register and outstanding review notesWhy this matters
Structural failure is rare, and when it happens the analysis is usually not the sole cause. But there is a recurring pattern in investigations of failures involving computation: the model produced an answer, the answer was used, and nobody had established what the model could not tell them.
This lesson is about the professional position that follows from that, and about what a review has to do in order to be worth having.
Responsibility for using the result
Before relying on a computational result, establish that the model and analysis method are suitable for the intended use. Errors can arise in the software, inputs, interpretation or assumptions about capability.
Check the documented scope of each function. A design module may omit punching shear, a diaphragm representation may omit openings, or an output may report displacement at only one point on each floor. A result can be correct within those limits and still leave part of the engineering question unanswered.
What a review should cover
A review should examine both whether the model was implemented as intended and whether its assumptions suit the structure. Use independent evidence where practical: a hand calculation, a simpler model, another analysis method or observations of behaviour.
Ask which load cases and behaviours are absent, and whether their omission could affect the conclusion. Reproducing a result with the same assumptions can help check implementation, but does not independently establish those assumptions.
Recording assumptions
Record the assumptions, limitations and exclusions, with reasons for decisions that may need review. This lets a checker challenge a decision without reconstructing the modelling history. Modules 6 and 13 develop the model record used for this purpose.
The engineer still needs to assess the analysis when computation is provided by a vendor or cloud service. Legal and contractual responsibilities are separate questions.
Practice
A model reports a total vertical reaction of 4,820 kN. The building is 6 storeys, 32 m × 18 m on plan, and the design permanent plus imposed load is 9.5 kN/m² at the serviceability limit state. Estimate the expected total reaction in kN, and state the ratio of model to estimate to two decimal places.
Check yourself
An analysis is run on a vendor's cloud solver. Where does responsibility for the analysis sit?
Summary
- The engineer is responsible for the output, including the parts nobody looked at
- The commonest serious failure is not a software bug but a capability the user assumed and the software never claimed
- A review is only worth its cost if it brings an independent route, independent eyes on assumptions, or the questions nobody asked
- Write down what you decided not to check, and why
- An unexplained disagreement is a finding, and it stays open