Computational Engineering / What computational engineering is

Choosing the method

Nine ways to answer a structural question, and how to pick the cheapest one that actually answers it.

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 notes

Why this matters

Choose the analysis method after defining the question. A short calculation may be enough for a simple member; a stability problem may require a model that represents second-order behaviour. Record the choice and its limitations so that a reviewer can assess it.

By the end of this lesson you should be able to

List the nine routes to a structural answer and what each costs
Match a question to the cheapest route that answers it
Recognise when the answer is 'no model at all'
Choose independent checks proportionate to the model and the consequences of error

The nine routes

Roughly in order of cost, and cost here means your time plus the checker's time, not computer time — computer time has been the cheap part for thirty years.

1. Mental estimate. Span-over-twenty for a steel beam depth. A tonne per square metre for a building's floor load. These are not calculations; they are the sense of scale that lets you notice a wrong answer. An engineer without them is defenceless against their own software.

2. Hand calculation. One equation, one answer, on one sheet. Fast, transparent, checkable by anyone in five minutes. Limited to structures a formula exists for.

3. Spreadsheet. Repetitive arithmetic with the numbers visible. Excellent for a load take-down or a schedule of member checks. Poor at anything with a chain of dependency, for reasons Lesson 4.3 spends its length on.

4. Script. Repetition with logic, and a record of what was done. Worth writing when the task will recur, when the number of cases is large, or when a person doing it by hand would get bored — boredom being an excellent predictor of error.

5. Dedicated design software. A program that knows what a portal frame is. Fast and reliable inside its remit; misleading outside it, because it will still produce an answer.

6. General finite element analysis. Any geometry, at the price of building and checking every assumption yourself.

7. Specialist nonlinear analysis. When the behaviour is genuinely nonlinear — large displacement, yielding, contact, tension-only. Expensive in time and in the expertise needed to judge the result.

8. Optimisation. When the question is 'which of these' rather than 'is this adequate'. Modules 14 to 17.

9. No model at all. Sometimes the right answer is to change the structure so the question does not arise. A layout that avoids torsion needs no torsion analysis.

The rule

Use the cheapest route that genuinely answers the question, and make sure you can check it by a different route.

The second half is the one that gets dropped. A finite element model checked by another finite element model built by the same person from the same assumptions has been checked against nothing.

Try it

Which method answers this question?

Twelve engineering questions. Choose a route, then read what that route would and would not tell you.

Question

Move through the twelve questions.

Your route

Choose the cheapest route you believe genuinely answers it.

Question 1. Is a 457 UB adequate for deflection on a 7.2 m simply supported span under 24 kN/m?

Choose a route to see what it would tell you. Nothing is revealed until you commit to an answer — deciding first is the skill being practised.

What this shows: The cheapest route that genuinely answers the question is the right route — and more than one answer is often defensible.

Worked example

Does this beam need a finite element model?

Given

A simply supported steel beam, 7.2 m span
Uniformly distributed load of 24 kN/m at the serviceability limit state
Deflection limit of span/360
No lateral restraint concerns: the beam is fully restrained by the slab

Find

The cheapest route that answers 'is a 457 × 191 × 67 UB adequate for deflection?'

Practice

A continuous three-span beam, spans 6 m, 8 m, 6 m, carries 18 kN/m. Estimate the hogging moment over the first internal support using the supplied preliminary estimate 0.100 wL² with the longer adjacent span. This coefficient is an exercise assumption, not an exact solution for these unequal spans. Give the answer in kN·m.

Check yourself

You need the peak stress at a re-entrant corner in a steel casting, to assess fatigue. Which route?

Check yourself

A question can be answered by a hand calculation, a frame model or a shell model. Which should be used?

Summary

There are nine routes, and computer time is the cheapest ingredient in all of them
State the question before choosing the route; a vague question gets an expensive answer
The check must come by a different route from the answer
'No model at all' is sometimes the correct engineering choice
An unnecessary model is a liability, not a spare