Module 17 · Lesson 17.1
Checking is not repeating
A second pass down the same road finds arithmetic slips and nothing else. What finds the errors that matter is a different road.
Why this matters
Structures very rarely fail because someone made an arithmetic mistake. They fail because a load was left out, a member was idealised as something it is not, or a number was carried in the wrong units. None of those is found by redoing the sum — the second attempt makes the same assumption the first one did, gets the same answer, and reports agreement. That is the trap this module exists to escape.
By the end of this lesson you should be able to
- Say why an independent route is what makes a check a check
- Run an order-of-magnitude check and interpret its bands honestly
- Audit a load path by equilibrium
- Use sanity bands to prompt questions rather than to pass or fail
What you should already know
- The integrated design of Module 16 — its building is the one checked here
- Load take-down and the governing check (Module 16)
- Actions and combinations (Module 2)
The two kinds of error
Errors in design divide sharply, and the division decides how to hunt them.
Errors of execution are slips: a transposed digit, a mis-keyed number, a line of a spreadsheet copied one row short. They are common, they are usually small, and a careful second pass finds them.
Errors of approach are different in kind: a load case nobody thought of, a beam idealised as simply supported when it is continuous, a span scaled off the wrong drawing, kilonewtons read as newtons. They are rarer and they are the ones that fail structures.
A second pass by the same route finds the first kind and is structurally incapable of finding the second, because it inherits the assumption that caused it.
That is the whole argument for independence. A check is worth its cost when it reaches the answer by a route that does not share the original's assumptions — a different idealisation, a different order of calculation, a rule of thumb, or simply equilibrium.
And the check does not need to be accurate. It needs to be independent. A rough answer by a different road is worth far more than a precise answer by the same one.
Worked example
Checking the Module 16 beam by a different route
Given
- The beam designed in Module 16: 350 × 650, 7.5 m span
- Design moment reported as 551 kNm; steel reported as 4 H32 (3217 mm²)
- The checker has the drawings and the brief, and has NOT seen the calculation
Find
Whether the reported values are believable.
Assumptions
- Rough routes only — the checker is not repeating the design
- Office loading on a 7.5 m × 5.0 m grid
Practice
A beam carries 75 kN/m over 7.5 m, simply supported. Estimate the design moment, in kNm.
Practice
Estimate the tension steel for a moment of 551 kNm with d = 589 mm, taking the lever arm as 0.95d and the steel stress as 0.87 × 500. Give the area in mm².
Practice
A calculation reports a beam moment of 5430 kNm where an approximate route gives 543 kNm. By what factor do they differ?
Check yourself
Why is a rough independent check often worth more than a careful repeat?
Equilibrium is the check you get for nothing
Add up everything applied to the structure. Add up everything arriving at the foundations. They must be equal.
This is worth stating on its own because of what it can find that nothing else can:
- a load applied to the slab and never passed to the beam;
- a tributary area counted twice, or a strip of floor belonging to no member at all;
- a self-weight added at the slab and again at the beam;
- an entire member omitted from the take-down.
No member check finds any of these, because each member is perfectly happy with the load it was given. Only the total knows.
Equilibrium does not care how the structure works, so a wrong idealisation cannot fool it. That is a rare property and it makes this the highest-value minute in checking.
Allow a small tolerance for rounding on the way down — a couple of per cent. A discrepancy larger than that is not rounding, and the sign tells you which way to look: less arriving than applied means something has been lost; more arriving than applied means something is being counted twice.
Sanity bands prompt questions, they do not pass or fail
Ordinary members land in ordinary ranges. A beam's span/depth ratio is usually between 10 and 20; a one-way slab's between 20 and 35. Tension steel in a beam runs about 0.4% to 2.5%. A concrete floor weighs 5 to 12 kN/m² all in.
These are not code limits, and a value outside one is not an error. It is unusual — and unusual things should have a reason the designer can state. "The beam is deep because it also carries a facade" is a complete answer. "I hadn't noticed" is not.
Two cautions matter more than the numbers.
Never quote a band you cannot justify. A band without a reason is superstition passed between engineers, and it will eventually be applied where it does not belong. Every band in this course carries its reason.
A value inside the band is not a pass. The Module 16 beam that failed its deflection check had a span/depth ratio of 15.3 — comfortably inside 10 to 20. Sanity bands catch the gross outlier and nothing finer.
Summary
- Errors of execution are found by repeating; errors of approach are not
- A check is worth its cost when its route does not share the design's assumptions
- Precision and reliability are unrelated — hunt factors, not percentages
- A lever arm of 0.95d and a stress of 0.87 fyk checks most flexural members in a minute
- Equilibrium finds what no member check can, and cannot be fooled by a wrong idealisation
- Sanity bands prompt questions; inside the band is not a pass
- State what a check SHARES with the design — that is where it is blind
This is educational material. It uses simplified examples to teach principles, and must not be relied on for real design or safety-critical decisions. Module overview and checkpoint