Engineering & Technology: Order-of-magnitude prediction

What you'll be able to do: Know roughly what the answer should be before you calculate, which catches the errors that matter and that no amount of careful arithmetic will.

Why engineers estimate and students don't

A student calculates a beam deflection of 4.7 metres and writes it down. A beam that deflects nearly five metres is not a beam; it's a rope. Nothing in their process stopped them, because their process was apply formula, report number.

Practising engineers estimate first, reflexively, and the estimate is a guard, It doesn't need to be close, it needs to be right to a factor of ten, because the errors worth catching are unit errors, setup errors and dropped factors, and all of those are wrong by orders of magnitude rather than percentages.

Careful arithmetic cannot catch a wrong setup. Estimation can, in about fifteen seconds.

The three things to predict

1. The order of magnitude. Millimetres or metres? Watts or kilowatts?
2. The direction. If I increase this, does that go up or down?
3. The sensitivity. Linear, squared, cubed, inverse? "Doubling the depth reduces deflection by a factor of eight" is a fact about the cube in the formula, and it's more useful than the formula.

Then calculate, and reconcile. A disagreement means one of them is wrong and you now know to look.

Before I calculate: ask me to predict the order of magnitude, the direction and the sensitivity, with a one-line reason for each. Wait. Then tell me the actual result and (if I was wrong) whether my reasoning was wrong or just my arithmetic.

The estimation toolkit

Dimensional analysis first. If the units don't work, nothing else matters, This catches more student errors than any other single check and takes seconds.

Anchor to something you know. A person is 80 kg. A car is 1.5 tonnes. Atmospheric pressure is 100 kPa. Steel is 8000 kg/m³ and about 200 GPa. A domestic circuit is a few kilowatts. Having twenty anchors makes estimation possible; having none makes it impossible.

Round brutally. π is 3, g is 10, everything is a power of ten. Precision is the enemy of the estimate.

Bound it. If you can't estimate, bracket it: it's more than X and less than Y. A bracket is often enough to catch the error.

Across the disciplines

Structural. Deflection should be span/300-ish for a serviceable beam. Anything wildly outside that is a setup error, not a design.

Mechanical. Estimate the torque, the power, the speed. A motor that needs 500 kW to turn a small shaft means a units error, almost always in RPM versus rad/s.

Electrical. Estimate current from power and voltage before analysing. Milliamps or amps? This catches component-value and decimal errors instantly.

Chemical. Estimate the heat of reaction and the temperature rise. A rise of 5000 K is a mass or molar error, and this is the check that prevents the dangerous kind of mistake.

Civil. Estimate the load. A floor is a few kPa; a crowd is a few hundred kilograms per square metre. Wrong by a factor of ten is a units error.

Fluids. Estimate the Reynolds number before choosing correlations, laminar or turbulent changes which equations apply at all. This is an estimate that selects the method.

Thermal. Estimate the time constant. Seconds, minutes or hours? Determines whether steady state is a reasonable assumption (article 01).

Software and systems. Estimate the data volume, the request rate, the latency budget. "How many rows?" before writing the query prevents the accidental cross join.

The habit that separates engineers from calculators

Ask, before every calculation: what's the biggest term, and why?

Engineering answers are usually dominated by one or two terms. Knowing which means you can sanity-check, simplify, and (most usefully) know where to spend effort. A student who computes all terms to four significant figures and cannot say which dominates has done arithmetic, not engineering.

Pitfalls

1. Estimating after calculating. Then it's rationalisation. Say it first.
2. Estimating precisely. An estimate to two significant figures is a slow calculation.
3. No anchors. Estimation is impossible without a stock of reference quantities. Build them deliberately.
4. Ignoring a disagreement. When the estimate and the calculation differ, find out which is wrong. That's the whole event.
5. Skipping the units check. Cheapest error-catching available and it's skipped constantly.
6. The tell: you've never caught an error this way. Either you're not estimating, or you're estimating after you've seen the answer.

Try this today

Before your next calculation, write three lines: order of magnitude, direction, sensitivity, with a reason each.

Then calculate. If they disagree, find out which was wrong before you do anything else.