# Thinking in estimates: units, scaling and educated guessing: Dimensions and easy cases: checking formulas and finding them

Source page: https://www.materialinsights.org/learn/engineering/thinking-in-estimates/p2/

[All courses](https://www.materialinsights.org/learn/) / [Understanding Engineering](https://www.materialinsights.org/learn/engineering/) / [Module 2: Thinking in estimates: units, scaling and educated guessing](https://www.materialinsights.org/learn/engineering/thinking-in-estimates/) Part 2 of 5

## Dimensions and easy cases: checking formulas and finding them

Units tell you whether a result is labelled correctly. Dimensions go further. They tell you what kinds of quantity a formula may contain, which makes them a filter for wrong formulas and, with a little care, a way to find right ones. Easy cases add a second filter: try the formula where you already know the answer. Together they let you trust or reject a result in a minute.

### 2.1 Only like can be added: dimensions catch wrong formulas

A unit is a particular scale for a quantity, such as the metre, the foot or the kilometre. A dimension is the kind of quantity underneath: a length, a time, a mass. Engineers write them with the letters L, T and M. Speed has dimensions L over T, whether you measure it in metres per second or kilometres per hour. Force has dimensions M x L over T squared, because force is mass times acceleration.

Three rules follow, and they are all one idea. Both sides of an equation must have the same dimensions. Only quantities of the same dimension can be added or subtracted. And anything inside a logarithm, an exponent or a sine must be a pure number, with no dimensions at all. Break any of these and the formula is wrong, whatever else is true of it.

Here is how this earns its keep. You remember that the power needed to lift water depends on the flow of water and the height, but you cannot recall whether gravity appears. Try mass flow multiplied by height: kilograms per second times metres. Power must have dimensions M x L squared over T cubed (a watt is a joule per second, and a joule is a newton-metre). The proposal has M x L over T, so it is short by L over T squared, which is exactly an acceleration. Gravity belongs in the formula. You have rebuilt the rule from the checking alone, and it matches the pump power calculation used in module 1.

Dimensions also explain why engineers like pure numbers. An efficiency of 60 per cent or a ratio of two lengths does not depend on the unit system, so it carries over from a model to a full-size machine.

Data Dimensions of quantities met in engineering (M mass, L length, T time)

| Quantity | Dimensions | SI unit |
|---|---|---|
| Speed | L / T | metre per second |
| Acceleration | L / T2 | metre per second squared |
| Force | M x L / T2 | newton |
| Energy or work | M x L2 / T2 | joule |
| Power | M x L2 / T3 | watt |
| Pressure | M / (L x T2) | pascal |
| Density | M / L3 | kilogram per cubic metre |

Each row can be rebuilt from the one above it: force is mass x acceleration, energy is force x length, power is energy / time.
 Source: Kosky et al., Exploring Engineering (Academic Press, 2013), ch. 2, 'Elements of Engineering Analysis', derived SI units; dimension notation added here  Common misreading

#### 'If the dimensions match, the formula is right'

Matching dimensions is necessary but not sufficient. Kinetic energy is half of mass times speed squared. The version with the half left out has exactly the same dimensions and is wrong by a factor of two. Dimensions cannot see pure numbers such as 2, one half or pi, and they cannot see a wrong sign. They are a filter that throws out many bad answers, not a proof of the good one. That is why the next two sections add further tests.

- Mahajan, Street-Fighting Mathematics (MIT Press, 2010) · ch. 1, 'Dimensions'
- Kosky et al., Exploring Engineering (Academic Press, 2013) · ch. 2, 'Elements of Engineering Analysis' (derived SI units)

### 2.2 Dimensional analysis can find a formula, up to a pure number

Now turn the filter into a tool. Take the overhead tank on a building roof, with a tap or small hole low down. How fast does water shoot out? Ask what the speed could depend on. The depth of water above the opening, h, is one candidate, with dimensions L. Gravity, g, is another, with dimensions L over T squared, since gravity is what drives the flow. Density, rho, is a third.

Density has a special property: it is the only candidate that contains mass. The speed we want has no mass in it, and nothing in the list can cancel density's mass. So density cannot appear in the formula. A heavier liquid is pushed harder but is also harder to move, and the two effects cancel.

Write the speed as a pure number C times g to some power a times h to some power b. Matching length gives a + b = 1. Matching time gives 2a = 1. So a and b are both one half, and v = C x the square root of (g x h). Nothing else is possible. The method cannot tell you C. A short energy argument, which module 3 will give you, finds C is about 1.4. Even without it, taking C as 1 puts the speed near 10 m/s for a tank 10 m above the tap. That already tells you doubling the height raises the speed only by about 40 per cent, and that water, petrol or oil would all leave at the same speed.

Real taps run slower because pipes, bends and valves eat some of the head. The table compares the head with what Indian guidance asks of town mains.

Data Worked estimate: speed and pressure from a rooftop tank 10 m above the tap

| Step | Working | Result |
|---|---|---|
| List what could matter | Speed v depends on gravity g and depth h; density cannot enter, since only it contains mass | v = C x g^a x h^b |
| Match dimensions | Length: a + b = 1. Time: 2a = 1 | a = 1/2, b = 1/2 |
| Formula | v = C x square root of (g x h), C a pure number between 1 and 1.4 |  |
| Put in numbers | g x h = 9.8 m/s2 x 10 m = 98 m2/s2; square root = 9.9 m/s | 10 m/s with C = 1, 14 m/s with C = 1.4 |
| Pressure behind the jet | density x g x h = 1,000 x 9.8 x 10 | 98,000 Pa, about 1 bar |
| Check against a real value | Indian water-supply guidance (CPHEEO) asks for a minimum residual head at the house connection of 7 m for one storey, 12 m for two and 17 m for three | A 10 m tank gives a head of the same size as a town supply; pipe friction takes part of it before the tap |

The CPHEEO figures are design minimums for the main at the ferrule, not pressures at a tap inside a building, so the comparison checks only the size of the answer. Friction loss in pipes depends on pipe size and length and is not estimated here.
 Source: Central Public Health and Environmental Engineering Organisation (CPHEEO), Manual on Water Supply and Treatment (Ministry of Urban Development, 1999), clause 2.2.8.3, as reproduced by the National Capital Region Planning Board, minimum residual pressure at the ferrule: 7 m for one storey, 12 m for two, 17 m for three, https://www.ncrpb.nic.in/NCRBP%20ADB-TA%207055/repository/pdf/clause/Clause.2.2.8.3e.pdf  Worked example

#### The method is only as good as the list of variables

Leave out a variable that matters and you get a neat wrong formula. For a thick oil in a narrow pipe, viscosity matters a great deal, and a list without it would fail. Add one that does not matter, such as the colour of the water, and the algebra simply finds no power for it. If two lengths are on the list, such as depth and hole diameter, their ratio is a pure number, and the formula might depend on it in a way dimensions cannot reveal. For a small hole in a large tank that ratio is tiny and drops out. Where it does not, you need an experiment or a limit.

- Mahajan, Street-Fighting Mathematics (MIT Press, 2010) · ch. 1, 'Dimensions' (the method of choosing variables and matching powers)
- Central Public Health and Environmental Engineering Organisation (CPHEEO), Manual on Water Supply and Treatment (Ministry of Urban Development, 1999), clause 2.2.8.3, as reproduced by the National Capital Region Planning Board · minimum residual pressure at the ferrule: 7 m for one storey, 12 m for two, 17 m for three, https://www.ncrpb.nic.in/NCRBP%20ADB-TA%207055/repository/pdf/clause/Clause.2.2.8.3e.pdf

### 2.3 Easy cases and limits test a formula where the answer is obvious

The second filter is the method of easy cases. Choose a simple or extreme setting for the quantities in a formula, where you can see the answer without calculating, and see whether the formula agrees.

Two taps fill a tank. One alone takes 20 minutes, the other alone takes 30. How long do both take together? Two formulas suggest themselves. Formula A is the average of the two times, 25 minutes. Formula B is the product of the times divided by their sum: 20 x 30 / 50 = 12 minutes. Test them on easy cases. If the second tap is dry, which is the same as its time going to infinity, the answer must be the first tap's own time. A fails, since the average runs off to infinity too. B gives the first tap's time. If both taps are identical, two of them should take half as long as one. A says the time is unchanged, which is absurd, since two taps cannot be slower than one. B halves it. Formula A would say 25 minutes for two taps when the faster alone needs 20. Formula B passes every test, and it also passes the dimensions test: time squared over time is time.

The same idea checks the tank-speed formula. With no gravity, nothing pushes the water, and the formula gives zero. With no depth, there is no head, and it gives zero again. Each limit agrees with common sense.

Use this on any model. In a spreadsheet for a plant, enter zero for the input and expect zero, or just the fixed part. Double the input and see whether the output doubles where it should.

Data Two candidate formulas for two taps filling a tank, tested on easy cases

| Test | What must happen | A: average, (t1 + t2) / 2 | B: t1 x t2 / (t1 + t2) |
|---|---|---|---|
| Second tap dry (t2 very large) | Time tends to t1 | Grows without limit: fails | Tends to t1: passes |
| Identical taps (t1 = t2 = 20) | Half of 20 = 10 minutes | 20 minutes: fails | 10 minutes: passes |
| Second tap instant (t2 tends to 0) | Time tends to 0 | Tends to t1 / 2: fails | Tends to 0: passes |
| Dimensions | Result is a time | Time: passes | Time squared over time: passes |
| The real case: 20 and 30 minutes | Less than 20 minutes | 25 minutes: fails | 12 minutes: passes |

 Source: Mahajan, Street-Fighting Mathematics (MIT Press, 2010), ch. 2, 'Easy cases' (the method of testing a formula at simple or extreme values); the two-tap example is this course's own  Case

#### A limit that caught a broken cell

A team builds a sheet for the monthly electricity bill of a small workshop: a fixed charge plus a rate times the units used. Setting the units to zero should leave just the fixed charge. The sheet showed zero, because a cell holding the fixed charge had been overwritten by mistake during editing. The check took ten seconds. Without it, the sheet would have under-billed every month, and nobody would have noticed until the first real invoice arrived. (An illustrative case, not a record of a particular firm.)

- Mahajan, Street-Fighting Mathematics (MIT Press, 2010) · ch. 2, 'Easy cases'

### See it in the table
[H₂O Water](https://www.materialinsights.org/commodities/water/)[El Electricity](https://www.materialinsights.org/energy/electricity/)

### Check your understanding

Why can the density of the liquid not appear in the formula for the speed of water leaving a hole in a tank, according to dimensional analysis?
 A formula for speed cannot contain mass, and nothing else in the list can cancel the mass in density, so density has to drop out.

A formula for the energy of a moving vehicle is proposed as mass x speed squared. It has the right dimensions. What does this show?
 Dimensions are blind to pure numbers, so a missing factor of one half would not be caught.

Which easy case would you try first on a proposed formula for the time two pumps take to empty a pond together?
 The limit with one pump doing nothing has an obvious answer, which any correct formula must reproduce.

[← Part 1: Why engineers estimate before they calculate](https://www.materialinsights.org/learn/engineering/thinking-in-estimates/p1/)[Mark read · Part 3: The big part first, then rounding, lumping and pictures →](https://www.materialinsights.org/learn/engineering/thinking-in-estimates/p3/)

## Other sections

- [Module overview](https://www.materialinsights.org/learn/engineering/thinking-in-estimates/index.md)
- [Why engineers estimate before they calculate](https://www.materialinsights.org/learn/engineering/thinking-in-estimates/p1/index.md)
- [The big part first, then rounding, lumping and pictures](https://www.materialinsights.org/learn/engineering/thinking-in-estimates/p3/index.md)
- [Scaling: why size changes everything](https://www.materialinsights.org/learn/engineering/thinking-in-estimates/p4/index.md)
- [When an estimate is good enough to decide](https://www.materialinsights.org/learn/engineering/thinking-in-estimates/p5/index.md)
- [Key terms, quiz and sources](https://www.materialinsights.org/learn/engineering/thinking-in-estimates/review/index.md)
