# Control and systems: feedback, stability and how machines are made to behave: Systems and the machines of today

Source page: https://www.materialinsights.org/learn/engineering/control-and-systems/p5/

[All courses](https://www.materialinsights.org/learn/) / [Understanding Engineering](https://www.materialinsights.org/learn/engineering/) / [Module 11: Control and systems: feedback, stability and how machines are made to behave](https://www.materialinsights.org/learn/engineering/control-and-systems/) Part 5 of 5

## Systems and the machines of today

A control loop never stands alone. It sits in a machine whose parts interact, in a plant with a slowest stage, in a network where one failure can spread. This part starts with how aircraft handling was turned into a specification, then asks how parts combine into systems, and ends with brakes, aircraft and a solar pump that use every idea in the module.

### 5.1 Before a quality can be designed in it must be turned into numbers: how aircraft handling became a specification

Every loop so far had a set point read off a gauge. Some targets are harder to write down. What a pilot wants is an aeroplane that feels right in the hand, and a feeling cannot be put on a drawing. Vincenti's history of American aircraft from 1918 to 1943 shows how that gap was closed.

In the early 1920s engineers at the US government's aeronautical laboratory at Langley Field fitted recording instruments to aircraft and measured, at each steady speed, the elevator angle and the stick force the pilot needed. These were numbers a designer could aim at, and they could be linked to what pilots reported. About two decades of flight testing followed, much of it by small teams of engineers and test pilots working side by side, and in 1942 and 1943 the US military services published the first written flying-quality requirements that new aircraft had to meet.

For control, the story carries three points. The first is about targets. Before anything can be designed or tuned, someone has to find a measurable quantity that stands for the need, as stick force against speed stood for a pilot's comfort. This is the step from a wish to a design target described in What engineers do: rules of thumb, the method and the kinds of engineering knowledge.

The second is about where the loop is drawn. Vincenti points out that engineers of the 1930s tended to treat the aircraft as a machine worked on from outside, while pilots sensed themselves as part of it through the stick and the seat. Modern practice counts pilot and aircraft as one closed-loop system.

The third is a trade-off. Inherent stability makes an aircraft return to steady flight by itself, but it also makes it resist the pilot's commands, so more stability buys calm at the price of agility. Since the 1970s some agile aircraft have been built slightly unstable on purpose and kept steady by Fly-by-wire flight control computers, so the stability comes from feedback rather than from the shape. The price is complete dependence on those computers and their sensors.
 Common misreading

#### Stable does not mean good to fly

It is tempting to rank aircraft, or any vehicle, by how stable they are. Stability is a tendency to return to the starting state, and to the machine a command is just one more push away from that state, so a design that shrugs off gusts also shrugs off its pilot. A heavily loaded lorry that holds its line in a crosswind is the same lorry that is slow to swerve round a pothole. The right amount depends on the task, and feedback has made it a setting chosen in software rather than one fixed by the shape of the wing and tail.

- Vincenti, What Engineers Know and How They Know It (Johns Hopkins University Press, 1990) · ch. 3, 'Establishment of Design Requirements: Flying-Quality Specifications for American Aircraft, 1918-1943', sections on Langley flight research from 1918 to 1936 and on the specifications of 1936 to 1943 (history stated briefly, in the module's own words)

### 5.2 A system is more than its parts: interactions, the slowest stage and the single point of failure decide how it behaves

A system is a set of parts, the connections between them and a purpose. Its behaviour comes from the connections as much as the parts. Interactions can be harmless or nasty. Flush a toilet while someone showers and the cold-water pressure dips, so the shower swings hot: two separate loops are coupled through the shared pipe. Two thermostats in one room, one heating and one cooling, will fight each other. Engineers draw the connections before they tune any single loop.

The next idea is the bottleneck. The output of a chain of stages is set by its slowest stage, however fast the others are. A faster filler on a bottling line does nothing if the labeller cannot keep up.

The third idea is redundancy: more than one part for a job, so that one can fail. The table shows the arithmetic, and a warning. Redundancy helps only if the failures are independent. Two pumps behind one power feeder fail together when the feeder does. Engineers call a part whose failure stops the whole system a single point of failure, and they hunt for them.

The warning also covers protection. On 30 July 2012 the northern grid collapsed, and on 31 July the northern, eastern and north-eastern grids went down together, with about 36,000 MW and then about 48,000 MW of load cut off. The official enquiry found that on both days the first trip was on the 400 kV Bina-Gwalior line, where a distance relay read heavy load as a fault and opened the line, although the line was not overheated; the committee called this an unintended operation. The network was already too weak to survive losing that line, so its flow moved onto neighbours, which tripped in turn. A safeguard set for one line had become a hazard to the whole, and the committee asked for an audit of protection settings across the system.

Data Worked estimate 6: a bottling line, a pumping station and the arithmetic of the slowest part

| Case | Working | Result |
|---|---|---|
| Line: filler 120, capper 150, labeller 90, cartoner 100 bottles per minute | line speed = slowest stage | 90 per minute |
| Replace the filler with a 200 per minute machine | slowest stage unchanged | still 90 per minute: nothing gained |
| Upgrade the labeller to 130 per minute | next slowest is the cartoner | 100 per minute, up 11 per cent |
| Series chain of four parts, each 99 per cent available | 0.99 x 0.99 x 0.99 x 0.99 | 96 per cent |
| Two pumps in parallel, each 95 per cent available | 1 - (0.05 x 0.05) | 99.75 per cent |
| Same pumps behind one 99 per cent power feeder | 0.99 x 0.9975 | 98.8 per cent: the common feeder is the weak point |

Availability values are illustrative round numbers, and every result is calculated. The method is the point: series chains multiply availabilities, parallel parts multiply failure chances, and a shared part limits both.

- Report of the Enquiry Committee on Grid Disturbance in Northern Region on 30th July 2012 and in Northern, Eastern and North-Eastern Region on 31st July 2012 (Ministry of Power, 16 August 2012) · full text at https://archive.prayaspune.org/collection/5/document/1144, sections on the disturbances of 30 and 31 July 2012 and the committee's findings

### 5.3 Today's machines wrap these ideas in silicon: anti-lock brakes, autopilots and a solar pump

Anti-lock braking is a loop whose set point is a slip, not a speed. The brake computer reads each wheel's speed from a toothed ring and compares it with the speed of the car, judged from all four wheels. When one wheel slows faster than the car could, it is about to lock, so valves drop that wheel's brake pressure, let it spin up and reapply, many times a second. The aim is to hold each tyre in a narrow band of slight slip. The first reason is steering: a front tyre that is sliding cannot turn the car. The second is distance, since rubber rolling with a little slip gives more retarding force than rubber skidding. Take 60 km/h (16.7 m/s) on a wet road with a friction coefficient of 0.6 near the best slip and 0.4 when locked. Distance = speed squared / (2 x coefficient x 9.8 m/s2): 278 / 11.8 = 24 m against 278 / 7.8 = 35 m (calculated). On loose gravel or snow the stop can be longer with the system than without, but the steering is kept.

An Autopilot uses nested loops, as in the robot joint. An outer loop compares altitude with the target and sets a pitch angle; an inner loop moves the elevator to hold that pitch. Sensors are gyroscopes and air-data probes. In Fly-by-wire flight control the pilot's stick is only a command to such loops, and several computers run in parallel and vote, since a single computer would be a single point of failure.

A solar pump controller shows many ideas at once. It converts the array's direct current into variable-frequency power for a motor, and tracks the array's best operating point as the sun moves by making small trial changes and keeping those that raise power. It also protects: it stops the pump if current falls, which signals a dry well, and when a float switch says the tank is full, the same idea as the float valve in Part 1. The table estimates the daily water.

Data Worked estimate 7: daily water from a 4.5 kW array on a farm borewell

| Step | Working | Result |
|---|---|---|
| Energy to lift 1 m3 through 40 m | 1,000 kg x 9.8 m/s2 x 40 m = 392 kJ | 0.109 kWh |
| Allow for pump and motor efficiency of 40 per cent | 0.109 / 0.40 | 0.27 kWh per m3 from the array |
| Daily array output at about 5 kWh per kW of panels in good sun | 4.5 kW x 5 | 22.5 kWh |
| Water per day | 22.5 / 0.27 | about 83 m3 (83,000 L) |
| Irrigation at 5 mm per day, 50 m3 per hectare | 83 / 50 | about 1.7 hectares |

Assumed: 40 m total head including friction and 40 per cent wire-to-water efficiency; every result is calculated. Check: a 3.7 kW pump at 40 per cent efficiency lifts about 14 m3/h from 40 m, and 6 sunny hours of that is about 80 m3, the same order. Pump power rises with the cube of speed, so the controller trades flow against sun with the affinity laws of Fluids: pressure, flow, lift and drag.
 Source: Bloomfield, How Things Work (Wiley, 2015), ch. 5, 'Fluids', pressure and pumping

- Bloomfield, How Things Work (Wiley, 2015) · ch. 2, 'The Laws of Motion, Part 2', section 2.2, 'Wheels', static and sliding friction
- Bloomfield, How Things Work (Wiley, 2015) · ch. 5, 'Fluids', section 5.2, 'Water Distribution', pumps and pressure
- Kosky et al., Exploring Engineering (Academic Press, 2013) · ch. 8, 'Computer Engineering', section 8.10, embedded computers

### See it in the table
[H₂O Water](https://www.materialinsights.org/commodities/water/)[Si Silicon](https://www.materialinsights.org/commodities/silicon/)

### Check your understanding

A line has stages running at 80, 120 and 100 units per minute. Upgrading only the 120 stage to 200 changes the line rate to:
 The slowest stage, 80, still sets the output. The upgrade has no effect until the 80 stage is improved.

Two identical pumps in parallel each work 90 per cent of the time but share one power feeder. Why is the pair less reliable than 99 per cent?
 Parallel pumps give 1 - 0.1 x 0.1 = 99 per cent only if failures are independent. A shared feeder fails both at once, so it caps system reliability.

Why does an anti-lock brake release the pressure briefly instead of holding the pedal force on?
 A locked tyre slides with lower friction and cannot steer. Brief releases let it spin up and stay near the slip where grip is greatest.

[← Part 4: Stability, delay and hunting: how a loop goes wrong](https://www.materialinsights.org/learn/engineering/control-and-systems/p4/)[Mark read · Wrap-up →](https://www.materialinsights.org/learn/engineering/control-and-systems/review/)

## Other sections

- [Module overview](https://www.materialinsights.org/learn/engineering/control-and-systems/index.md)
- [Open loop and closed loop: measure, compare, correct](https://www.materialinsights.org/learn/engineering/control-and-systems/p1/index.md)
- [Sensors, actuators and on-off control: the hardware of a loop](https://www.materialinsights.org/learn/engineering/control-and-systems/p2/index.md)
- [Proportional, integral, derivative: what each action fixes and what it costs](https://www.materialinsights.org/learn/engineering/control-and-systems/p3/index.md)
- [Stability, delay and hunting: how a loop goes wrong](https://www.materialinsights.org/learn/engineering/control-and-systems/p4/index.md)
- [Key terms, quiz and sources](https://www.materialinsights.org/learn/engineering/control-and-systems/review/index.md)
