Material Insights · Understanding the Material World

Design, failure and safety: how engineering knowledge grows: Key terms, quiz and sources

All courses / Understanding Engineering / Module 12: Design, failure and safety: how engineering knowledge grows Wrap-up

Design, failure and safety: how engineering knowledge grows: review

Key terms

Design hypothesis
The view that a design is a prediction that a thing will carry its loads, which a successful life confirms only up to the loads seen so far.
Return period
The average time between events as large as a stated one; a 50-year wind has a 1 in 50 chance in any year.
Forensic engineering
The investigation of failures, working back from the broken parts and records to the decisions that led to them.
Normalisation of deviance
The drift by which a departure from the design, once survived, comes to be accepted as normal.
Selection by stand-in
Testing design variations by calculation, models and rigs instead of in service, so that bad ideas fail cheaply.
Scatter factor
A divisor applied to a test life because identical parts fail at different cycle counts.
Coefficient of variation
The standard deviation of a quantity as a fraction of its average, a measure of its spread.
Reliability index
The average margin between strength and load divided by its standard deviation; a higher index means a smaller chance of failure.
Safe life
A design approach that retires a part at its tested life divided by a scatter factor.
Fail-safe and damage tolerance
A design approach that assumes cracks will start, and ensures the structure survives them and that they are found by inspection.
Inspection interval
The time between inspections, set so that a crack has several chances to be found before it grows critical.
Prototype
A first working version built to be tested and revised, before production begins.
Test to failure
A test that loads a part until it breaks, to learn its margin and its way of failing.
Code or standard
An agreed document that sets loads, strengths, margins, details and tests, and records what past failures taught.
Partial factor
A separate margin applied to one quantity, such as a load or a material strength, in a limit-state code.
Risk
The chance of a harm multiplied by its size; targets are stated for the chance, matched to the consequence.
ALARP
As low as reasonably practicable: risk between the intolerable and broadly acceptable levels must be cut until further cuts cost grossly more than they gain.
Common-cause failure
A single fault, such as a shared power supply or design error, that disables several redundant parts together.
Failure rate, MTBF and FIT
The failure rate is failures per hour; MTBF is its inverse for a constant rate; FIT is failures per billion device-hours.
Bathtub curve
The pattern of failure rate over a product's life: high at the start, low and flat in use, rising with wear-out.

Try it in the table

  • Open 'Iron suspension bridge' and 'Cable-stayed bridge'. Pick a footbridge or road bridge you know and list the checks you would want before it reopens after repair, using the Morbi table in Part 2. Say which of them would matter just as much for the cable-stayed type.
  • Open 'Safety elevator' and 'Railway air brake'. For each, write down the fail-safe state and the failure that triggers it. Then find a third device in the Products table that is built to fail into a safe state, and say what the 'safe state' is.
  • Open 'Passenger car' and 'Three-point seat belt'. Take a mass from the page, work out the kinetic energy at 64 km/h as in estimate 9, and find how far the front must crumple to hold the average deceleration to 20 g. (The answer is about 0.8 m for the 1,400 kg of the module.)
  • Open 'Aeroplane' and 'Aircraft engine'. Choose a part that is inspected at intervals and use estimate 7 to set an inspection interval if a crack doubles in length every 2,000 cycles, can be found at 5 mm, and is critical at 60 mm.
  • Open 'LED lamp' and 'Smartphone'. Count the main parts in the bill of materials, assume 50 FIT for each, and work out the failure rate and MTBF of a controller board as in estimate 12. Then say which single part you would make redundant first, and what the copies would have in common.

Final quiz

A bridge has stood for 30 years and was designed for the 50-year wind. Which statement is best supported?

The chance of no 50-year wind in 30 years is about 55 per cent. A quiet record shows only that the worst day has not yet come.

By the investigators' figures, the as-built Hyatt Regency connection could carry about what share of the capacity the code required?

Average capacity of 18,600 lb against an expected 68,000 lb is 27 per cent. The unchecked change to the rod arrangement doubled the load on the weak detail.

What did the Comet tank test show?

The tank fuselage failed after 3,060 cycles in all, against a 10,000-cycle aim, with the crack starting at a corner of the forward escape hatch. Fatigue of whole pressurised structures had not been tested.

Why was the Challenger launch an extrapolation?

The coldest earlier joint was about 53 F against about 16 F calculated for the failed joint. Nothing in the flight record covered that range.

With strength known to 10 per cent and load to 25 per cent, what does raising the factor of safety from 2 to 3 do to the chance of failure?

The reliability index rises from 3.1 to 5.1, which lowers the chance of failure by thousands of times. The effect is steep but depends on the spreads being as assumed.

How does a railway air brake illustrate fail-safe design?

The failure of the working element, a burst hose or a parted coupling, puts the system in its safe state, the brakes on.

A controller board has 200 parts, each at 50 FIT, all needed. What is its approximate MTBF?

200 x 50 = 10,000 FIT = 10^-5 per hour, so MTBF is 100,000 hours, about 11.4 years. Series parts add their failure rates.

Which statement about codes and standards is best supported?

Codes compress experience into rules, and changes such as the 2002 revision of IS 1893 followed the Bhuj earthquake. They do not remove the need for checking the structure as built.
Next module: Cost, trade-offs and resources: why the best design is rarely the strongest →

Sources for this module