AERO61021 Chap.1 Why Engineers Experiment
Why Engineers Experiment
Why Engineers Experiment sets out where physical testing fits among the three ways engineers gather information: analysis, modelling and simulation, and experimentation. Analytical theory is elegant but reaches only simpler systems, and simulation is limited by computing power, so small-scale physics such as turbulence is usually modelled rather than resolved.
Experiments supply what neither can guarantee: benchmark data, certification evidence, a first look at a new concept and, increasingly, the validation of numerical models. The chapter separates verification, which checks that a model has been built and solved correctly, from validation, which checks that it represents reality well enough within its intended domain.
It explains why confidence in a model becomes expensive as it approaches certainty, why testing therefore clusters at the edges of an operating envelope, and how aircraft and motorsport examples showed models drifting where conditions were extreme.
It closes with practical habits: search earlier experimental papers, handbooks and databases before building a rig, estimate the expected answer in advance, and reject any result that breaks physical limits whatever produced it. Together these ideas justify the time and money a physical test demands.
What this chapter covers
- 01
Analysis, simulation and experimentation as information routes
- 02
The stages of a typical experiment from rig to conclusion
- 03
Verification versus validation of models
- 04
Why confidence in a model costs more near certainty
- 05
Testing at the edges of an operating envelope
- 06
Finding prior experiments, handbooks and data sources
- 07
Plausibility checks against physical limits
Worked example · free
Estimate the cost of extra model confidence
- 1Fix the constant from the known point: k = 25,000 x (1 - 0.80) = 5,000.
- 1At 90% confidence the cost is 5,000/0.10 = £50,000, twice the 80% figure, matching the proportion quoted in the module notes.
- 1At 96% confidence the cost is 5,000/0.04 = £125,000.
- 1The six points from 90% to 96% add £75,000, three times the whole cost of the first 80%, so test effort is best spent where the model is weakest, usually at the envelope edges.
Key terms
- Verification
- Checking that a program or analytical model is implemented correctly and solves the problem it was intended to solve.
- Validation
- Showing that a model is accurate enough for its intended use within its domain of applicability.
- Independent variable
- A parameter the experimenter changes deliberately in order to observe its effect on measured quantities.
- Operating envelope
- The full range of conditions over which a vehicle or system is designed to work.
Why Engineers Experiment FAQ
Why run experiments when CFD and FEA are so capable?
Simulations resolve only part of the physics and are limited by computing power, so they tend to depart from reality at the edges of a design envelope and on genuinely new problems. Experiments provide benchmark data, certification evidence and the comparison needed to validate a model, which is why most modern projects combine both.
Is matching one wind tunnel result enough to validate a model?
No. Agreement at a single operating condition shows the model is acceptable there, but it says nothing about other conditions. Validation builds confidence over a domain step by step, and the cost of adding confidence rises steeply as complete coverage is approached.
Where should I look before designing a new experiment?
Start with earlier experimental papers and reviews in journals such as Experiments in Fluids or Experimental Mechanics, then handbooks of experimental fluid and solid mechanics, the ESDU engineering data collection and National Physical Laboratory guidance. Experienced experimentalists can save weeks if you arrive with specific questions.
Assessment move
Treat this chapter as the frame that every later technique hangs on, and practise explaining it in your own words rather than memorising definitions. Start by drawing the experiment chain from memory: building the rig, installing and calibrating sensors, running the test, watching the reading respond, digitising and saving the signal, then analysing it and drawing a conclusion.
For each link, write one thing that can go wrong and how you would detect it; later chapters on calibration, sampling and acquisition give you the detail for each failure. Next, build a short list of contrasting statements about verification and validation and sort them quickly. Good practice items describe a real activity, such as a mesh study, a code review or a comparison with tunnel data, and ask which question it answers.
Work the confidence and cost relation with two or three invented budgets so the runaway near full confidence becomes intuitive, but remember it is an illustration that reproduces the proportions in the notes rather than a published formula.
When you read the guest lecture topics on industrial and structural testing, note where the speakers chose a physical test because a model could not be trusted, and keep those cases as evidence for coursework discussion. For the group report, prepare a paragraph that justifies your experiment against the alternatives of analysis and simulation, names what you will validate and states the limits of that validation.
Finally, adopt the plausibility habit now: before each lab session write the value you expect, the sign it should have and the physical limits it cannot cross, such as negative absolute pressure or temperatures below absolute zero. Compare the logged data with that note before you analyse anything.
This small routine catches reference mistakes, wiring faults and unit errors early, and it gives you a ready sentence for the discussion section explaining why your results are credible.