Engineering & Technology · Sketches, Models & Technical Communication
Assumptions, simplification and checking model predictions
The constant-rate model predicts 5.00 L after 20 seconds. A comparison record gives 4.84 L. That difference is useful information. It could indicate that inflow was lower than assumed, that some water escaped, that timing differed or that the reading method was biased. Changing the equation immediately would skip the task of understanding the disagreement.
Use units and simple arithmetic from Measurement, Units & Engineering Evidence. This chapter introduces its own coordinates, projections and model vocabulary; multiplication and division are enough for the worked cases. The geometry, system, data and results are original teaching constructions. They are not observed experiments, build specifications or operating instructions. No conformance to a drafting, piping or NASA standard is claimed.
Back to the chapterAfter reading, you can
Expose assumptions, compare predictions with independent evidence and describe the conditions under which a model has actually been checked.
Make assumptions answerable
An assumption is something accepted as a basis for reasoning. Our model assumes 1 L initially, a constant incoming rate of 0.20 L/s, no outflow or leakage, and enough capacity that overflow does not occur during the interval. Its intended use here is an early estimate of stored volume from 0 to 20 s. It does not predict surface height, valve response or behaviour after the tank fills.
Each assumption suggests a check: record the initial volume, inspect whether an outlet was open, compare incoming volume over equal intervals and identify the usable capacity. The purpose guides which details matter. A small decorative mark on the tank may be irrelevant to this volume equation; an unnoticed outlet is not. Simplification is useful when it removes detail that does not materially change the intended answer.
Check implementation and representation separately
Verification asks whether the model has been implemented according to its specified rules. For our calculation, check units, arithmetic, the initial condition and a few known outputs. Validation asks how well the model represents the system for its intended use. That calls for a comparison with suitable observations or another justified referent, under recorded conditions.
The plotted comparison data are invented for this lesson: at 0, 5, 10, 15 and 20 s the listed volumes are 1.00, 1.98, 2.95, 3.91 and 4.84 L. They demonstrate the method of comparison; they are not a real validation experiment. The differences from the formula are 0.00, −0.02, −0.05, −0.09 and −0.16 L. The lower plot makes the growing disagreement easier to see.
Ask what the difference can support
Suppose, for this constructed exercise, the final reading has an assumed bound of ±0.05 L. Its interval is 4.79 to 4.89 L, which does not include the model’s 5.00 L prediction. The assumed measurement bound alone does not account for the difference. This does not identify the cause: timing, flow inputs, model omissions and the bound itself still need examination.
Adjusting a model to match a set of measurements is fitting, sometimes called model calibration. This use of calibration differs from the instrument calibration introduced in the preceding chapter: here we are choosing model parameters. Reproducing the same data is weaker evidence than predicting a new, suitable comparison kept separate from tuning. Likewise, acceptable agreement near one setting does not establish agreement at every setting. Record the checked range and the size of error that the intended decision can tolerate before using the results more widely.
Calculate a difference without explaining it away
At 20 s, predict V = 1 + 0.20 × 20 = 5.00 L. Subtract prediction from comparison value: 4.84 − 5.00 = −0.16 L. The negative sign means the comparison is below the prediction. As a fraction of predicted total volume, 0.16 ÷ 5.00 = 0.032, or 3.2%. State the denominator when reporting a percentage.
An average rate fitted to these endpoints is (4.84 − 1.00) ÷ 20 = 0.192 L/s. It exactly matches the two endpoints by construction. It does not prove that the rate was constant between them.
Record this as a candidate revised model, then plan a separate comparison over the intended time and operating range. Do not relabel the fitted endpoint as independent confirmation. Retain the original model and its disagreement so that another reviewer can follow why the revision was proposed.
Try the reasoning
A program gives 50 L at 20 s for the stated rule. Which check should begin the investigation?
Show the answer and reasoning
Verification: compare the program’s units and operations with V = 1 + 0.20t. The stated rule gives 5 L, so its implementation is already suspect.
Can the fitted average 0.192 L/s prove that flow stayed constant?
Show the answer and reasoning
No. The same total increase can arise from many changing flow histories. Additional observations are required to examine variation within the interval.
Does the hypothetical ±0.05 L final-reading bound cover the 0.16 L difference?
Show the answer and reasoning
No. The corresponding 4.79–4.89 L interval excludes 5.00 L. Other sources of disagreement still require examination; the calculation alone does not diagnose them.
What to carry forward
Keep assumptions, implementation checks and evidence comparisons visible. Agreement has a purpose and a range; disagreement calls for investigation.
References
Teaching resources and research supporting this lesson’s explanations and examples.