From a property claim to a measurement

“Stronger” needs an endpoint

A claim that one film is stronger could refer to tensile yield, maximum tensile stress, tear resistance or puncture response. Define the quantity, specimen and conditions before choosing equipment. The measurand is the quantity intended to be measured.

Two film strips reach maximum pulls of 12.0 N and 9.60 N. It is tempting to call the first film stronger. But a larger strip can carry a larger force without having a larger strength per area. There is another choice even after geometry is included: do we want the stress at a small specified strain, or the greatest engineering stress in the whole tensile record?

An endpoint is the defined observation or event at which a quantity is extracted. Here engineering stress is σ = F/A₀, with axial force F and original cross-sectional area A₀ perpendicular to that force. Engineering strain is ε = ΔL/L₀, where L₀ is the original separation of two gauge marks and ΔL is the increase in their separation. Applied force and stress over an area and Deformation and relative strain develop those quantities.

For the case below, choose maximum engineering tensile stress S_max = F_max/A₀ as the primary measurand. Keep a second, different endpoint σ₂: the engineering stress when the measured gauge strain first reaches +0.0200, or 2.00%, during the specified loading. Neither is automatically a yield stress, elastic modulus, tear resistance or puncture force. A well-defined experiment can answer either question; the label “stronger” cannot decide which was intended.

Specify the object and the response before collecting a number

Use two fictional uniform rectangular film specimens P and Q. P has original width w = 10.0 mm and thickness t = 0.0500 mm; Q has w = 8.00 mm and t = 0.0400 mm. Their original measured gauge lengths are both 20.0 mm. These are illustrative single-specimen records, not certified film grades or estimates of every specimen made by either process.

The defined comparison uses axial tension along the marked manufacturing direction, matched specimen preparation and conditioning, measured specimen temperature 23.0 °C, and the same monitored 50% relative humidity environment. The loading system maintains a gauge-strain rate of 0.0100 s⁻¹ in the compared region. The rate is a stipulated controlled specimen quantity, not inferred just from a crosshead speed setting. These chosen conditions define the teaching record; they are not a prescribed standard test procedure.

The record includes force, time and the changing separation of the gauge marks, with the force and length observations aligned in time. Alignment, gripping, local width and thickness, and where deformation or failure occurred are retained. A check that the instrument responds correctly is different from checking whether a grip slipped or the selected specimen represents the question.

P and Q are the comparison specimens; neither is an absolute strength standard. A matched reference specimen or a check on the force system can reveal changes in the measurement process, but cannot make different conditioning, direction or endpoints equivalent. Specimens, dimensions and test conditions continues those preparation questions. Here the supplied conditions and records are sufficient for the stated numerical illustration only.

An indication becomes a result through an explicit model

An indication is a value supplied by the measuring system. The force channel in this invented valid record reads 0.200 N high over the relevant range. A separately established correction is therefore c_F = −0.200 N, giving F = I_F + c_F, where I_F is the indicated force in newtons. This correction is supplied, not estimated by assuming that one specimen’s response is known. Its uncertainty belongs in the eventual result.

For a rectangular uniform gauge, A₀ = wt. P has A₀ = 0.500 mm² and Q has A₀ = 0.320 mm². The primary model is S_max = (I_F,max + c_F)/(wt). With force in N and area in mm², the result is in N/mm² = MPa. Original area is used; a stretched or locally thinned area would define a different stress measure.

The table summarises selected values from complete, inspected, unclipped loading records. Their stated maxima are resolved from those full records, not inferred to be global maxima merely because a sparse table contains no larger entry. For P, 12.20 − 0.200 = 12.00 N, so S_max = 12.00/0.500 = 24.0 MPa. For Q, 9.800 − 0.200 = 9.600 N, so S_max = 9.600/0.320 = 30.0 MPa.

The force ranking and normalised maximum-stress ranking differ: P carries 25% more maximum force than Q, while Q has 25% higher maximum engineering stress than P in these records. They use different denominators. That does not show Q is a better component under every load or that one specimen establishes a whole production population.

Now select the second endpoint. At measured ε = 0.0200, the corrected forces are 10.00 N for P and 5.120 N for Q. Their σ₂ values are 20.0 and 16.0 MPa. P is higher at this strain, even though Q has the higher S_max. At a 20.0 mm original gauge length, 2.00% corresponds to ΔL = 0.400 mm. A comparison that mixes P’s maximum stress with Q’s stress at 2% would not compare the same measurand.

Keep the units, correction, dimensions and endpoint with the result. The displayed significant figures alone do not state uncertainty. A measurement model also needs relevant effects such as nonuniform thickness, force calibration, gauge interpretation and how the endpoint was extracted. Calibration, uncertainty and repeatability develops the correction and uncertainty assessment in more depth.

A plausible number can come from the wrong motion or a limited channel

A crosshead is a moving part of the testing machine. Its travel need not equal the extension between marks on the specimen: grip motion, machine deformation, specimen portions outside the gauge or slippage can contribute. In a separate check, crosshead travel is 0.400 mm, but a 20.0 mm specimen gauge extends only 0.100 mm. The measured gauge strain is 0.100/20.0 = 0.00500, or 0.500%, not 2.00%.

The difference of 0.300 mm shows that the two movement definitions are not interchangeable. It does not by itself assign all that difference to grip slip; additional observations of the grips, specimen and machine are needed to locate it. Consequently, the force at 0.400 mm crosshead travel cannot be labelled the force at 2% specimen-gauge strain on this evidence.

Clipping creates a different problem. For a separate idealised demonstration with exact area A₀ = 0.500 mm², suppose a validated force-channel description is F_recorded = min(F_actual, 10.0 N) for positive force. The function min chooses the smaller value: all actual forces at or above 10.0 N are recorded as 10.0 N. This is a deliberately limited measurement channel, not a material law.

If the record reaches that ceiling, it shows the actual maximum was at least 10.0 N, so S_max ≥ 20.0 MPa in this exact-area demonstration. It does not establish S_max = 20.0 MPa or supply an upper bound. A true plateau at 10.0 N and a rising force above 10.0 N can produce the same clipped trace. Additional identical readings cannot reconstruct the missing peak.

A symmetric uncertainty added to the ceiling reading does not restore the absent data. Retain the limitation and obtain an appropriate measurement range or independent evidence rather than presenting an ordinary complete maximum. The earlier P/Q comparison explicitly excluded this fault. These are method-validity checks, not alternative explanations to hide inside a tidy average.

A reported value and an acceptance decision are different statements

For Q’s particular S_max result, now supply a complete independently justified combined standard uncertainty u_c = 1.00 MPa. Standard uncertainty expresses uncertainty on a standard-deviation scale. With specified coverage factor k = 2, expanded uncertainty is U = ku_c = 2.00 MPa. Report s = 30.0 MPa with U = 2.00 MPa and k = 2, together with the endpoint and conditions. The central estimate is denoted s here to distinguish it from the measurand S_max.

The corresponding symmetric uncertainty interval is s ± U, here 28.0–32.0 MPa. It is not a guaranteed hard bound or a measured range of different film specimens. A coverage probability requires assumptions and information about the uncertainty assessment; k = 2 alone is not a universal 95% claim. The supplied budget is not derived from the two specimen values or from the number of display decimals.

Imagine a deliberately chosen bench requirement S_max ≥ 29.0 MPa and a declared decision rule: accept this result as meeting the requirement only if s − U ≥ 29.0 MPa. The equality is included. With s = 30.0 and U = 2.00 MPa, the lower endpoint is 28.0 MPa, so the result is not accepted under this rule even though its central estimate exceeds 29.0 MPa.

This guarded decision moves the minimum acceptable measured estimate to 29.0 + U = 31.0 MPa for that uncertainty. It does not change the physical requirement from 29.0 MPa, change the specimen’s properties or prove the specimen’s true value is below the requirement. It defines how the available evidence is used. Another declared rule can make a different decision and carry different risks; this one is an invented example, not a mandatory standard or a safety criterion.

Suppose a new independently justified assessment of the same endpoint returns s = 30.0 MPa with U = 1.00 MPa and the same stated k. Now s − U = 29.0 MPa, which meets the chosen rule at equality. The central measured estimate did not rise. Better supported measurement information changed this particular decision. Merely reporting more decimal places would not justify the smaller U.

An uncertainty assessment assumes that the method addresses the intended quantity. It cannot turn tear initiation into uniaxial tensile strength, crosshead movement into verified gauge strain, or a clipped maximum into a fully observed one. Define the quantity and validate the measurement path before applying a decision rule.

Make the procedure answer the question

Specify geometry, material state, loading path, rate and endpoint. Record how raw indications become the reported value. A good procedure distinguishes an actual response from grip slippage, instrument limits or a calculation based on an unsuitable area.

A completed question could be: “What is the maximum engineering tensile stress of this identified film strip, referred to its measured original rectangular area, along this direction and under these conditioning and loading conditions, using a valid complete force record?” The result then includes the value, uncertainty and method record. A broader question about a process or population additionally needs an appropriate sampling design.

For a different use, define a different observable before borrowing the adjective. A puncture comparison might ask for the force at first through-thickness penetration with a specified probe shape, support opening, film state and loading rate. That force in N is a component-test endpoint; dividing by an arbitrary tensile area does not turn it into S_max. Decide which event matters and retain the raw evidence that identifies it.

Tensile curves, strength and ductility develops the physical interpretation of a complete tensile record. This lesson’s task is earlier: select the question, connect observations to the quantity and preserve what the method cannot resolve. Sampling, scale and extrapolation limits then asks how far the specimen result can reasonably be extended.

Worked examples

Try the reasoning

1. Is maximum load automatically tensile strength?

   Show the answer and reasoning

   No.

   It must be related to the specified cross-sectional area and definition.

2. Why keep raw records?

   Show the answer and reasoning

   They make processing and anomalies inspectable.

   A final number can hide clipping or slippage.

3. A new specimen has w = 6.00 mm, t = 0.0300 mm and maximum force indication 5.600 N with the same valid correction −0.200 N. Its complete unclipped record supplies that maximum. Find S_max and compare its maximum force with Q’s.

   Show the answer and reasoning

   A₀ = 0.180 mm², F_max = 5.400 N and S_max = 30.0 MPa. It has the same central S_max as Q but a smaller maximum force.

   Use the measured original geometry: 6.00 × 0.0300 = 0.180 mm², then (5.600 − 0.200)/0.180 = 30.0 N/mm². Equal engineering-stress endpoints need not mean equal strip load capacity or establish identical full material behaviour.

4. At crosshead travel 0.400 mm, a 20.0 mm specimen gauge extends only 0.100 mm. Find its engineering strain. Does the remaining 0.300 mm prove grip slip?

   Show the answer and reasoning

   The specimen-gauge strain is 0.500%; the remainder does not uniquely prove slip.

   Divide the measured gauge extension by its original length: 0.100/20.0 = 0.00500. Machine, grip and out-of-gauge specimen motion can contribute to the difference. Track the relevant parts before assigning a cause or labelling this the 2% endpoint.

5. In the idealised clipped-channel case, three successive recorded force values are 10.0 N and A₀ = 0.500 mm² exactly. Can S_max = 20.0 MPa be reported as a fully resolved maximum, or can averaging the three values supply it?

   Show the answer and reasoning

   No. The model supplies S_max ≥ 20.0 MPa, with no upper bound from that trace.

   The channel reports the same ceiling for every actual force at or above 10.0 N. Both an actual plateau and a higher peak can fit those observations. Averaging does not invert a many-to-one clipped response. This ideal bound omits geometry uncertainty by stipulation; it is not the earlier expanded-uncertainty report.

6. A separately justified result for the same endpoint is s = 31.5 MPa with U = 2.00 MPa and k = 2. Apply the same s − U ≥ 29.0 MPa rule. Does the decision supply a service lifetime?

   Show the answer and reasoning

   Its lower endpoint is 29.5 MPa, so the result is accepted under this rule. No service lifetime follows.

   31.5 − 2.00 = 29.5 MPa exceeds the declared limit. A single tensile endpoint and its uncertainty do not cover cyclic loading, environment or the complete component. No probability or safety certification was supplied by this arithmetic.

7. A report changes 30.0 MPa to 30.000 MPa without new data or model evidence and halves U. Is that a justified improvement in measurement information?

   Show the answer and reasoning

   No. Additional printed decimals alone do not reduce uncertainty.

   The force, dimensions, endpoint and other relevant contributions still need their justified uncertainty assessment. A changed decision near the limit must rest on improved evidence, not a formatting choice. A valid correction also does not automatically remove its own uncertainty.

8. A package film is called “more puncture resistant”. Propose one measurable comparison and explain why the P/Q tensile result cannot answer it alone.

   Show the answer and reasoning

   For example, compare force at first through-thickness penetration using matched probe geometry, support opening, film state and loading rate, with a recorded event criterion.

   State the actual object, preparation, direction where relevant, contact geometry, endpoint and how force/event observations are obtained. This test uses a different load path from uniaxial tension. Its result cannot be generated just by renaming S_max or dividing an unrelated force by an arbitrary area.

What to carry forward

• Define the measurand before measuring.
• Retain the connection from observations to results.
• Different endpoints can give different correct rankings for the same specimens.
• Specimen-gauge extension, machine travel and a force-channel ceiling are not interchangeable observations.
• A valid value, its uncertainty and a declared decision rule serve distinct roles.
• An uncertainty interval is not a guaranteed physical bound or a range of all future specimens.