Process Measurement, Instrumentation and Thermal Control

Learning objective: Relate temperature, pressure, residence-time distributions and defect mechanisms to observability, control and quality.

What this unit requires

Process quality involves dynamic hidden variables, measurement delay and disturbance sources. The response of one sensor or a first-order numerical model should not be promoted into a conclusion about entire melt history or an accepted-product population.

An engineering statement is useful only after its units, system boundary, assumptions and evidentiary status are defined. This unit deliberately distinguishes mathematical identity, conditional model prediction, laboratory observation and verified process claim. Calculated values in these pages are synthetic worked examples, not measured commercial-resin data. Seek grade- and application-specific property evidence before any equipment or product decision.

Four-lesson pathway

1. [Residence-time distributions and mixing](../lessons/unit-06-lesson-01.html) — [numerical model 16](../labs/lab-16.html)
2. [Thermal history and degradation kinetics](../lessons/unit-06-lesson-02.html) — [numerical model 17](../labs/lab-17.html)
3. [Temperature-loop dynamics and disturbance rejection](../lessons/unit-06-lesson-03.html) — [numerical model 18](../labs/lab-18.html)
4. [Defects, measurement systems and statistical control](../lessons/unit-06-lesson-04.html) — [numerical model 18](../labs/lab-18.html)

Demonstrate learning

• [Sixteen multi-part assessment questions](../assessments/unit-06.html): numerical reconstruction, equation audit, model challenge and design trade-off.
• [Unit-level engineering project](../projects/unit-06.html): a practical but fictional application with traceable mass, quality and evidence requirements.
• [Interactive offline laboratory manual](../LABORATORY-MANUAL.html): adjustable, source-visible calculations with CSV export.

Completion standard

Explain a material mechanism, reconstruct the featured calculation with consistent dimensions, test one limiting case and specify an experiment that could invalidate the conclusion. An equation that reproduces its own arithmetic has been software-verified; it has not thereby been validated against polymer specimens or factory operation.

Unit 06 · Lesson 01 — Residence-time distributions and mixing

[Unit 06 course map](../units/unit-06.html) · [Interactive simulation 16](../labs/lab-16.html) · [Assessment](../assessments/unit-06.html) · [Applied case](../projects/unit-06.html)

1. The polymer-processing question

Polymer melt residence history affects thermal exposure, additive mixing and degradation. Residence time is not merely barrel volume divided by throughput when channels have backmixing, dead zones or bypass. A tanks-in-series model provides an educational comparison of age distributions; a tracer step or pulse test is required for real equipment, with careful handling of tracer-solids compatibility and response deconvolution. Long tails can produce gels or discoloration even when mean residence time is acceptable. Investigate distribution width as well as its mean.

This lesson belongs to Process measurement and control, whose goal is to relate temperature, pressure, residence-time distributions and defect mechanisms to observability, control and quality. The challenge is to connect chemical structure or thermomechanical history to an output that can be measured, rather than reporting attractive numerical results without a clear object of study. Name the material family and the actual specimen state before using a correlation. A number obtained on dry pellets can be inappropriate for a moist, filled, highly oriented or degraded finished article.

Process quality involves dynamic hidden variables, measurement delay and disturbance sources. The response of one sensor or a first-order numerical model should not be promoted into a conclusion about entire melt history or an accepted-product population.

2. Mathematical description and physical quantities

Starting relationship: For n ideal equal CSTRs in series, variance σt²=τ²/n.

A formula is meaningful only with its variable definitions. Identify the dependent variable, its SI unit, the independent variables and the dimensional basis of every fitted coefficient. For a ratio, inspect positive denominators and the zero-input limit; for a rate law, distinguish a rate per mass from a rate per volume; for a thermal transition, use kelvin if a reciprocal temperature is used. Reconstruct the equation's origin from counting, conservation, constitutive behavior or an explicitly stated empirical fit. Those origins cannot be interchanged. In particular, a pure mass-balance identity may survive changes in polymer chemistry while a viscosity correlation usually does not.

Measurement latency and material age matter — An instrument at one barrel location is not a direct measurement of the molecular experience of every polymer parcel. Residence-time tails couple to oxidation and gel formation; slow thermocouples can conceal melt temperature overshoot. A good control strategy names what it observes, what it can actuate and what hidden state cannot be inferred from a setpoint alone.

Sketch a control volume, measurement chain, or molecular population appropriate to Residence-time distributions and mixing. Annotate the sketch with the actual controlled settings and the measured outputs, and mark each coefficient as an assumed teaching value, a fitted value or a possible independent measurement. The resulting sketch should make clear what extra measurement would transform a numerical demonstration into an experimentally testable model. Avoid assigning industrial meaning to a symbol solely because its unit appears plausible.

3. Worked numerical investigation

If mean residence time τ=120 s and n=4, the ideal residence-time standard deviation is 60 s. Increasing n to 9 narrows it to 40 s at the same τ, but does not eliminate a true stagnant zone.

Independent calculation procedure: Extract the supplied initial conditions into a table and convert all reported units before substituting numbers. Record at least one intermediate result; then recompute the output in a calculator or spreadsheet without copying the answer. Vary the first independent quantity by 10% while all others remain fixed. Compute both the absolute response and percentage response; decide whether a local linear approximation is adequate. Reverse the question by asking whether the underlying measurement has enough resolution to distinguish the two numerical responses.

Constraint check: Identify any physically impossible input: a negative molecule count, mass recovery above the feed, mass fraction exceeding unity, a temperature on an incorrect absolute scale, a stress beyond a known material domain, or a nonfinite denominator. Not all of these appear in this model; choose those that apply. Discuss whether the original example is conservative or optimistic for real material variation. Values here are synthetic to explain algebra and must not be presented as published resin property tables.

4. Model verification is not material validation

The tanks-in-series representation cannot automatically capture stagnant pockets, pressure dynamics, phase change or nonlinear residence-dependent viscosity.

Rearrange the equation to infer one input from the target output. If the rearrangement is ill-conditioned, explain how small measurement error may produce a large inferred-parameter error. Identify a pair of conceptually different mechanisms capable of moving the observed output in the same direction. A sound diagnosis cannot select one mechanism solely from that output; it needs an orthogonal measurement. This is the central protection against overinterpreting a polymer melt curve, a DSC peak, a strength measurement or a recycling-yield statistic.

Compare Residence-time distributions and mixing with Thermal history and degradation kinetics as a neighboring subject in this unit. Do the two relationships describe the same physical quantity or two independent descriptors? Would their characteristic times, spatial scales, material states or experimental procedures be comparable? Propose one experiment that measures both and write what a disagreement would imply. A second model is useful only if it provides independent information rather than repeating the same assumption in a new notation.

5. Evidence and experimental design

Use an appropriate tracer, include line dead-volume correction and compare measured E(t) with model family rather than assuming a perfect plug-flow response.

Measurement planning: Define specimens, sampling locations, conditioning, calibration method, measurement resolution and number of independent replicates. Specify one reference material or control and the expected direction of a genuine effect. Organize a data ledger with columns for parameter, unit, instrument, method, timestamp, operator, uncertainty and calibration status. Do not fit and validate a model on the same observations without acknowledging that the validation is not independent. Where the property depends on processing history, preserve the exact prior temperature, deformation and storage conditions.

A robust plan explicitly separates structural uncertainty, parameter uncertainty and observation uncertainty. Structural uncertainty asks whether important physics are absent; parameter uncertainty asks whether fitted numbers are poorly identified; observation uncertainty asks whether instrumentation or sampling is noisy or biased. Repeated sampling can reduce some random errors but cannot repair an incorrect equation or uncalibrated specimen history. For Residence-time distributions and mixing, name one of each, and predict which category will dominate extrapolation into an unmeasured processing regime.

6. Interactive investigation and a falsifiable prediction

Open [Laboratory 16](../labs/lab-16.html). Read its exact equation and domain, then predict the trend with the first slider before viewing the curve. Change only that slider at three positions, record numeric outputs and use CSV export for the complete sweep. Change a second parameter next; compare whether the response retains its direction and curvature. Treat visible invalid-condition warnings as part of the lesson, not software errors to ignore. The simulation operates entirely in the browser with no external requests.

Investigation: Contrast n=1 and n=8 response widths while documenting evidence required for a real extruder.

Create a two-column evidence note: one column lists predictions directly supported by the stated mathematical relationship; the second lists interpretations that require independent polymer measurements. State what observation would most clearly falsify the assumed relation. If no practical experiment is yet proposed, the numerical model is a consistency exercise rather than a validated predictor. Do not infer equipment tolerances, worker-safety margins or compliance from fictional sliders.

7. Design decision with limitations

A process engineer evaluating Residence-time distributions and mixing faces at least two objectives: a desired quantity or geometry and the assurance that material quality remains fit for intended service. Build a trade-off table containing accepted-product quality, throughput or test cost, risk of extrapolation and measurement effort. For each alternative, name the assumption on which the apparent advantage depends. Identify a stop condition, such as failure to reproduce the response with a new batch or detection of a qualitatively different material phase or deformation regime. This makes the decision conditional and auditable.

Read the technical reference titles on the [references page](../REFERENCES.html) and check whether they discuss terminology, measurement method or product specifications. A terminology reference cannot supply experimental constants; a standard's title and abstract do not grant permission to claim full procedural compliance. The book is an editorial draft and a verified industrial decision requires current complete standards, manufacturer information and professional review.

8. Independent lesson problems

1. Derive or reconstruct For n ideal equal CSTRs in series, variance σt²=τ²/n from the most appropriate physical or statistical principle. Define every symbol and every unit, then identify a limiting case.
2. Reproduce this lesson's worked example and change one input by ±10%; report output changes with justified significant figures.
3. Identify two competing physical explanations for a deviation from the model, and one measurement that distinguishes them.
4. Design a sampling plan that tests applicability after a change in material grade, temperature or processing history. Show how you would preserve calibration traceability.
5. Write a six-sentence engineering review: objective, model result, credible restriction, alternative, required experiment and conditional conclusion.

Next: return to the [unit pathway](../units/unit-06.html) or complete the [unit assessment](../assessments/unit-06.html).

Unit 06 · Lesson 02 — Thermal history and degradation kinetics

[Unit 06 course map](../units/unit-06.html) · [Interactive simulation 17](../labs/lab-17.html) · [Assessment](../assessments/unit-06.html) · [Applied case](../projects/unit-06.html)

1. The polymer-processing question

Polymer molecular structure can change during overheating, oxygen exposure, hydrolysis and repeated reprocessing. A first-order surrogate for retained property or mass is useful for sensitivity analysis, but the observable might be tensile strength, molar mass or functional-group content and each can follow different kinetics. Exponential Arrhenius temperature sensitivity must be applied using kelvin. Long residence tails magnify damage; low-viscosity product may reflect chain scission rather than improved processing. Experimental characterization is indispensable.

This lesson belongs to Process measurement and control, whose goal is to relate temperature, pressure, residence-time distributions and defect mechanisms to observability, control and quality. The challenge is to connect chemical structure or thermomechanical history to an output that can be measured, rather than reporting attractive numerical results without a clear object of study. Name the material family and the actual specimen state before using a correlation. A number obtained on dry pellets can be inappropriate for a moist, filled, highly oriented or degraded finished article.

Process quality involves dynamic hidden variables, measurement delay and disturbance sources. The response of one sensor or a first-order numerical model should not be promoted into a conclusion about entire melt history or an accepted-product population.

2. Mathematical description and physical quantities

Starting relationship: First-order retained fraction f=exp(−k0 t exp[−Ea/(RT)]).

A formula is meaningful only with its variable definitions. Identify the dependent variable, its SI unit, the independent variables and the dimensional basis of every fitted coefficient. For a ratio, inspect positive denominators and the zero-input limit; for a rate law, distinguish a rate per mass from a rate per volume; for a thermal transition, use kelvin if a reciprocal temperature is used. Reconstruct the equation's origin from counting, conservation, constitutive behavior or an explicitly stated empirical fit. Those origins cannot be interchanged. In particular, a pure mass-balance identity may survive changes in polymer chemistry while a viscosity correlation usually does not.

Measurement latency and material age matter — An instrument at one barrel location is not a direct measurement of the molecular experience of every polymer parcel. Residence-time tails couple to oxidation and gel formation; slow thermocouples can conceal melt temperature overshoot. A good control strategy names what it observes, what it can actuate and what hidden state cannot be inferred from a setpoint alone.

Sketch a control volume, measurement chain, or molecular population appropriate to Thermal history and degradation kinetics. Annotate the sketch with the actual controlled settings and the measured outputs, and mark each coefficient as an assumed teaching value, a fitted value or a possible independent measurement. The resulting sketch should make clear what extra measurement would transform a numerical demonstration into an experimentally testable model. Avoid assigning industrial meaning to a symbol solely because its unit appears plausible.

3. Worked numerical investigation

For a rate coefficient k=0.005 s⁻¹ at the specified temperature and hold t=100 s, retained fraction exp(−0.5)≈0.607. This is only a kinetic proxy and is not a safety threshold.

Independent calculation procedure: Extract the supplied initial conditions into a table and convert all reported units before substituting numbers. Record at least one intermediate result; then recompute the output in a calculator or spreadsheet without copying the answer. Vary the first independent quantity by 10% while all others remain fixed. Compute both the absolute response and percentage response; decide whether a local linear approximation is adequate. Reverse the question by asking whether the underlying measurement has enough resolution to distinguish the two numerical responses.

Constraint check: Identify any physically impossible input: a negative molecule count, mass recovery above the feed, mass fraction exceeding unity, a temperature on an incorrect absolute scale, a stress beyond a known material domain, or a nonfinite denominator. Not all of these appear in this model; choose those that apply. Discuss whether the original example is conservative or optimistic for real material variation. Values here are synthetic to explain algebra and must not be presented as published resin property tables.

4. Model verification is not material validation

One Arrhenius channel does not account for complex stabilizer chemistry, autocatalysis, chain branching, crosslinking or spatial gradients.

Rearrange the equation to infer one input from the target output. If the rearrangement is ill-conditioned, explain how small measurement error may produce a large inferred-parameter error. Identify a pair of conceptually different mechanisms capable of moving the observed output in the same direction. A sound diagnosis cannot select one mechanism solely from that output; it needs an orthogonal measurement. This is the central protection against overinterpreting a polymer melt curve, a DSC peak, a strength measurement or a recycling-yield statistic.

Compare Thermal history and degradation kinetics with Temperature-loop dynamics and disturbance rejection as a neighboring subject in this unit. Do the two relationships describe the same physical quantity or two independent descriptors? Would their characteristic times, spatial scales, material states or experimental procedures be comparable? Propose one experiment that measures both and write what a disagreement would imply. A second model is useful only if it provides independent information rather than repeating the same assumption in a new notation.

5. Evidence and experimental design

Measure molecular weight, carbonyl bands or mechanical retention after controlled temperature-time histories; include oxygen and moisture conditions.

Measurement planning: Define specimens, sampling locations, conditioning, calibration method, measurement resolution and number of independent replicates. Specify one reference material or control and the expected direction of a genuine effect. Organize a data ledger with columns for parameter, unit, instrument, method, timestamp, operator, uncertainty and calibration status. Do not fit and validate a model on the same observations without acknowledging that the validation is not independent. Where the property depends on processing history, preserve the exact prior temperature, deformation and storage conditions.

A robust plan explicitly separates structural uncertainty, parameter uncertainty and observation uncertainty. Structural uncertainty asks whether important physics are absent; parameter uncertainty asks whether fitted numbers are poorly identified; observation uncertainty asks whether instrumentation or sampling is noisy or biased. Repeated sampling can reduce some random errors but cannot repair an incorrect equation or uncalibrated specimen history. For Thermal history and degradation kinetics, name one of each, and predict which category will dominate extrapolation into an unmeasured processing regime.

6. Interactive investigation and a falsifiable prediction

Open [Laboratory 17](../labs/lab-17.html). Read its exact equation and domain, then predict the trend with the first slider before viewing the curve. Change only that slider at three positions, record numeric outputs and use CSV export for the complete sweep. Change a second parameter next; compare whether the response retains its direction and curvature. Treat visible invalid-condition warnings as part of the lesson, not software errors to ignore. The simulation operates entirely in the browser with no external requests.

Investigation: Compare exposures that have equal mean residence but different long-tail fractions.

Create a two-column evidence note: one column lists predictions directly supported by the stated mathematical relationship; the second lists interpretations that require independent polymer measurements. State what observation would most clearly falsify the assumed relation. If no practical experiment is yet proposed, the numerical model is a consistency exercise rather than a validated predictor. Do not infer equipment tolerances, worker-safety margins or compliance from fictional sliders.

7. Design decision with limitations

A process engineer evaluating Thermal history and degradation kinetics faces at least two objectives: a desired quantity or geometry and the assurance that material quality remains fit for intended service. Build a trade-off table containing accepted-product quality, throughput or test cost, risk of extrapolation and measurement effort. For each alternative, name the assumption on which the apparent advantage depends. Identify a stop condition, such as failure to reproduce the response with a new batch or detection of a qualitatively different material phase or deformation regime. This makes the decision conditional and auditable.

Read the technical reference titles on the [references page](../REFERENCES.html) and check whether they discuss terminology, measurement method or product specifications. A terminology reference cannot supply experimental constants; a standard's title and abstract do not grant permission to claim full procedural compliance. The book is an editorial draft and a verified industrial decision requires current complete standards, manufacturer information and professional review.

8. Independent lesson problems

1. Derive or reconstruct First-order retained fraction f=exp(−k0 t exp[−Ea/(RT)]) from the most appropriate physical or statistical principle. Define every symbol and every unit, then identify a limiting case.
2. Reproduce this lesson's worked example and change one input by ±10%; report output changes with justified significant figures.
3. Identify two competing physical explanations for a deviation from the model, and one measurement that distinguishes them.
4. Design a sampling plan that tests applicability after a change in material grade, temperature or processing history. Show how you would preserve calibration traceability.
5. Write a six-sentence engineering review: objective, model result, credible restriction, alternative, required experiment and conditional conclusion.

Next: return to the [unit pathway](../units/unit-06.html) or complete the [unit assessment](../assessments/unit-06.html).

Unit 06 · Lesson 03 — Temperature-loop dynamics and disturbance rejection

[Unit 06 course map](../units/unit-06.html) · [Interactive simulation 18](../labs/lab-18.html) · [Assessment](../assessments/unit-06.html) · [Applied case](../projects/unit-06.html)

1. The polymer-processing question

Barrel and die temperature respond to heater power, cooling, thermal mass, sensor lag and material flow. A first-order model can describe approximate step responses around one operating point; it cannot reproduce all heater dead zones, exothermic viscous heating and thermal coupling. Controller tuning must account for sensor placement and delay, and apparent stability in a simulator does not establish product thermal uniformity. Treat pressure and residence-time changes as disturbances that might shift material temperature even when a thermocouple reads the same setpoint.

This lesson belongs to Process measurement and control, whose goal is to relate temperature, pressure, residence-time distributions and defect mechanisms to observability, control and quality. The challenge is to connect chemical structure or thermomechanical history to an output that can be measured, rather than reporting attractive numerical results without a clear object of study. Name the material family and the actual specimen state before using a correlation. A number obtained on dry pellets can be inappropriate for a moist, filled, highly oriented or degraded finished article.

Process quality involves dynamic hidden variables, measurement delay and disturbance sources. The response of one sensor or a first-order numerical model should not be promoted into a conclusion about entire melt history or an accepted-product population.

2. Mathematical description and physical quantities

Starting relationship: First-order surrogate T(t)=Tset+(T0−Tset)exp(−t/τ).

A formula is meaningful only with its variable definitions. Identify the dependent variable, its SI unit, the independent variables and the dimensional basis of every fitted coefficient. For a ratio, inspect positive denominators and the zero-input limit; for a rate law, distinguish a rate per mass from a rate per volume; for a thermal transition, use kelvin if a reciprocal temperature is used. Reconstruct the equation's origin from counting, conservation, constitutive behavior or an explicitly stated empirical fit. Those origins cannot be interchanged. In particular, a pure mass-balance identity may survive changes in polymer chemistry while a viscosity correlation usually does not.

Measurement latency and material age matter — An instrument at one barrel location is not a direct measurement of the molecular experience of every polymer parcel. Residence-time tails couple to oxidation and gel formation; slow thermocouples can conceal melt temperature overshoot. A good control strategy names what it observes, what it can actuate and what hidden state cannot be inferred from a setpoint alone.

Sketch a control volume, measurement chain, or molecular population appropriate to Temperature-loop dynamics and disturbance rejection. Annotate the sketch with the actual controlled settings and the measured outputs, and mark each coefficient as an assumed teaching value, a fitted value or a possible independent measurement. The resulting sketch should make clear what extra measurement would transform a numerical demonstration into an experimentally testable model. Avoid assigning industrial meaning to a symbol solely because its unit appears plausible.

3. Worked numerical investigation

For T0=180°C, Tset=200°C and τ=80 s, after 160 s the model predicts T≈197.3°C. The equation assumes one constant setpoint, not a complete PID controller.

Independent calculation procedure: Extract the supplied initial conditions into a table and convert all reported units before substituting numbers. Record at least one intermediate result; then recompute the output in a calculator or spreadsheet without copying the answer. Vary the first independent quantity by 10% while all others remain fixed. Compute both the absolute response and percentage response; decide whether a local linear approximation is adequate. Reverse the question by asking whether the underlying measurement has enough resolution to distinguish the two numerical responses.

Constraint check: Identify any physically impossible input: a negative molecule count, mass recovery above the feed, mass fraction exceeding unity, a temperature on an incorrect absolute scale, a stress beyond a known material domain, or a nonfinite denominator. Not all of these appear in this model; choose those that apply. Discuss whether the original example is conservative or optimistic for real material variation. Values here are synthetic to explain algebra and must not be presented as published resin property tables.

4. Model verification is not material validation

No dead-time, saturation, actuator limits, coupling or controller dynamics are represented, so model agreement does not justify final control tuning.

Rearrange the equation to infer one input from the target output. If the rearrangement is ill-conditioned, explain how small measurement error may produce a large inferred-parameter error. Identify a pair of conceptually different mechanisms capable of moving the observed output in the same direction. A sound diagnosis cannot select one mechanism solely from that output; it needs an orthogonal measurement. This is the central protection against overinterpreting a polymer melt curve, a DSC peak, a strength measurement or a recycling-yield statistic.

Compare Temperature-loop dynamics and disturbance rejection with Defects, measurement systems and statistical control as a neighboring subject in this unit. Do the two relationships describe the same physical quantity or two independent descriptors? Would their characteristic times, spatial scales, material states or experimental procedures be comparable? Propose one experiment that measures both and write what a disagreement would imply. A second model is useful only if it provides independent information rather than repeating the same assumption in a new notation.

5. Evidence and experimental design

Collect heat-up/cool-down step tests and feed-rate disturbances; compare thermocouple calibration and melt-temperature measurements.

Measurement planning: Define specimens, sampling locations, conditioning, calibration method, measurement resolution and number of independent replicates. Specify one reference material or control and the expected direction of a genuine effect. Organize a data ledger with columns for parameter, unit, instrument, method, timestamp, operator, uncertainty and calibration status. Do not fit and validate a model on the same observations without acknowledging that the validation is not independent. Where the property depends on processing history, preserve the exact prior temperature, deformation and storage conditions.

A robust plan explicitly separates structural uncertainty, parameter uncertainty and observation uncertainty. Structural uncertainty asks whether important physics are absent; parameter uncertainty asks whether fitted numbers are poorly identified; observation uncertainty asks whether instrumentation or sampling is noisy or biased. Repeated sampling can reduce some random errors but cannot repair an incorrect equation or uncalibrated specimen history. For Temperature-loop dynamics and disturbance rejection, name one of each, and predict which category will dominate extrapolation into an unmeasured processing regime.

6. Interactive investigation and a falsifiable prediction

Open [Laboratory 18](../labs/lab-18.html). Read its exact equation and domain, then predict the trend with the first slider before viewing the curve. Change only that slider at three positions, record numeric outputs and use CSV export for the complete sweep. Change a second parameter next; compare whether the response retains its direction and curvature. Treat visible invalid-condition warnings as part of the lesson, not software errors to ignore. The simulation operates entirely in the browser with no external requests.

Investigation: Vary thermal time constant and find when sensor placement could make a simple step test misleading.

Create a two-column evidence note: one column lists predictions directly supported by the stated mathematical relationship; the second lists interpretations that require independent polymer measurements. State what observation would most clearly falsify the assumed relation. If no practical experiment is yet proposed, the numerical model is a consistency exercise rather than a validated predictor. Do not infer equipment tolerances, worker-safety margins or compliance from fictional sliders.

7. Design decision with limitations

A process engineer evaluating Temperature-loop dynamics and disturbance rejection faces at least two objectives: a desired quantity or geometry and the assurance that material quality remains fit for intended service. Build a trade-off table containing accepted-product quality, throughput or test cost, risk of extrapolation and measurement effort. For each alternative, name the assumption on which the apparent advantage depends. Identify a stop condition, such as failure to reproduce the response with a new batch or detection of a qualitatively different material phase or deformation regime. This makes the decision conditional and auditable.

Read the technical reference titles on the [references page](../REFERENCES.html) and check whether they discuss terminology, measurement method or product specifications. A terminology reference cannot supply experimental constants; a standard's title and abstract do not grant permission to claim full procedural compliance. The book is an editorial draft and a verified industrial decision requires current complete standards, manufacturer information and professional review.

8. Independent lesson problems

1. Derive or reconstruct First-order surrogate T(t)=Tset+(T0−Tset)exp(−t/τ) from the most appropriate physical or statistical principle. Define every symbol and every unit, then identify a limiting case.
2. Reproduce this lesson's worked example and change one input by ±10%; report output changes with justified significant figures.
3. Identify two competing physical explanations for a deviation from the model, and one measurement that distinguishes them.
4. Design a sampling plan that tests applicability after a change in material grade, temperature or processing history. Show how you would preserve calibration traceability.
5. Write a six-sentence engineering review: objective, model result, credible restriction, alternative, required experiment and conditional conclusion.

Next: return to the [unit pathway](../units/unit-06.html) or complete the [unit assessment](../assessments/unit-06.html).

Unit 06 · Lesson 04 — Defects, measurement systems and statistical control

[Unit 06 course map](../units/unit-06.html) · [Interactive simulation 18](../labs/lab-18.html) · [Assessment](../assessments/unit-06.html) · [Applied case](../projects/unit-06.html)

1. The polymer-processing question

Melt fracture, splay, weld lines, sink, voids, gels, haze and color drift have overlapping causes; a defect image is not a unique diagnosis. Reliable troubleshooting ranks hypotheses from measurable variables and separates evidence from anecdotes. Statistical process control tracks shifts only after measurement-system analysis and a stable operating definition. A percent reject rate needs a denominator, detection sensitivity and sampling plan. Even a low average reject rate can hide correlated failures after start-up or material-lot changes.

This lesson belongs to Process measurement and control, whose goal is to relate temperature, pressure, residence-time distributions and defect mechanisms to observability, control and quality. The challenge is to connect chemical structure or thermomechanical history to an output that can be measured, rather than reporting attractive numerical results without a clear object of study. Name the material family and the actual specimen state before using a correlation. A number obtained on dry pellets can be inappropriate for a moist, filled, highly oriented or degraded finished article.

Process quality involves dynamic hidden variables, measurement delay and disturbance sources. The response of one sensor or a first-order numerical model should not be promoted into a conclusion about entire melt history or an accepted-product population.

2. Mathematical description and physical quantities

Starting relationship: Observed reject fraction p̂=number rejected/number inspected.

A formula is meaningful only with its variable definitions. Identify the dependent variable, its SI unit, the independent variables and the dimensional basis of every fitted coefficient. For a ratio, inspect positive denominators and the zero-input limit; for a rate law, distinguish a rate per mass from a rate per volume; for a thermal transition, use kelvin if a reciprocal temperature is used. Reconstruct the equation's origin from counting, conservation, constitutive behavior or an explicitly stated empirical fit. Those origins cannot be interchanged. In particular, a pure mass-balance identity may survive changes in polymer chemistry while a viscosity correlation usually does not.

Measurement latency and material age matter — An instrument at one barrel location is not a direct measurement of the molecular experience of every polymer parcel. Residence-time tails couple to oxidation and gel formation; slow thermocouples can conceal melt temperature overshoot. A good control strategy names what it observes, what it can actuate and what hidden state cannot be inferred from a setpoint alone.

Sketch a control volume, measurement chain, or molecular population appropriate to Defects, measurement systems and statistical control. Annotate the sketch with the actual controlled settings and the measured outputs, and mark each coefficient as an assumed teaching value, a fitted value or a possible independent measurement. The resulting sketch should make clear what extra measurement would transform a numerical demonstration into an experimentally testable model. Avoid assigning industrial meaning to a symbol solely because its unit appears plausible.

3. Worked numerical investigation

If 18 parts fail inspection out of 600 inspected, p̂=0.03. A binomial standard error sqrt[p̂(1−p̂)/600] is approximately 0.0070, if independent sampling assumptions hold.

Independent calculation procedure: Extract the supplied initial conditions into a table and convert all reported units before substituting numbers. Record at least one intermediate result; then recompute the output in a calculator or spreadsheet without copying the answer. Vary the first independent quantity by 10% while all others remain fixed. Compute both the absolute response and percentage response; decide whether a local linear approximation is adequate. Reverse the question by asking whether the underlying measurement has enough resolution to distinguish the two numerical responses.

Constraint check: Identify any physically impossible input: a negative molecule count, mass recovery above the feed, mass fraction exceeding unity, a temperature on an incorrect absolute scale, a stress beyond a known material domain, or a nonfinite denominator. Not all of these appear in this model; choose those that apply. Discuss whether the original example is conservative or optimistic for real material variation. Values here are synthetic to explain algebra and must not be presented as published resin property tables.

4. Model verification is not material validation

Binomial intervals can be misleading under clustering, inconsistent inspection or changing defect criteria.

Rearrange the equation to infer one input from the target output. If the rearrangement is ill-conditioned, explain how small measurement error may produce a large inferred-parameter error. Identify a pair of conceptually different mechanisms capable of moving the observed output in the same direction. A sound diagnosis cannot select one mechanism solely from that output; it needs an orthogonal measurement. This is the central protection against overinterpreting a polymer melt curve, a DSC peak, a strength measurement or a recycling-yield statistic.

Compare Defects, measurement systems and statistical control with Residence-time distributions and mixing as a neighboring subject in this unit. Do the two relationships describe the same physical quantity or two independent descriptors? Would their characteristic times, spatial scales, material states or experimental procedures be comparable? Propose one experiment that measures both and write what a disagreement would imply. A second model is useful only if it provides independent information rather than repeating the same assumption in a new notation.

5. Evidence and experimental design

Use traceability by lot, setting and timestamp, repeatability/reproducibility of visual grades, and controlled one-factor challenges.

Measurement planning: Define specimens, sampling locations, conditioning, calibration method, measurement resolution and number of independent replicates. Specify one reference material or control and the expected direction of a genuine effect. Organize a data ledger with columns for parameter, unit, instrument, method, timestamp, operator, uncertainty and calibration status. Do not fit and validate a model on the same observations without acknowledging that the validation is not independent. Where the property depends on processing history, preserve the exact prior temperature, deformation and storage conditions.

A robust plan explicitly separates structural uncertainty, parameter uncertainty and observation uncertainty. Structural uncertainty asks whether important physics are absent; parameter uncertainty asks whether fitted numbers are poorly identified; observation uncertainty asks whether instrumentation or sampling is noisy or biased. Repeated sampling can reduce some random errors but cannot repair an incorrect equation or uncalibrated specimen history. For Defects, measurement systems and statistical control, name one of each, and predict which category will dominate extrapolation into an unmeasured processing regime.

6. Interactive investigation and a falsifiable prediction

Open [Laboratory 18](../labs/lab-18.html). Read its exact equation and domain, then predict the trend with the first slider before viewing the curve. Change only that slider at three positions, record numeric outputs and use CSV export for the complete sweep. Change a second parameter next; compare whether the response retains its direction and curvature. Treat visible invalid-condition warnings as part of the lesson, not software errors to ignore. The simulation operates entirely in the browser with no external requests.

Investigation: Construct a fault tree for one defect and identify one observation that differentiates two possible causes.

Create a two-column evidence note: one column lists predictions directly supported by the stated mathematical relationship; the second lists interpretations that require independent polymer measurements. State what observation would most clearly falsify the assumed relation. If no practical experiment is yet proposed, the numerical model is a consistency exercise rather than a validated predictor. Do not infer equipment tolerances, worker-safety margins or compliance from fictional sliders.

7. Design decision with limitations

A process engineer evaluating Defects, measurement systems and statistical control faces at least two objectives: a desired quantity or geometry and the assurance that material quality remains fit for intended service. Build a trade-off table containing accepted-product quality, throughput or test cost, risk of extrapolation and measurement effort. For each alternative, name the assumption on which the apparent advantage depends. Identify a stop condition, such as failure to reproduce the response with a new batch or detection of a qualitatively different material phase or deformation regime. This makes the decision conditional and auditable.

Read the technical reference titles on the [references page](../REFERENCES.html) and check whether they discuss terminology, measurement method or product specifications. A terminology reference cannot supply experimental constants; a standard's title and abstract do not grant permission to claim full procedural compliance. The book is an editorial draft and a verified industrial decision requires current complete standards, manufacturer information and professional review.

8. Independent lesson problems

1. Derive or reconstruct Observed reject fraction p̂=number rejected/number inspected from the most appropriate physical or statistical principle. Define every symbol and every unit, then identify a limiting case.
2. Reproduce this lesson's worked example and change one input by ±10%; report output changes with justified significant figures.
3. Identify two competing physical explanations for a deviation from the model, and one measurement that distinguishes them.
4. Design a sampling plan that tests applicability after a change in material grade, temperature or processing history. Show how you would preserve calibration traceability.
5. Write a six-sentence engineering review: objective, model result, credible restriction, alternative, required experiment and conditional conclusion.

Next: return to the [unit pathway](../units/unit-06.html) or complete the [unit assessment](../assessments/unit-06.html).

Unit 06 applied engineering design study — Process measurement and control

Case status: fictional scenario. All measurements, market costs, process coefficients and material specifications are synthetic unless properly sourced by the learner. No calculated output authorizes a machine setting, product certification or environmental claim.

Case brief

An extruder occasionally makes black specks and off-color startup material. The indicated barrel temperature looks stable, and the operator assumes the problem must be a bad pigment lot.

The essential discipline is to prevent a model from silently becoming evidence. A design team should draw a boundary, define a function, preserve material and energy units, and identify a measurement capable of changing its conclusion. In polymer systems the same nominal resin can behave differently after drying, heating, oxidation, stretching and reprocessing. Product acceptance must be assessed on the relevant final article and sampling plan.

Technical analyses

Compare residence-time mean and distribution width, degradation time response and delayed temperature indication. Build a cause-and-effect hypothesis map and a batch/time-stamped data collection plan.

Describe where the physical state changes, where material is lost, where energy crosses the boundary and where quality is gained or degraded. If two alternatives differ by material type or final function, normalize them before comparing. State which inputs are experimentally measured, independently sourced, nominal manufacturer data or assumed illustrative coefficients. Use an uncertainty interval for any decision-critical property and do not invent a measured value.

Three interactive investigations

1. [Simulation 16](../labs/lab-16.html): sweep the first parameter and reconstruct the default numerical output from its equation; record an analytical limit.
2. [Simulation 17](../labs/lab-17.html): assess which second physical mechanism the model adds; state whether both calculations use the same material state.
3. [Simulation 18](../labs/lab-18.html): deliberately test parameter sensitivity or an invalid-domain warning and explain why treating a fictitious curve as a specification would be unsafe.

Calculate at least one output independently in a spreadsheet or short code segment, save the CSV curves, and annotate differences between numerical rounding and structural model assumptions. The three simulations may complement one another without forming a complete virtual factory.

Evidence protocol

Design a measurement schedule with sensor or analytical method, instrument range, calibration date, specimen preparation, sampling interval and acceptable data completeness. Define a control sample and a replicate strategy. For process comparisons, separate startup and steady production and keep separate numerator/denominator definitions for material yield, rejected-part rate and energy use. A small systematic bias can outweigh an impressive-looking simulated benefit.

Propose at least two counterhypotheses that could explain the same visible process defect or laboratory trend. Specify the smallest experiment that can discriminate them. State which claim you would withdraw if an independent material or product test contradicts the predicted model curve. When the claim concerns safety, food contact, biodegradation or a regulated end use, require the relevant complete, current standard and qualified expert review.

Deliverables

Submit RTD model comparisons, temperature-step sensitivity, a fault tree, an instrumentation audit and a sampling protocol designed to reject at least one competing hypothesis.

In addition, submit a one-page management summary: intended function and specification, explicit assumptions, normalized numerical comparison, evidence gaps, the chosen next experiment and the decision that cannot yet be made. Append all formulas, spreadsheet inputs and raw exports so another person can reconstruct the case. A recommendation is not complete until it specifies what new evidence would change it.

Assessment rubric

• 25% correctness of mass, dimensional and mathematical reasoning.
• 20% physically appropriate model scope and identification of omissions.
• 20% reproducible model sweeps and independent numerical checks.
• 20% design of experiments and measurement-system integrity.
• 15% appropriate conditional decision and transparent communication.

Unit 06 assessment — Process measurement and control

Format: 16 extended, multi-part problems, four per lesson. The responses are assessed for correct definitions (20%), conservation or mathematical reasoning (25%), reconstructable numerical work (25%), limitation-aware interpretation (20%) and evidence-based decision (10%). Every example is explicitly synthetic, and a student can correctly reject an unsupported model claim.

Lesson 1 — Residence-time distributions and mixing

Question 01: Numerical reconstruction and sensitivity

Reproduce the worked computation for Residence-time distributions and mixing: If mean residence time τ=120 s and n=4, the ideal residence-time standard deviation is 60 s. Increasing n to 9 narrows it to 40 s at the same τ, but does not eliminate a true stagnant zone. Produce a three-column ledger (given, calculated and unit), check the result using a different numerical route, then perturb one input by +15% and −15%. Report whether the effects are symmetric in output space. Explain why arithmetic agreement cannot establish the full measured distribution, property or grade suitability. Include one sign or limit check based on For n ideal equal CSTRs in series, variance σt²=τ²/n.

Question 02: Physical derivation and dimensional audit

Develop the equation used in Residence-time distributions and mixing from an appropriate physical basis: For n ideal equal CSTRs in series, variance σt²=τ²/n. Enumerate the assumptions needed to omit alternative effects, show the dimensional units of every input, and identify at least two plausible conditions under which the expression becomes unsuitable. Compare the idea with a neighboring topic in the same unit; which variable would be shared, and which material or instrument information cannot be transferred? Explain which conservation law or identity remains valid even if the fitted relation fails.

Question 03: Simulation challenge and falsification

Run [Laboratory 16](../labs/lab-16.html) and export its entire 41-point parameter sweep. Select one low, one intermediate and one high value; calculate a finite-difference slope, verify one output independently and test a boundary or failure condition. Use this known limitation: The tanks-in-series representation cannot automatically capture stagnant pockets, pressure dynamics, phase change or nonlinear residence-dependent viscosity. Construct a plausible synthetic dataset that violates the model trend and state whether you would investigate wrong coefficients, missing mechanism or instrument bias first. Design a discriminating test to distinguish the possibilities.

Question 04: Engineering decision and measurement plan

An engineering review proposes changing a material or processing condition because of the following evidence recommendation: Use an appropriate tracer, include line dead-volume correction and compare measured E(t) with model family rather than assuming a perfect plug-flow response. Write two alternatives, a quality target, a measurement cost estimate expressed as assumptions, and a stop condition. Tie the decision to the explicit lesson investigation: Contrast n=1 and n=8 response widths while documenting evidence required for a real extruder. Describe at least three separate observations needed before adoption. State a condition under which your preferred option would reverse and a limitation that makes extrapolation unacceptable without new data.

Lesson 2 — Thermal history and degradation kinetics

Question 05: Numerical reconstruction and sensitivity

Reproduce the worked computation for Thermal history and degradation kinetics: For a rate coefficient k=0.005 s⁻¹ at the specified temperature and hold t=100 s, retained fraction exp(−0.5)≈0.607. This is only a kinetic proxy and is not a safety threshold. Produce a three-column ledger (given, calculated and unit), check the result using a different numerical route, then perturb one input by +15% and −15%. Report whether the effects are symmetric in output space. Explain why arithmetic agreement cannot establish the full measured distribution, property or grade suitability. Include one sign or limit check based on First-order retained fraction f=exp(−k0 t exp[−Ea/(RT)]).

Question 06: Physical derivation and dimensional audit

Develop the equation used in Thermal history and degradation kinetics from an appropriate physical basis: First-order retained fraction f=exp(−k0 t exp[−Ea/(RT)]). Enumerate the assumptions needed to omit alternative effects, show the dimensional units of every input, and identify at least two plausible conditions under which the expression becomes unsuitable. Compare the idea with a neighboring topic in the same unit; which variable would be shared, and which material or instrument information cannot be transferred? Explain which conservation law or identity remains valid even if the fitted relation fails.

Question 07: Simulation challenge and falsification

Run [Laboratory 17](../labs/lab-17.html) and export its entire 41-point parameter sweep. Select one low, one intermediate and one high value; calculate a finite-difference slope, verify one output independently and test a boundary or failure condition. Use this known limitation: One Arrhenius channel does not account for complex stabilizer chemistry, autocatalysis, chain branching, crosslinking or spatial gradients. Construct a plausible synthetic dataset that violates the model trend and state whether you would investigate wrong coefficients, missing mechanism or instrument bias first. Design a discriminating test to distinguish the possibilities.

Question 08: Engineering decision and measurement plan

An engineering review proposes changing a material or processing condition because of the following evidence recommendation: Measure molecular weight, carbonyl bands or mechanical retention after controlled temperature-time histories; include oxygen and moisture conditions. Write two alternatives, a quality target, a measurement cost estimate expressed as assumptions, and a stop condition. Tie the decision to the explicit lesson investigation: Compare exposures that have equal mean residence but different long-tail fractions. Describe at least three separate observations needed before adoption. State a condition under which your preferred option would reverse and a limitation that makes extrapolation unacceptable without new data.

Lesson 3 — Temperature-loop dynamics and disturbance rejection

Question 09: Numerical reconstruction and sensitivity

Reproduce the worked computation for Temperature-loop dynamics and disturbance rejection: For T0=180°C, Tset=200°C and τ=80 s, after 160 s the model predicts T≈197.3°C. The equation assumes one constant setpoint, not a complete PID controller. Produce a three-column ledger (given, calculated and unit), check the result using a different numerical route, then perturb one input by +15% and −15%. Report whether the effects are symmetric in output space. Explain why arithmetic agreement cannot establish the full measured distribution, property or grade suitability. Include one sign or limit check based on First-order surrogate T(t)=Tset+(T0−Tset)exp(−t/τ).

Question 10: Physical derivation and dimensional audit

Develop the equation used in Temperature-loop dynamics and disturbance rejection from an appropriate physical basis: First-order surrogate T(t)=Tset+(T0−Tset)exp(−t/τ). Enumerate the assumptions needed to omit alternative effects, show the dimensional units of every input, and identify at least two plausible conditions under which the expression becomes unsuitable. Compare the idea with a neighboring topic in the same unit; which variable would be shared, and which material or instrument information cannot be transferred? Explain which conservation law or identity remains valid even if the fitted relation fails.

Question 11: Simulation challenge and falsification

Run [Laboratory 18](../labs/lab-18.html) and export its entire 41-point parameter sweep. Select one low, one intermediate and one high value; calculate a finite-difference slope, verify one output independently and test a boundary or failure condition. Use this known limitation: No dead-time, saturation, actuator limits, coupling or controller dynamics are represented, so model agreement does not justify final control tuning. Construct a plausible synthetic dataset that violates the model trend and state whether you would investigate wrong coefficients, missing mechanism or instrument bias first. Design a discriminating test to distinguish the possibilities.

Question 12: Engineering decision and measurement plan

An engineering review proposes changing a material or processing condition because of the following evidence recommendation: Collect heat-up/cool-down step tests and feed-rate disturbances; compare thermocouple calibration and melt-temperature measurements. Write two alternatives, a quality target, a measurement cost estimate expressed as assumptions, and a stop condition. Tie the decision to the explicit lesson investigation: Vary thermal time constant and find when sensor placement could make a simple step test misleading. Describe at least three separate observations needed before adoption. State a condition under which your preferred option would reverse and a limitation that makes extrapolation unacceptable without new data.

Lesson 4 — Defects, measurement systems and statistical control

Question 13: Numerical reconstruction and sensitivity

Reproduce the worked computation for Defects, measurement systems and statistical control: If 18 parts fail inspection out of 600 inspected, p̂=0.03. A binomial standard error sqrt[p̂(1−p̂)/600] is approximately 0.0070, if independent sampling assumptions hold. Produce a three-column ledger (given, calculated and unit), check the result using a different numerical route, then perturb one input by +15% and −15%. Report whether the effects are symmetric in output space. Explain why arithmetic agreement cannot establish the full measured distribution, property or grade suitability. Include one sign or limit check based on Observed reject fraction p̂=number rejected/number inspected.

Question 14: Physical derivation and dimensional audit

Develop the equation used in Defects, measurement systems and statistical control from an appropriate physical basis: Observed reject fraction p̂=number rejected/number inspected. Enumerate the assumptions needed to omit alternative effects, show the dimensional units of every input, and identify at least two plausible conditions under which the expression becomes unsuitable. Compare the idea with a neighboring topic in the same unit; which variable would be shared, and which material or instrument information cannot be transferred? Explain which conservation law or identity remains valid even if the fitted relation fails.

Question 15: Simulation challenge and falsification

Run [Laboratory 18](../labs/lab-18.html) and export its entire 41-point parameter sweep. Select one low, one intermediate and one high value; calculate a finite-difference slope, verify one output independently and test a boundary or failure condition. Use this known limitation: Binomial intervals can be misleading under clustering, inconsistent inspection or changing defect criteria. Construct a plausible synthetic dataset that violates the model trend and state whether you would investigate wrong coefficients, missing mechanism or instrument bias first. Design a discriminating test to distinguish the possibilities.

Question 16: Engineering decision and measurement plan

An engineering review proposes changing a material or processing condition because of the following evidence recommendation: Use traceability by lot, setting and timestamp, repeatability/reproducibility of visual grades, and controlled one-factor challenges. Write two alternatives, a quality target, a measurement cost estimate expressed as assumptions, and a stop condition. Tie the decision to the explicit lesson investigation: Construct a fault tree for one defect and identify one observation that differentiates two possible causes. Describe at least three separate observations needed before adoption. State a condition under which your preferred option would reverse and a limitation that makes extrapolation unacceptable without new data.