Processing, voids and property variation

Impregnation and cure establish structure

Manufacture must place reinforcement, fill spaces and develop the intended matrix state. Incomplete impregnation, entrapped gas, fibre waviness or uneven cure can change local load transfer. Processing conditions are therefore part of the material description.

Consider one route for a fibre-reinforced thermosetting polymer: place a dry fibre preform, introduce liquid resin so that it penetrates the spaces, then develop the cured matrix. A preform is the arranged reinforcement before it has become the finished composite. Its open space is intended to receive resin; it is not automatically the finished part’s void fraction. Other routes start with already impregnated material or a thermoplastic and have different histories.

Impregnation means penetration around and between fibres, rather than merely covering the outside. A tow is a bundle of fibres. Gaps between tows and the much smaller spaces inside a tow can fill at different rates. Pressure differences drive liquid flow against viscous resistance; wetting and capillary forces at the advancing liquid–gas boundary also matter. Permeability describes how readily a particular connected porous arrangement permits flow.

In one possible filling pattern, resin advances rapidly along an easier channel before fully entering a neighbouring bundle. The fronts can surround gas and leave an incompletely wetted region. Another balance of capillary, viscous and geometric effects gives another sequence. Resin at an outlet or a full-looking mould surface therefore does not prove that every internal path was filled. The first figure is a mechanism sketch, not a flow simulation.

Temperature can change liquid viscosity while cure changes the resin’s molecular connectivity and response. These effects interact: material that could flow earlier may become unable to complete the same redistribution later. Gelation marks the formation of a connected network in the reacting system; it does not mean every reaction is complete. More time, more pressure or more heat is not a universal repair for a trapped region or an unsuitable formulation. This lesson supplies no manufacturing settings.

Different defects interrupt different parts of the path

A finished void replaces material that otherwise could carry force or contact a fibre. Its surface changes the surrounding stress field, and an unwetted patch can remove a transfer path across part of a fibre–matrix contact. Shape, orientation, location, neighbouring fibres and the loading mode matter as well as total gas volume. A collection of round isolated pores is not mechanically equivalent to a thin connected unfilled region just because their volumes match.

Fibre waviness changes the local direction of reinforcement relative to the load. A resin-rich region contains less reinforcement on its local volume basis and redirects the local sharing of force. Either can affect response without changing the nominal fibre type. They can also occur together with voids, so attributing a changed response to void volume alone may hide a second change. Load transfer through fibres and interfaces develops the transfer mechanism; Damage, delamination and interface failure develops competing damage paths.

Cure variation is another state difference. A reacting thermoset changes connectivity, viscosity and stiffness, and the reaction can release heat. The temperature inside a thick region need not follow the surrounding tool or oven temperature exactly. Resin reaction shrinkage and different cooling contractions can produce internal stresses when other constituents or the tool restrict their free changes. A sound-looking surface and one resin trade name do not identify this history.

Keep the causal chain explicit: a process condition may alter flow or reaction; this may leave a particular structure or state; that structure may alter the relevant deformation or failure path. An observed correlation between a processing setting and a test result does not determine every intervening step. Compare local evidence and state as well as final response. Why processing history matters prepares this distinction.

A bulk density needs a matching constituent account

Use original fictional finished specimens, not measurements of a named laminate. At one common dry condition, suppose constituent masses are independently known: fibre mass m_f = 10.0 g and cured-matrix mass m_m = 6.96 g. Their actual densities in that condition are ρ_f = 2.50 g/cm³ and ρ_m = 1.20 g/cm³. The external body volume is V_b = 10.0 cm³. Assume their constituent volumes add, with only these two nongaseous constituents and negligible gas mass, no additional coating, absorbed liquid or reaction mass change.

From m = ρV, the fibre volume is V_f = 10.0/2.50 = 4.00 cm³ and matrix volume V_m = 6.96/1.20 = 5.80 cm³. Their total solid-constituent volume V_s = 9.80 cm³ leaves V_v = V_b − V_s = 0.200 cm³ not occupied by those constituents. The body-based void fraction is φ_v = V_v/V_b = 0.0200, or 2.00%. The complete account is 40.0% fibre, 58.0% matrix and 2.00% void by external body volume.

The total mass is 16.96 g, so bulk density is ρ_b = 16.96/10.0 = 1.696 g/cm³. The void-free density for this same constituent mass composition is instead ρ_s = 16.96/9.80 ≈ 1.73061 g/cm³. Dividing the two definitions gives ρ_b/ρ_s = V_s/V_b, hence φ_v = 1 − ρ_b/ρ_s. Density and properties relative to mass prepares the distinction between bulk and solid-constituent volume.

The reference must match the actual composition and state. A different void-free mixture with 40.0% fibre by its own total volume would have density 0.400 × 2.50 + 0.600 × 1.20 = 1.720 g/cm³. Substituting that value gives 1 − 1.696/1.720 ≈ 1.40%, not 2.00%. In the actual porous body, the fibre share of solid volume is 4.00/9.80 ≈ 40.82%, rather than 40.0%. Removing pores while keeping the same constituent amounts is not the same as silently replacing their space with matrix.

These equations are a material-volume account, not a void-strength law or an identification procedure. Unknown constituent proportions, using uncured rather than cured resin density, moisture, a third phase or an unsuitable external-volume measurement can invalidate the inference. Even a correct total gives no pore map, and the inferred unoccupied volume does not identify its cause or connectivity.

An average can hide a critical region

Bulk void fraction and average fibre fraction are useful but do not locate defects. A pore near a highly stressed interface may matter differently from an isolated pore elsewhere. Compare reproducible process records, spatial inspection and representative tests.

Divide each fictional 10.0 cm³ body into five equal 2.00 cm³ zones, with zone 3 representing the region of interest at a joint. All three records below have the same total constituent volumes and mass account just calculated. Treat these as exact volume inventories, not micrographs, specimens cut for a test or predictions of their performance.

In A, every zone is 40% fibre, 58% matrix and 2% void. In C, the four outer zones are 40% fibre, 60% matrix and no void, while the joint zone is 40% fibre, 50% matrix and 10% void. All the void volume is now in one fifth of the body. The overall void fraction remains (4 × 0 + 10)/5 = 2%. A and C keep the fibre fraction uniform while changing where the unoccupied volume lies.

B retains C’s void distribution but also changes reinforcement distribution. Its four outer zones are 45% fibre and 55% matrix; the joint zone is 20% fibre, 70% matrix and 10% void. The overall fibre fraction remains (4 × 45 + 20)/5 = 40%, and overall matrix fraction is (4 × 55 + 70)/5 = 58%. The original worked case’s resin-rich joint is now visible in the account: its matrix fraction is 70%, although the whole-body fraction is still 58%.

The second figure plots those inventories on a common volume basis. A versus B changes both local reinforcement and void placement, so a response difference would not isolate a void-location effect. C versus B holds the void map fixed and changes the fibre/matrix distribution. A versus C is the more informative pair for investigating the supplied change in void placement at fixed local fibre fraction. Actual pore shape, interface state and fibre orientation would still need to be controlled or measured.

No numerical joint strength follows from these fractions. The joint geometry and loading determine which region and mechanism matter, and two equal-volume zones need not be equally stressed. For unequal zone volumes, calculate a volume-weighted total: sum each zone’s void volume and divide by the sum of its body volumes. A plain average of percentages works here only because all five volumes are equal.

A repeated reading is not a new manufactured panel

Use a separate original response record, unrelated to the three spatial maps. Three independently manufactured panels from route P and three from route Q each provide one matched coupon. Dimensions, direction, conditioning, marked gauge and a 100 N tensile reading are common; all six have a stipulated measured overall void fraction of 2.00%. The supplied extensions are observations, not values predicted from void content.

P gives 0.098, 0.100 and 0.102 mm; Q gives 0.090, 0.100 and 0.110 mm. The arithmetic mean, sum divided by three, is 0.100 mm for each route. The observed range, largest minus smallest, is 0.004 mm for P and 0.020 mm for Q. Equal means therefore conceal a fivefold difference in this observed spread.

For an invented extension screen δ ≤ 0.105 mm at that same 100 N, all three P observations pass while only two Q observations pass. Judging only the mean would miss Q’s supplied 0.110 mm observation. These are fractions of six recorded outcomes, not probabilities that future panels will pass, and three panels do not identify a population distribution or establish a process capability.

Measuring one unchanged coupon three times mainly examines reading repeatability under those conditions; it does not create three independent manufacturing outcomes. Repeating regions within one panel examines local heterogeneity, while comparing separately made panels addresses between-panel variation. Measurement error, gauge/fixture movement, different conditioning or a systematic process change can also contribute. Record which unit was repeated before explaining a spread.

Because all six bulk void fractions match, those fractions alone cannot explain the extension spread. Inspect local void morphology and fibre distribution, verify cure and other state differences, and assess the measurement. A useful process comparison keeps these records connected rather than replacing them with a single average or assuming every observed defect caused the response.

Combine process, spatial and response evidence

A process record says what was imposed: material batch/state, preform placement, resin preparation, flow and reaction history, and the actual local temperatures or pressures that were monitored. It does not by itself prove the resulting interior. Spatial inspection asks what formed; a response test asks what that state does under the relevant load and conditions.

A polished section can reveal a local unwetted area or fibre waviness but samples a plane and can be altered by preparation. X-ray tomography can examine a three-dimensional field. Segmentation assigns image regions to inferred phases; resolution, contrast, thresholds and the chosen field affect what is counted. A few-pixel feature is not automatically a resolved pore, and an image of one region is not a complete population survey. Microscopy, diffraction and structure evidence prepares the distinction between an observation and its interpretation.

Cure or reaction-state evidence needs a suitable thermal, chemical or mechanical observation and a material-specific interpretation; bulk density is not a direct cure meter. Likewise, detecting low total void volume does not rule out fibre waviness, a resin-rich joint or a residual-stress problem. More complete information can change which proposed mechanism is plausible.

Choose a comparison that can discriminate explanations: matched local fibre arrangement with different measured void patterns; matched void pattern with different matrix states; or repeated locations and independent panels with the same inspection method. Do not change several factors and attribute the entire response difference to the one shown in a photograph. Keep unresolved alternatives explicit. The arithmetic here supplies no processing recipe, defect acceptance criterion, component lifetime or safety guarantee.

Worked examples

Try the reasoning

1. Does complete external fill prove complete internal impregnation?

   Show the answer and reasoning

   No.

   Internal voids can remain. Resin can pass along an easier channel while a neighbouring bundle remains incompletely wetted. An exterior observation and an internal phase map answer different questions.

2. Why retain cure history?

   Show the answer and reasoning

   It affects matrix structure and residual stress.

   A resin label does not specify its final state. Reaction, gelation and the local thermal history influence matrix response and shrinkage. A tool or oven reading does not necessarily describe every internal region.

3. A different body has a 1.00 cm³ joint region with 10.0% void and a 9.00 cm³ remainder with no void. What is whole-body void fraction? Is averaging 10% and 0% correct?

   Show the answer and reasoning

   Void volume is 0.100 cm³ and whole-body fraction 0.100/10.0 = 1.00%. The plain two-number average, 5%, is incorrect.

   The two regions do not have equal volumes. Add actual void volumes and actual body volumes separately, or use weights of 1/10 and 9/10. A local high fraction can coexist with a low bulk fraction.

4. An unfamiliar dry body has 8.00 g fibre of density 2.00 g/cm³, 5.40 g cured matrix of density 1.00 g/cm³ and external volume 10.0 cm³. On the same two-constituent assumptions, find void fraction, bulk density and composition-matched void-free density.

   Show the answer and reasoning

   V_f = 4.00 cm³, V_m = 5.40 cm³ and void volume = 0.600 cm³, giving 6.00%. Bulk density is 1.340 g/cm³; the matched void-free density is 13.40/9.40 ≈ 1.42553 g/cm³.

   The solid volume is 9.40 cm³ and mass is 13.40 g. Substituting the matched densities into 1 − ρ_b/ρ_s reproduces 0.0600. These are volume accounts, not a 6% loss of every mechanical property.

5. Only a bulk density of 1.696 g/cm³ is known. A report assigns 2.00% void by borrowing the preceding specimen’s composition. Could a no-void mixture of the same two constituents have that density with different proportions?

   Show the answer and reasoning

   Yes. A no-void fibre volume fraction of (1.696 − 1.20)/(2.50 − 1.20) ≈ 0.38154 gives the same density. Density alone has not identified 2.00% void.

   On a unit-volume no-void account, density is 2.50φ_f + 1.20(1 − φ_f). This alternative has about 38.15% fibre by volume. The two bodies can share a bulk density while differing in both composition and void content; independent composition/state evidence is needed.

6. A calculation uses ρ_b = 1.80 g/cm³ and a purported matching void-free density ρ_s = 1.73 g/cm³, producing a negative void fraction. Should the report silently replace it with zero and declare the body void-free?

   Show the answer and reasoning

   No. The negative result contradicts the assumed nonnegative missing-volume account and requires checking the inputs and model.

   Possible causes include wrong composition or constituent state, a significant additional phase or absorbed mass, unsuitable volume/density measurement, or uncertainty. The arithmetic cannot establish a physical negative pore volume or prove zero voids.

7. For P and Q at 100 N, compare observed mean and range, then change the extension screen to δ ≤ 0.100 mm. Which recorded panels pass, and may those fractions be used as future pass probabilities?

   Show the answer and reasoning

   Both means are 0.100 mm; ranges are 0.004 and 0.020 mm. With equality allowed, two of three recorded panels pass in each route. These fractions do not establish future probabilities.

   P’s 0.102 and Q’s 0.110 mm exceed the new screen. The mean did not change when the requirement changed, and the common 2/3 outcome does not erase their different observed ranges. Independent future sampling and an inference model would be needed for population claims.

8. A revised process gives fewer detected voids, a different measured matrix reaction state and a different joint response. Can the entire response change be assigned to reduced void content, or repaired by copying the process settings to another thickness?

   Show the answer and reasoning

   Neither follows. Void population and matrix state changed together, and another geometry can have a different flow and thermal history.

   Retain the actual local state/inspection and matched response records. Separate the competing changes with an appropriate comparison or validated coupled model. A source process cycle for one material and part is not a universal recipe.

What to carry forward

• Control and inspect local architecture.
• Connect process records with property variation.
• Impregnation, gelation and complete reaction are different observations and stages.
• Use a complete finished-body volume account and a composition-matched density reference.
• A global fraction does not specify local pore, fibre or matrix distribution.
• Attach variation to its sampling unit; a mean and observed range are not future reliability.
• Use process records, spatial inspection, state evidence and response tests together.