Author: [Bruce Zhou]
Affiliation: [Jota Machinery Composites Material Prepreg Solution]
Corresponding Author: [jotamachinery@gmail.com]
Published : December 05 , 2025

Abstract

Robotic automated fiber placement (R-AFP) of carbon-fiber towpregs offers a route to high-throughput, repeatable manufacture of aerospace-grade laminates, but process parameters must be tuned to balance consolidation quality, fiber integrity, and interlaminar performance. This study investigates an R-AFP process for carbon/epoxy towpregs, combining in-situ infrared thermography with tensile testing of cross-ply laminates. Towpregs are produced from 12K PAN-based carbon fibers impregnated with an epoxy matrix using a dedicated impregnation line, with resin weight fraction (RWF) varied between 25.1 and 35.9 wt%. Flat [0°/90°]₁₀ laminates are deposited by a 6-axis industrial robot under compaction pressures of 0.8, 1.6, and 2.4 bar and cured in an autoclave. Infrared imaging during lay-up reliably detects intentional gaps, overlaps, twisted tows, and local detachment through simple thermal contrast, confirming thermography as a practical online quality-monitoring tool. Quasi-static tensile tests reveal linear elastic responses up to fracture for all conditions. Lower resin content systematically increases tensile modulus and ultimate tensile strength (UTS), but shifts failure towards extensive delamination due to limited matrix volume. Increasing compaction pressure reduces both UTS and modulus, with microscopy showing transverse fiber damage at higher pressures that outweighs any reduction in porosity. The results define a qualitative process window in which modest compaction pressure and intermediate resin content provide the best compromise between stiffness, strength, and interlaminar integrity. The combined use of robotic placement, thermography, and microstructural analysis provides a framework for tuning AFP parameters in structurally critical composite laminates.

Keywords: automated fiber placement; towpreg; carbon-fiber reinforced polymer; infrared thermography; compaction pressure; resin weight fraction; tensile failure; delamination.

1. Introduction

Automated fiber placement (AFP) has become a key technology for manufacturing high-performance composite structures where weight, repeatability, and dimensional accuracy are critical. By steering narrow, pre-impregnated tows or tapes along numerically defined paths, AFP can build tailored laminates with minimal manual intervention and consistent fiber orientation. When AFP is combined with industrial robots, robotic automated fiber placement (R-AFP) offers additional flexibility in workspace and component geometry.

Despite these advantages, AFP processes are sensitive to defects generated during deposition. Gaps, overlaps, twisted tows, and local detachment from the substrate are common placement imperfections that can degrade interlaminar strength, initiate delamination, and reduce fatigue life. At the same time, process variables such as compaction pressure, lay-up speed, temperature, and resin content influence the balance between void removal, fiber alignment, and matrix flow. Process windows are often defined empirically, and conservative settings are used to avoid scrap at the expense of productivity or performance.

Non-destructive evaluation (NDE) techniques capable of monitoring the process in situ are therefore highly desirable. Infrared thermography is attractive because it is non-contact, can cover the entire placement region, and is sensitive to small variations in local thickness, contact quality, and heating history. However, its integration into R-AFP and its quantitative link to mechanical performance remain under-explored.

This work focuses on an R-AFP process for thermoset carbon-fiber towpregs and addresses two questions:

1. Can online infrared thermography reliably flag typical AFP defects during deposition of towpregs?
2. How do resin weight fraction (RWF) and compaction pressure influence tensile properties and failure modes of cross-ply laminates manufactured by R-AFP?

By combining controlled towpreg impregnation, robotic lay-up, thermographic monitoring, and post-cure mechanical testing, the study proposes a process–structure–property link that is directly relevant to industrial AFP applications.

2. Materials and Methods

2.1 Towpreg Material and Impregnation

The reinforcement is a 12K PAN-based high-strength carbon fiber supplied as a dry tow and slit to a nominal width of 5 mm. The matrix is an aerospace-grade epoxy system suitable for autoclave curing at 130 °C.

Towpregs are produced on a dedicated automated fiber impregnation line. The epoxy resin is pre-heated at 60 °C for 8 h to reach a stable viscosity. The carbon tow is fed from an unwinding creel under a nominal tension of about 15 N and passes beneath a spreader roll where the fiber bundle is flattened. Resin is metered onto the spread tow through a controlled gap, ensuring a uniform film along the width. The impregnated tow is then pulled through a short heated section to promote partial resin flow into the fiber bundle before being cooled and wound onto spools together with a polypropylene release paper to prevent blocking.

The resin weight fraction (RWF) is controlled through resin metering settings and line speed, which is kept at 20 m/min for all runs. RWF is measured according to ASTM D3171 by digesting the resin in a chemical bath and weighing the remaining fiber mass. Two target compositions are used in this study, corresponding to RWF ≈ 25.1 wt% and ≈ 35.9 wt%. Optical microscopy of polished cross-sections confirms that the towpregs are well impregnated, with fibers fully embedded in the matrix and no obvious intra-tow voids at both compositions.

2.2 Robotic Automated Fiber Placement Cell

Lay-up is carried out using an industrial 6-axis robot (KUKA KR 360), equipped with a custom fiber placement head. The head integrates:

• a set of guide rollers to route the towpreg;
• a compaction roller mounted on a load-controlled mechanism to apply a known normal pressure;
• a local heating system to raise the surface temperature of the towpreg and promote tack during placement;
• cutting and clamping units for tow start/stop operations.

The robot operates in position control, with path planning ensuring constant deposition speed along straight courses. The towpreg is placed onto a flat, heated substrate tool.

Three nominal compaction pressures p are investigated: 0.8, 1.6, and 2.4 bar, measured as the average normal pressure under the compaction roller. The pressure is adjusted by regulating the force applied by the roller mechanism and is monitored with a load cell during process set-up. The deposition speed is fixed at 100 mm/s for all experiments.

2.3 Laminate Architecture and Curing

Flat cross-ply laminates are produced with a stacking sequence [0°/90°]₁₀ (20 plies in total). Towpregs are placed in adjacent courses with nominal butt-joint contact; the 0° plies are oriented along the future loading direction, and 90° plies are orthogonal. Each panel has a nominal planform of 300 × 300 mm.

Following lay-up, panels are vacuum bagged and cured in an autoclave. The cure cycle consists of a ramp to 130 °C under 4 bar overpressure, a 2 h dwell, and controlled cooling to ambient temperature. After cure, laminate thickness is measured at multiple locations using a digital micrometer. As expected, lower RWF and higher compaction pressure produce slightly thinner laminates.

2.4 Infrared Thermographic Monitoring

An infrared camera (InfraTec VarioCAM® HD) is mounted on a fixed support such that its field of view covers the deposition zone. The camera records thermal images at up to 240 Hz during placement, capturing the heat signature of the freshly placed towpreg and underlying layers.

To assess detection capability, four typical AFP defects are intentionally introduced by modifying the robot path or process conditions:

• Gap – a programmed lateral offset between adjoining courses.
• Overlap – a deliberate lateral intrusion of one course over its neighbour.
• Twisted tow – a controlled twist of the towpreg between guide rollers to disturb its flatness.
• Detachment – insufficient local contact pressure near an edge, producing poor tack.

For each scenario, thermographic data are recorded and compared with optical cross-sections of the cured laminates, allowing correlation between the observed thermal patterns and actual geometric defects.

2.5 Tensile Testing

Quasi-static tensile properties are measured according to ASTM D3039. Rectangular specimens are waterjet-cut from cured panels with the long axis aligned either with the 0° or with the 90° direction of the laminate. Aluminium end tabs are bonded to reduce grip-induced damage.

Tests are performed on an MTS 810 servo-hydraulic machine under displacement control at a crosshead speed of 2 mm/min. Axial strain is measured either by an extensometer or via machine crosshead displacement over a calibrated gauge length. For each combination of RWF, compaction pressure, and loading direction, at least five specimens are tested to failure. Ultimate tensile strength (UTS) and Young’s modulus E are derived from the stress–strain curves.

2.6 Microstructural Analysis

Polished cross-sections of selected laminates are examined by optical microscopy to identify gaps, overlaps, and delamination patterns. Fracture surfaces from tensile specimens are observed using scanning electron microscopy (SEM) to assess fiber fracture, matrix cracking, and interfacial debonding. Particular attention is paid to transverse fiber damage in specimens produced at higher compaction pressures.

3. Results

3.1 Thermographic Detection of Deposition Defects

Infrared imaging reveals distinct thermal signatures for each intentional defect type:

• Gaps between adjacent courses appear as relatively cool, narrow bands in the thermal map. Because the uncoupled substrate under the gap region has lower heat capacity than a well-consolidated tow, the temperature drops more rapidly, producing a visible contrast. Cross-sectional microscopy confirms that these regions are indeed unreinforced strips flanked by slightly distorted neighbouring plies.
• Overlaps generate local hot bands where the thickness is greater than nominal. The additional material retains heat longer and presents a higher apparent surface temperature. Optical sections show two superimposed tows with the same orientation and small secondary gaps in the adjacent courses caused by the lateral displacement.
• Twisted tows are characterised by elongated hot zones along the feed direction, reflecting locally increased thickness and reduced contact area. The twisted region disrupts both the tow itself and the conformability of subsequent plies, causing local deviations in fiber alignment and resin pooling.
• Detached regions resulting from poor tack exhibit sharp, localised hot spots. Because the tow is not in intimate contact with the underlying layer, heat transfer to the substrate is reduced, and the surface cools more slowly. After cure, these regions correspond to partially disbonded plies or regions with entrapped air.

Defect-free lay-up produces a homogeneous temperature field, with gradual, uniform cooling and no local maxima or minima beyond small variations associated with tow edges. These observations indicate that qualitative thermography is sufficient to detect and classify typical R-AFP defects as they occur.

3.2 Tensile Stress–Strain Response

For all tested conditions, the stress–strain curves are linear up to catastrophic failure, consistent with brittle fracture of carbon/epoxy laminates under tension. No macroscopic yielding or plateau behaviour is observed. Within experimental scatter, specimens loaded along 0° and 90° directions exhibit similar global responses, reflecting the symmetric [0°/90°] stacking and the relatively balanced contribution of the two orientations in these cross-ply laminates.

Both process variables—RWF and compaction pressure—have a clear influence on stiffness and strength. Lower resin content (25.1 wt%) leads to higher initial slope of the stress–strain curve and higher UTS compared with high resin content (35.9 wt%). In contrast, increasing compaction pressure from 0.8 to 2.4 bar results in a progressive reduction in both modulus and strength for a given RWF.

3.3 Effect of Resin Weight Fraction

Reducing RWF from 35.9 to 25.1 wt% increases the fiber volume fraction and decreases overall laminate thickness. At fixed compaction pressure, lower RWF consistently yields higher tensile modulus and UTS. This trend is physically intuitive: more fibers per unit cross-section carry the tensile load more effectively.

However, fracture observations reveal that this gain in axial performance comes at a cost in interlaminar integrity. Laminates with low RWF show pronounced layer-wise debonding after first ply failure, especially in specimens loaded transverse to the local fiber direction. Once transverse matrix cracks form in the 90° plies, delamination readily propagates along interfaces because the reduced resin content offers limited resistance to crack growth. Fracture surfaces are characterised by step-like features and broad regions of clean fiber surfaces indicative of fiber/matrix separation.

In high RWF laminates, transverse matrix cracking is still present, but delamination is less extensive. The additional resin provides more bridge ligaments and higher interfacial area, which delays interlaminar separation. Final failure often occurs when 0° plies fracture after progressive damage in the off-axis layers. Overall stiffness and strength remain lower than in low-RWF laminates due to the higher proportion of non-load-bearing matrix.

3.4 Effect of Compaction Pressure

Across all RWF levels, increasing compaction pressure from 0.8 to 2.4 bar produces only modest further reductions in cured laminate thickness, suggesting that porosity levels are already low at the lowest pressure. Nevertheless, mechanical properties deteriorate significantly as pressure increases.

SEM examination of fracture surfaces from high-pressure specimens reveals evidence of transverse fiber damage attributed to the compaction roller. Localized crushing or micro-buckling of filaments is observed near the surface, particularly where tows change direction or pass over slight geometric irregularities. These damaged regions act as stress concentrators during tensile loading and promote premature fiber breakage. As a result, the benefit of slightly increased consolidation at high pressure is overshadowed by the detrimental effect of fiber damage.

At the lowest pressure (0.8 bar), fracture surfaces display more intact fibers and fewer indications of compaction-induced damage. Cracking is dominated by matrix fracture and fiber/matrix debonding, with fiber fracture occurring closer to the nominal ultimate load of the material.

4. Discussion

4.1 Process–Structure–Property Relationships

The combined results highlight a nuanced process–structure–property relationship for R-AFP of carbon-fiber towpregs:

1. Resin content controls the balance between axial performance and delamination resistance.
   Lower RWF increases fiber volume fraction, thereby increasing stiffness and strength. However, the reduced matrix volume weakens interlaminar bonding and favours delamination once matrix cracks appear. Higher RWF provides more robust interlaminar bonding but dilutes the reinforcing phase and increases the propensity for matrix-dominated cracking.
2. Compaction pressure has an optimum window.
   A common intuition is that higher compaction pressure improves laminate quality by reducing voids and promoting tow consolidation. In this study, once a baseline level of consolidation is achieved, further increasing pressure primarily introduces transverse fiber damage and residual stresses, leading to lower tensile performance. This behaviour implies that compaction pressure should be optimised rather than maximised, and that roller design and contact kinematics play critical roles.
3. Defects detectable by thermography are strongly linked to mechanical performance.
   Although mechanical testing in this work was performed on panels manufactured without intentionally left defects, the thermal images demonstrate that gaps, overlaps, twisted tows, and detachments produce clear signatures. In practice, such defects would introduce local stress concentrations and reduce net section strength, especially under fatigue or impact. Integrating thermography into an automated feedback loop could prevent such defects from being embedded into critical components.

4.2 Implications for Industrial R-AFP

From an industrial perspective, the findings suggest several guidelines:

• Online thermography is practical and informative.
  The ability to visually classify defects in real time using temperature contrast alone makes infrared systems attractive for production cells. Even without advanced image processing, operators or supervisory software can flag regions requiring repair or rework.
• Intermediate resin content is a pragmatic compromise.
  While the lowest tested RWF delivers the highest tensile properties, interlaminar behaviour must be considered for real components subject to multi-axial loading, impact, or fatigue. An intermediate resin content—between the two extremes studied here—may offer a better overall compromise between axial stiffness, strength, and delamination resistance, especially when combined with toughened resins or z-reinforcements.
• Compaction settings must be tuned to the material system.
  The pressure levels that are safe for one fiber/matrix combination, tow width, or roller geometry may be excessive for another. Process qualification should therefore include systematic variation of compaction pressure with parallel mechanical and microstructural assessment, rather than adopting generic “high-pressure” settings.
• Extension to complex geometries requires further study.
  The present work focuses on flat cross-ply laminates. Curved geometries, thickness variations, and steering paths typical of pressure vessels and aerodynamic shells will influence the local pressure distribution, temperature field, and defect formation. Process windows identified here provide a starting point, but further work is needed to map these trends to realistic component shapes.

4.3 Future Work

Several directions emerge from this study:

• Quantitative defect evaluation via thermography.
  The present thermographic analysis is qualitative. Future work should aim to establish detection thresholds, size estimates, and probability of detection for different defect types and depths, possibly combining active thermography with controlled heating pulses.
• Broader parameter space.
  Additional process variables—including placement speed, preheating temperature, roller hardness, and tow width—should be incorporated into a more comprehensive design of experiments to define robust process windows for different laminate architectures.
• Additional loading scenarios.
  Complementary tests in compression, open-hole tension, interlaminar shear, fatigue, and impact would provide a more complete picture of how RWF and compaction pressure affect structural performance in representative service conditions.

5. Conclusions

A robotic automated fiber placement process for carbon-fiber towpregs has been studied with respect to online defect monitoring and tensile performance of cross-ply laminates. The main conclusions are:

1. Infrared thermography is effective for in-situ quality monitoring.
   Gaps, overlaps, twisted tows, and local detachment each produce distinctive thermal signatures during placement, making them readily identifiable in real time.
2. Resin weight fraction strongly influences stiffness, strength, and failure mode.
   Lower resin content increases tensile modulus and UTS by raising fiber volume fraction but promotes extensive delamination due to limited matrix volume. Higher resin content improves interlaminar cohesion but reduces axial properties and shifts damage towards matrix cracking.
3. Compaction pressure must be carefully controlled.
   Within the studied range, increasing compaction pressure reduces tensile properties, primarily due to transverse fiber damage introduced during lay-up. Optimal settings minimise voids without crushing fibers.
4. A balanced process window is required for structural applications.
   For cross-ply laminates, a combination of moderate resin content and low to moderate compaction pressure appears to offer the best compromise between axial performance and interlaminar integrity.

Overall, the integration of R-AFP, thermographic monitoring, and mechanical characterisation provides a practical framework for tailoring process parameters in the manufacture of high-quality CFRP laminates and forms a foundation for future work on more complex geometries and loading conditions.

References

1. Forcellese, A.; Mancia, T.; Russo, A.C.; Simoncini, M.; Vita, A. Robotic automated fiber placement of carbon fiber towpregs. Materials and Manufacturing Processes 2022, 37(5), 539–547.
2. Lukaszewicz, D.H.-J.; Ward, C.; Potter, K.D. The engineering aspects of automated prepreg layup: History, present and future. Composites Part B: Engineering 2012, 43, 997–1009.
3. Stelzer, S.; Schaefer, T.; Hühne, C.; Rolfes, R. Automated fiber placement—Defect types and their influence on the mechanical properties of aerospace laminates. Composite Structures 2018, 200, 46–61.
4. Cherniaev, A.; Hallett, S.R. Performance of automated fibre placement laminates containing manufacturing defects. Composites Part A: Applied Science and Manufacturing 2019, 124, 105480.
5. Meola, C.; Boccardi, S.; Carlomagno, G.M. Infrared thermography in the evaluation of aerospace composite materials. Infrared Physics & Technology 2015, 71, 453–461.
6. Kim, J.-K.; Mai, Y.-W. Engineered Interfaces in Fiber Reinforced Composites; Elsevier: Oxford, 1998.
7. Daniel, I.M.; Ishai, O. Engineering Mechanics of Composite Materials; Oxford University Press: New York, 2006.