When an edge comes out bad, most people's first reaction is to "tweak the parameters." But lay out the analysis of variance from the literature and the answer is fairly clear: feed rate and tool geometry are the two main causes, and cutting speed is not even significant in most studies. What fewer people know is that delamination and burrs are often opposite trade-offs: at the same fiber angle, suppressing delamination tends to produce burrs.
The One-Minute Version
- Feed rate is the number-one factor. An ANOVA on 10.23mm-thick aerospace-grade CFRP shows exit delamination is dominated by feed rate (46.71%), followed by tool diameter (35.00%), with cutting speed not significant (6.60%, p=0.205).
- Tool wear has split effects. At flank wear VB=0.3mm, thrust force rises about 100%, yet entry delamination is actually the lowest while exit delamination gets significantly worse. It can't be reduced to "more wear is always worse."
- Backing support is the highest-return fix. With a backing plate pre-drilled with a matching hole array, exit delamination stayed at 1.00–1.02 over a full tool life of 1,403 holes.
- But a backing plate won't fix burrs. Burrs come from cutting-edge rounding and have nothing to do with support. These are two problems that must be handled separately.
- Delamination and burrs are opposite trade-offs. Fiber cutting angles of 0–90° give little delamination but more burrs; 90–180° give severe delamination but few burrs.
- Dull tools damage the material itself. After an uncoated carbide tool trimmed 7.2m, coupon flexural strength dropped 10.5% (p=0.029); with a CVD diamond-coated tool, strength did not drop over 16.2m.
Key 1: Find out which variables actually matter
Rather than adjusting parameters by feel, start with what designed experiments conclude. The two independent studies below each ran a full analysis of variance and calculated each variable's percentage contribution to delamination:
Delamination factor contributions: two studies side by side
ANOVA percentage contributions from two independent studies; bar length equals contribution
Not every study produces the same ranking. In experiments by Cerqueira et al. (2025) on HexPly 8552 prepreg, cutting speed was the most significant factor (p=0.040), while feed rate was borderline (p=0.051). A likely reason for the difference is that the study tested only two feed levels, 0.05 and 0.20 mm/rev, which covers a narrower range.
The practical reading: treat feed rate as the first knob to try, but not as the only answer. The truly reliable approach is still a small set of test cuts on your own material and tools.
Key 2: Delamination and burrs are opposite trade-offs
This is the most overlooked point in this article and the most useful in practice. The same fiber cutting angle θ affects delamination and burrs in opposite directions:
Fiber cutting angle: the opposing delamination vs. burr trade-off
Over the same angle range, delamination and burrs trend in opposite directions
You can't drive both delamination and burrs to zero at the same time; you can only decide which one you can less afford. For structural parts (arms, frame plates), prioritize suppressing delamination, because delamination eats directly into strength. For cosmetic parts or surfaces that will be bonded later, prioritize suppressing burrs, because burrs interfere with fit-up. Make this call before you lay out the toolpaths.
Key 3: Tool wear, and its effects split in two directions
A 2014 study by Feito et al. did something clever: instead of waiting for tools to dull naturally, they used grinding to create a controlled amount of wear (VB = 0.3 mm). That is the only way to isolate wear as a single variable. The results were surprising:
| Tool condition | Change in thrust | Entry delamination Fd | Exit delamination | Interpretation |
|---|---|---|---|---|
| New tool | 50–150 N (baseline) | — | Baseline | Best overall |
| Flank wear VB=0.3mm | +approx. 100% | 1.14–1.18 (lowest) | Significantly higher, highest of all trials | Entry better, exit worse |
| Edge rounding 0.05mm | +50–65% | approx. 1.3 (highest) | — | Worst entry |
| These data show that "a dull tool makes delamination worse" is an oversimplification. Wear changes the cutting-edge geometry and how loads are applied, and the effects on entry and exit can go in opposite directions. In practice, the side to watch is the exit side: that is what keeps getting worse as wear progresses. | ||||
A dull tool doesn't just cut ugly edges; it damages the material itself
This is the most important takeaway from a 2022 study by Ashworth et al. published in Wear. They ran flexural tests on trimmed CFRP coupons:
| Tool | New-tool edge radius | Cutting length | Change in coupon flexural strength | Statistical significance |
|---|---|---|---|---|
| Uncoated carbide | 6.08 ± 0.49 µm | 7.2 m | −10.5% | p = 0.029 (significant) |
| CVD diamond-coated | 16.64 ± 0.58 µm | 16.2 m | +2.2% | Not significant (no degradation) |
| The tools were φ6mm, 8-flute burr-style router end mills at 8,000 rpm and 0.015mm feed per tooth. Note that the coated tool's new edge radius is actually larger than the uncoated tool's (the coating thickens the edge), but it wears much more slowly, so it performs better over long cutting distances. | ||||
An arm cut with a dull uncoated tool already has about 10% less flexural strength than the material itself, and there is no way to see it from the outside. You won't find out until it breaks in flight. This is one of the most fundamental differences between machining CFRP and machining metal: machining quality doesn't just affect appearance; it directly changes the mechanical performance of the finished part.
The countermeasure is clear: change tools preventively, based on cutting length, not on how the edge looks.
Key 4: Backing support, the highest-return fix
The physical cause of exit-side delamination (push-down / push-out) is simple: the axial force of the tool pushing downward pries the bottom few unsupported plies off as a sheet. The fix is just as simple: give it support.
A long-duration experiment by Gutiérrez et al. in 2023 provides convincing data. They placed an extruded polystyrene foam board pre-drilled with a matching hole array under the workpiece and then drilled 1,403 holes in a row:
How backing support suppresses exit delamination, and where the returns diminish
Left: exit delamination level over a full tool life. Right: backing preload stress vs. delamination factor.
All six factors side by side, and the order to fix them
| Priority | Factor | Effect on delamination | Effect on burrs | Concrete action |
|---|---|---|---|---|
| 1 | Backing support | Exit delamination can be held to 1.00–1.02 | No effect | Spoilboard in full contact, vacuum hold-down, pre-drilled backing plate |
| 2 | Tool condition | Exit delamination keeps worsening with wear | Edge rounding is the main cause of burrs | Change tools preventively by cutting length, not by appearance |
| 3 | Feed rate | ~46% contribution, the largest variable | Secondary effect | Start at 600 mm/min and reduce 20% at a time |
| 4 | Tool geometry | 20–35% contribution | Flute form decides shearing vs. pushing fibers aside | Switch to multi-flute diamond, compression, or a smaller diameter |
| 5 | Fiber cutting angle | Worst at 135° | Opposite direction | Decide at layout which angle the critical edges fall on |
| 6 | Cutting speed | Not significant in most studies | Limited effect | Change it last; lowest payoff |
| Priority is ranked by "cost of the change ÷ effect," which is not exactly the same as ANOVA contribution. Backing support ranks first because it costs almost nothing yet eliminates an entire class of defects outright; cutting speed ranks last because the literature generally shows it is not significant. | ||||
One last thing: there is no standard for burrs
If you're looking for a standard that says how much burr is acceptable, the answer is that there isn't one. A 2021 review by Poór et al. in Composites Part B states directly that industry and academia rely mainly on digital image processing to evaluate burrs, but there is no standardized consensus on measurement techniques or metrics for CFRP burrs.
The same paper notes another reality: burrs usually do not reduce material strength, but industry standards still recommend removing them, and the deburring step often adds substantially to labor time and cost.
Since there is no external standard, set an internal one, and make it something that can be judged repeatably. A workable method: take a photo of an edge you consider acceptable as a reference sample and post it next to the machine; anything worse gets recut or deburred. It sounds low-tech, but it is far more reliable than "looks okay to me," and it is currently the only workable approach in this field.
Another improvement backed by data is tool tilting: Tima and Geier (2024) found that tilting the tool at an angle during edge trimming reduced the burr extent by about 50% compared with the conventional approach, and it was especially effective with tools near the end of their life.
FAQ
I don't have a vacuum table. Does a spoilboard with double-sided tape count as backing support?
Yes, and it works well. The key to backing support is that nothing under the workpiece is left unsupported, so the tool's axial thrust has somewhere to go. A spoilboard in full contact with the underside of the workpiece achieves that. The catch is that the tape itself must be thin: foam-backed tape is springy, which effectively leaves a compressible gap under the workpiece and reduces the support.
How do I measure the delamination factor Fd without specialized equipment?
Fd = maximum delamination diameter ÷ nominal hole diameter. That is the definition for drilling, and it can be measured with an ordinary toolmaker's microscope or a graduated magnifier. For milled edges there is no equivalent accepted metric; each study defines its own. In practice, the most workable approach is to photograph the edge, measure the maximum depth of the damaged zone with image software, and keep the photo conditions fixed so batches can be compared.
How long does it take for a more expensive diamond-coated tool to pay for itself?
It depends on your volume and quality requirements. Data from Ashworth et al. (2022) show that an uncoated tool reduced coupon flexural strength by 10.5% after 7.2m of cutting length, while a diamond-coated tool showed no degradation over 16.2m. If your parts are structural, that difference is a safety issue, not a cost issue. If you only make cosmetic parts or early prototypes, uncoated tools changed frequently are fine.
Why is my entry face clean but my exit face a mess?
That is classic exit delamination (push-down). On the entry side the material is held down by the bulk of the sheet above it; on the exit side there is no support at all, and the tool's axial thrust pushes the bottom ply off as a sheet. Tool wear keeps making it worse: in the experiments by Feito et al. (2014), the worn tool produced the highest exit delamination of all trials. Adding a backing plate is the most direct fix.
What's the best way to remove burrs?
By hand, lightly run fine sandpaper or a diamond file along the edge. Avoid high-speed rotary tools: they tend to round the edge and end up ruining dimensional accuracy. Better still, reduce burrs at the source: use a fresh tool, adjust the fiber cutting angle, or tilt the tool as found by Tima and Geier (2024). And always wear P100-rated protection, since deburring dust is generated even closer to your breathing zone than cutting dust.
Sources
- Taşcı, Bayar, Yüksel et al. (2026), Polymers 18(8):938 (ANOVA contributions for drilling AA7050+CFRP stacks). Accessed 2026-09-03. https://pmc.ncbi.nlm.nih.gov/articles/PMC13120667/
- Huang, Liao, Kai, Su & Hung (2022), Materials 15(9):3392 (ANOVA contributions for ultrasonic-assisted drilling of CFRP). Accessed 2026-09-03. https://pmc.ncbi.nlm.nih.gov/articles/PMC9104629/
- Feito, Díaz-Álvarez, Díaz-Álvarez, Cantero & Miguélez (2014), Materials 7(6):4258 (quantified effect of tool wear on thrust force and entry/exit delamination). Accessed 2026-09-03. https://pmc.ncbi.nlm.nih.gov/articles/PMC5455907/
- Ashworth, Fairclough, Meredith, Takikawa & Kerrigan (2022), Wear 498-499:204340 (effect of tool coating and wear on coupon flexural strength). Accessed 2026-09-03. https://wrap.warwick.ac.uk/id/eprint/164926/1/WRAP-Effects-tool-coating-tool-wear-surface-flexural-strength-slotted-2022.pdf
- Gutiérrez, Meseguer, Muñoz-Sánchez & Feito (2023), Coatings 13(8):1440 (1,403-hole long-duration test; effect of a foam backing plate). Accessed 2026-09-03. https://doi.org/10.3390/coatings13081440
- Zhang, Shi, Wang & Shao (2026), Materials 19(15):3230 (backing preload stress vs. delamination factor; material is 2.5D woven C/SiC). Accessed 2026-09-03. https://pmc.ncbi.nlm.nih.gov/articles/PMC13467053/
- Cerqueira, Matos et al. (2025), Materials 18(7):1595 (pilot-hole strategies and ANOVA on HexPly 8552; a counterexample where cutting speed is more significant). Accessed 2026-09-03. https://pmc.ncbi.nlm.nih.gov/articles/PMC11990906/
- Poór, Geier, Pereszlai & Xu (2021), A critical review of the drilling of CFRP composites: Burr formation, characterisation and challenges, Composites Part B 223:109155 (no standardized consensus on CFRP burrs). https://doi.org/10.1016/j.compositesb.2021.109155
- Tima & Geier (2024), Machining-Induced Burr Suppression in Edge Trimming of CFRP Composites by Tool Tilting, JMMP 8(6):247 (tool tilting reduces burrs by about 50%). https://doi.org/10.3390/jmmp8060247
- Hintze, Hartmann & Schütte (2011), Composites Science and Technology (fiber cutting angle and the critical delamination zone). https://doi.org/10.1016/j.compscitech.2011.08.002
Edge quality is decided by the machine, the tool, and the workholding together
Carbon-CNC machines come with carbon fiber dust protection as standard, with adjustable spindle speed and feed, and can be paired with a vacuum table and dust collection module to tackle CFRP edge quality. To discuss practical parameters and workholding for your parts, contact us.
Get a Quote