When Your Monocoque Core Geometry Conflicts With Your Automated Fiber Placement Paths

You've got a beautiful monocoque core design—curved, tapered, maybe even doubly curved. Then the AFP programmer sends back red flags: tow steering violations, gaps, overlaps, or paths that just don't fit. The geometry and the placement algorithm are fighting each other. Who wins? It depends on where you're in the workflow. This article walks through the choices, the trade-offs, and the gotchas when core geometry and AFP paths clash.

Who Must Decide and By When?

Role of the design engineer vs. manufacturing engineer

Design owns the core geometry. Manufacturing owns the fiber paths. The conflict sits squarely between them—and neither side can resolve it alone. I have watched design engineers lock a core profile for aerodynamic perfection, only to hand it to manufacturing and discover the AFP head can't navigate the curvature at all. The catch is that both teams must agree on a single decision: which gets priority, and by how much. There is no neutral ground here. One side bends, or the part fails.

Honestly—most teams skip this negotiation. They assume the software will find a workaround. Then tooling gets cut, the first ply steers off course, and suddenly you're scrapping a twenty-thousand-dollar mandrel. That hurts. The decision can't be delegated to a simulation default; it requires a human trade-off meeting. Design engineers need to understand that a 0.5° draft angle change can save two weeks of path programming. Manufacturing engineers need to accept that a local curvature constraint might degrade the laminate’s load path by 3%. Neither is right. Neither is wrong. But the call must be made together.

‘A core-AFP conflict is not a bug in the model. It's a boundary condition your team must own before the spindle turns.’

— senior composites engineer, after a 400-hour rework cycle

Decision gates in the NPI timeline

New product introduction has three critical gates: concept freeze, detailed design release, and tooling commit. The conflict must be resolved before tooling commit. I have seen teams push the decision past gate two, assuming the AFP vendor will “figure it out.” They don't. The fiber placement head can't bend the rules of physics—only the team’s assumptions. What usually breaks first is the radius at the core’s trailing edge: too tight for the tow width, and you get gaps or overlaps that require manual rework. That rework doubles the cycle time per part. Worse, it introduces variation you can't inspect away.

The right gate is detailed design release. At that point, you still control the core draft, the ply sequence, and the tooling surface. After tooling commit, any geometry change triggers a change order that can take three weeks and cost five figures. One rhetorical question belongs here: is your schedule tight enough to absorb that delay? Most teams answer “no” after the fact. Do it before the steel is cut.

Consequences of delaying the choice

Delaying feels safe. It's not. The immediate consequence is a re-run of the AFP simulation—same geometry, different parameters, same dead end. That eats days. The hidden consequence is worse: the longer you wait, the more entrenched each side becomes. Design argues the core is frozen. Manufacturing argues the path is impossible. Neither budges, and the project manager is forced to escalate to a director who has never touched a ply book. Wrong order. The director makes a call based on budget, not physics, and you end up with a compromise that satisfies nobody.

We fixed this by setting a hard deadline: seven days after detailed design release, the core geometry gets a “manufacturing veto” review. If AFP can't lay a continuous 0° ply over the primary load path, the design must add a 2° draft or a flat facet. No exceptions. That rule cut our conflict resolution time from three weeks to four days. The trick is not to negotiate the deadline—it's to enforce it before tooling commit. That's the only gate that matters.

Three Ways to Resolve the Conflict

Option A: Modify core geometry to suit AFP paths

The most direct fix — alter the core itself. You shave a radius here, flatten a curvature there, or even redesign a whole panel segment so the fiber placement head doesn't have to fight the surface. I have watched teams spend three weeks on AFP programming only to realize a 2° draft-angle tweak on the core would have saved them fourteen days of path optimization. The trade-off is brutal, though: change the core and you risk invalidating the structural FEA, the mold machining, and the bond-tooling that was already ordered. One aerospace shop I worked with shaved 4 mm off a honeycomb ramp radius to accommodate a 12.7 mm tow width — the AFP ran beautifully, but the subsequent shear-test panel failed because the core-to-skin bond line got too thin. That fix cost them a full re-qualification cycle. So ask yourself: can the core tolerate a geometry change without cascading into re-certification hell? If yes, do it. If no, look at Option B.

Option B: Adjust AFP parameters to fit core geometry

Keep the core as-is; force the AFP machine to adapt. This means narrower tows, steered courses, variable angle-ply sequences, or even segmented placement where the head lifts and re-approaches mid-ply. The catch is speed — or rather, the loss of it. A machine that normally lays 30 kg of carbon per hour might drop to 8 kg per hour when it has to steer around a core radius tighter than its natural tow-carrying capacity. Most teams miss this: AFP parameters are not infinitely flexible. Tow wrinkling at the apex of a steep core ramp is almost guaranteed if the steering radius dips below 1.5× the tow width. I have seen a fuselage barrel section scrapped because the programmer tried to squeeze a 6.35 mm tow into a 4 mm core curvature — the fibers buckled, the void content hit 11%, and the ultrasonic scan looked like a lightning storm. So Option B preserves the core design but eats cycle time and may introduce defects. You need a realistic defect-acceptance threshold upfront, not after the first five parts fail NDI.

Field note: motorsport plans crack at handoff.

Option C: Hybrid approach with local reinforcements

The pragmatic middle ground — modify neither the core nor the AFP path completely. Instead, you leave the core geometry untouched but add local dry-fabric patches, pre-cured doublers, or hand-laid overwraps exactly where AFP path conflict occurs. This is not elegant. It works. A racing-yacht mast manufacturer I visited uses precisely this trick: the AFP head lays 90% of the monocoque automatically, but at the core-ramp transitions where the placement head can't maintain compaction pressure, they insert a 0/90° woven patch by hand before the final cure. That patch adds 0.3 mm thickness locally — and has never caused a porosity failure in three production years. The downside is obvious: hybrid means manual steps, which means process control variance, which means you now have two QA signatures per ply instead of one. The hybrid approach also complicates your cure cycle because the local reinforcement changes the thermal mass and resin flow in that region. One Formula One supplier tried hybrid patches without adjusting the debulk sequence — the reinforcement zone starved of resin while the surrounding AFP laminate became resin-rich. A complete rework of the cycle parameters was needed. So the trade-off is complexity for flexibility: you keep your core and your AFP program mostly intact, but you introduce manual intervention and a new family of process variables. That said — when done right, hybrid is the only path that doesn't force a geometry or speed sacrifice. It just forces you to manage a more complicated manufacturing process. Most teams skip this option because it feels like a half-measure. It's not. It's a tactical retreat that lets you hit structural targets without redesigning the entire core or accepting 60% slower layup rates.

“The hybrid patch approach saved our monocoque core design when AFP steering failed at the saddle point. Two years later, we still use it on every new program.”

— Composite process engineer, marine OEM (private correspondence, 2023)

What to Compare When Selecting a Path

Mechanical performance impact

The first lens is structural. A fiber path that fights the core geometry creates local stress concentrations — and those spots fail first. I have watched a part pass static load tests only to delaminate at the 47th thermal cycle because the tows were draped across a radius they were never designed to follow. Compare ply-by-ply strain maps, not just ultimate strength numbers. The catch: a path that aligns perfectly with the core may produce zero defects on the machine but shift the load path into a resin-rich zone. That shift kills fatigue life faster than a visible wrinkle ever could. Most teams skip this — they compare layup speed instead of asking whether the laminate will hold under 10,000 cycles. Don't be that team.

Manufacturing cycle time

Time is the second axis, but it lies if you measure it wrong. Raw deposition speed means nothing when the robot pauses every four meters to recut a bridging tow. The real metric is floor-to-floor time per part, including rework. A conflict that forces the AFP head to lift and re-engage adds 30 seconds per event — that compounds fast on a 200-layer stack. However, a slower path that runs without interruption often wins on total throughput. I have seen a shop double output by accepting a 15% slower feed rate because the machine never stopped. One rhetorical question: would you rather watch a fast robot stall repeatedly or a steady robot finish before lunch?

Cost and scrap rate

Money talks last. A path that avoids all core conflicts might require a custom mandrel or a pre-formed insert — that tooling cost hits the budget once, but the scrap savings recur every shift. Compare the cost of one scrapped monocoque against the amortized expense of a modified core. The numbers shift fast when your material is prepreg running at $150 per kilogram. What usually breaks first is the scrap budget, not the engineering schedule. Fragments tell the story: wrong path equals scrapped shell. Right path equals repeatable output. One team I worked with chose a path that added 12% to cycle time but cut scrap from 18% to 3% — that trade-off paid back in fourteen days. Run your own numbers before you commit to a decision based on deposition rate alone.

We optimized for machine speed and buried a stress riser that cracked in qualification. Three months of work, gone.

— Senior manufacturing engineer, aerospace tier‑1 supplier

Trade-Offs Table: Core Geometry vs. AFP Paths

Steering radius vs. core curvature

The AFP head can only bend tows so far before buckling or wrinkling. Your core geometry, meanwhile, might demand a tighter curve than the steer radius can handle. That mismatch shows up fast: tows lift at corners, gaps open along the inner radius, or worse—the fiber slips entirely and you scrap a part mid-layup. I have seen teams spend two weeks dialing in a core profile, only to discover the AFP head physically can't follow it. The catch is that steering limits are not negotiable. You can slow the layup speed, adjust tow tension, or pre-heat the material, but the minimum radius is a hard floor. If your core curvature dips below that number, you either redesign the core or accept that the AFP path will deviate. Most teams skip this check until the first dry run fails—then they panic.

Gap tolerance vs. structural integrity

Gaps happen. The AFP head leaves deliberate spaces between tows to manage steering and prevent overlap—but those gaps become stress concentrators. A 0.5 mm gap in a highly loaded region can reduce compressive strength by 12–18 percent. The trade-off is harsh: tight gap tolerances force the AFP path into straighter, simpler routes that may not match the core's complex curves. Looser tolerances let the head follow the geometry but punch holes in the laminate's load path. What usually breaks first is the shear web—a gap there initiates delamination under cyclic load. One client ignored this for a wing rib core; the seam blew out during static testing. They lost a week re-laying the part. The fix was a blend: tight gaps on the primary load axis, wider gaps everywhere else. Not ideal, but structural integrity won.

Tow width vs. core draft angle

Wider tows cover more area faster—great for production speed, terrible for conforming to steep draft angles. A 12 mm tow on a 45-degree core draft leaves exposed substrate at the edges; the AFP head can't steer the tow's full width into the valley. Narrower tows (3 mm or 6 mm) follow the draft better but double layup time and increase splice count. That sounds fine until you realize each splice is a potential defect site. I have watched a shop switch from 12 mm to 6 mm to solve a draft-fit issue, only to introduce so many tow splices that the NDT reject rate hit 22 percent. The real question: can you reorient the core's draft to something shallower than 30 degrees? If not, pick the narrowest tow that keeps splice density under your internal limit. Most engineers overestimate the core's manufacturability—they design for stiffness, not for the AFP head's actual reach.

'We cut the draft to 25 degrees and dropped tow width from 12 mm to 6 mm. Splice count went up 40 percent, but reject rate dropped from 18 percent to 3 percent. Worth the slower layup.'

— Senior manufacturing engineer, aerospace tier-1 supplier, composite monocoque program

Reality check: name the engineering owner or stop.

Right order: check draft angle first, then pick tow width, then validate gap tolerances against structural requirements. Wrong order: start with a two-day AFP simulation and find the conflict during the first physical layup. The table above is not academic theory—it's a checklist that keeps your core geometry and AFP paths from fighting each other on the factory floor. Pick the trade-off that matches your part's critical load path, not the one that looks cleanest in the simulation.

Steps After You've Chosen

Update the CAD model with approved changes

Open your native CAD—CATIA, NX, or whatever lives on your desktop—and freeze the approved compromise. I have watched teams argue for two weeks over a 3-millimeter core shift and then leave the model untouched for another three days. Don't do that. The moment your cross-functional meeting signs off on either recontouring the core or adjusting the AFP tow-steering angles, someone needs to lock that geometry into the digital twin. Export a fresh STEP or .CATPart with a clear version tag: ‘CORE_v4_AFP_compromise’. That tag saves phone calls at 2 AM later. Delete the old candidate files. Ambiguity here multiplies downstream waste—machinists cut the wrong core, robot programmers load stale ply books, and suddenly the conflict you just resolved reappears in the CMM report. Clean house fast.

Re-run AFP simulation with new parameters

Now push the updated model through your fiber placement simulation suite—Vericut Composite Programming, CGTech, or whatever handles your tow-by-tow logic. The catch is: don't use the same simulation settings that failed before. Change one variable at a time. Start with tow width or steering radius limits. I once saw a shop re-run the same 15-minute simulation four times without updating the collision tolerance—they were debugging ghosts. What usually breaks first is the transition zone where the core taper meets the flat laminate: the AFP head either lifts off or overcompresses the prepreg. Tweak the roller compaction force there. If your simulation software offers a ‘drape failure’ heat map, turn it on. Let the red patches tell you where the next conflict will surface. Run three iterations minimum: nominal, plus a +2% and −2% core thickness tolerance band. Production parts rarely land exactly on nominal. Your simulation should survive that slop.

‘We re-ran the simulation after changing the core chamfer by 1.5 degrees—the AFP paths suddenly converged. Two hours of work versus two weeks of scrap.’

— Senior manufacturing engineer, aerospace Tier 1 supplier

Validate with test coupons before production

Simulation passes on screen. Real fiber doesn't forgive. Cut three test coupons—one at the worst-case conflict zone, one at the nominal area, one at the opposite structural boundary. Lay up each coupon on the shop floor using the exact robot program you just validated. Watch the first ply go down. Listen for the hiss of trapped air or the snap of a tow breaking under tension. I have stood next to an AFP head and heard that snap—it's the sound of a rejected part. Subject those coupons to ultrasonic C-scan and, if your budget allows, short-beam shear testing. Track porosity levels and fiber waviness. If the conflict zone shows void content above 2%, go back and soften the core-AFP transition. One more loop: tweak, simulate, cut, scan. That rhythm is cheaper than scrapping an entire fuselage panel. The validation step is where your decision becomes real data—or real pain. Choose the data.

What Goes Wrong If You Ignore the Conflict

Delamination at steering points

You force the tow to turn where the core says no—and the material answers with silence. Then a crack. Then a non-destructive inspection flag that kills the part. I have watched a single steering point delamination cascade through an entire cure cycle, turning a twelve-hour autoclave run into a scrap tag. The problem is simple physics: the AFP head can steer a tow by twisting it, but if the core geometry below that tow has a sudden radius change, the fibers lift. They want to straighten. The core doesn't give them room. That gap becomes a void, then a delamination under load. What usually breaks first is the interface between the first ply and the core surface—not the core itself, not the laminate stack, but the millimeter-thin bond line that nobody checked after the path was written.

Wrinkles and bridging in corners

Corners are liars. They look fine in the CAD model. In the physical layup, the tow bridge spans a concave radius like a suspension bridge, touching nowhere in the middle. I fixed one of these on a prototype wing rib by hand-adding darts every six millimeters—a full shift of work that the automation was supposed to eliminate. The catch is that bridging doesn't show up until the first ply is compacted, and by then the AFP program has already run three more plies on top. You get a wrinkle train. Those wrinkles create local thickness buildups that shift subsequent ply boundaries, and suddenly your steered paths no longer align with the core contours they were optimized for. Wrong order: the core was cut first, the path was simulated second, and nobody re-ran the compaction simulation after the first physical layup trial.

“The tow path was beautiful on screen. On the tool it looked like a failed road project—potholes and detours everywhere.”

— process engineer, after a third iteration on a monocoque floor beam

High rework and scrap rates

Rework is the quiet killer. Not the dramatic delamination, not the visible wrinkle—the slow burn of cutting out bad plies, recutting new tows, re-running the AFP head for one section while the rest of the part waits. That hurts cycle time more than any single failure mode. I have seen a shop spend four hours patching a single corner that should have taken twelve minutes. The core geometry conflict was written off as a “minor offset”—until the offset multiplied across forty plies and every tow end landed in the wrong place. Most teams skip this: they compare core and path alignment once, during the design review, and never again. Then the first production run hits, and the scrap bin fills with parts that looked correct on paper but delaminated under vacuum.

Honestly—the worst outcome is not the scrap itself. It's the lost confidence. When operators start adding manual overwraps to every AFP course because they don't trust the path, you have lost the entire value of automation. The machine becomes an expensive tow dispenser, not a production system. That's the real cost of ignoring a core-geometry conflict: you pay for automation, but you get hand-layup labor with robotic packaging.

Field note: motorsport plans crack at handoff.

Quick FAQ on Core-AFP Conflicts

Can I always modify the core?

No — and that assumption burns a lot of schedules. If the core is a variable-thickness syntactic foam or a shaped honeycomb block, you can often machine or shim it. But if your core is a cured co-cured structure, say a stiffened panel that already passed NDI, touching it means re-qualifying the whole bond. I have seen teams spend three weeks debating a core modification only to learn the customer’s spec forbids it. The catch: core changes cascade into mold rework and diaphragm re-layups, which cost more than an AFP path tweak. Ask first: is the core a frozen datum? If yes, you move the path.

Do AFP machines have adaptive algorithms?

Some do, but not the kind that fix serious geometry conflicts. Modern AFP heads can steer tows within ±3° and compensate for local draping — that handles gentle double curvature. What they can't do is wrap a 6-mm-thick core edge that drops 45° in 10 mm of travel. The head will either bridge the gap (leaving a void) or buckle the tow at the radius. I fixed one job by writing a custom layup rule that split the zone into three sub-paths — the machine’s “adaptive” algorithm only had two variables. So: trust the built-in logic for surface waviness; distrust it for abrupt core steps. Manual path editing or intermediate tow cuts usually save the day.

“We assumed the robot would handle the core ramp. Instead, it laid a 50-mm-long wrinkle that looked like a crow’s nest. We cut it out by hand — two shifts lost.”

— Manufacturing engineer, aerospace Tier 1 supplier

Is manual ply cutting ever better?

Yes — but only in narrow, predictable zones. If the conflict lives on a core edge that's ≤20 mm wide and the adjacent AFP path is stable, cut a slit in the ply at the boundary. That lets the tow relax instead of fighting the drop-off. The pitfall: manual cuts introduce dry-fiber edges and potential delamination sites, so you must inspect the cut zone with a 10× loupe afterwards. What usually breaks first is the operator’s consistency — three different technicians produce three different slit lengths. Better approach: write a small NC program that pauses the head at the conflict line, lets you snip, then resumes. That hybrid workflow works when full AFP re-layup would cost double the cycle time. Just don't use manual cutting for more than 5 % of the part surface; beyond that the seam quality scatters. A single short anecdote: we once avoided a core re-machining by slitting four plies at a 15-mm radius — saved 14 hours of AFP reprogramming, passed ultrasonic inspection. The client still uses that trick on similar parts.

Wrong question to ask: “Which is always better?” The right question: “Which moves the conflict to a lower-risk location in the trade space?” Modify the core only if the path is frozen and the change is shallow. Let the AFP adapt if the geometry deviation is smooth, not a step. Cut manually only when the zone is small and the team can standardize the cut depth. Pick one, test it on a 300-mm coupon, then scale. That's how you break the standoff without re-running the whole FEA model.

What to Take Away (No Hype)

Start the conversation early

The single most expensive mistake I keep seeing? Teams let the core geometry get finalized before AFP ever looks at the part. That feels efficient — separate teams, separate tools, separate deadlines. Then the collision shows up six weeks later, three prototypes deep. You don't need a full digital twin on day one. You need a handshake — a quick surface check between the core modeller and the AFP programmer before anyone stamps a revision. That meeting costs an hour. Rewinding a cured monocoque costs a week and a lot of carbon dust.

Most shops skip this because they assume the geometry is 'close enough' and AFP can adapt. It can't — not without penalties. The core's local curvature dictates how a tow bends; AFP machines enforce a minimum turning radius. If those two numbers fight, the machine stops or the material wrinkles. Honest—I have watched a perfectly designed core sit on a shelf for three months because nobody checked the tow-steering limit at the concave fillet. A thirty-second measurement would have caught it.

Use simulation to explore options

Simulation here is not about making a glossy animation for the client. It's about asking 'what breaks first' on your specific geometry — not generic rules of thumb. Run the AFP path generator on the as-designed core. Watch where the simulation flags a gap, a wrinkle, or a steering violation. Then ask: can we shift the ply boundaries by 15 mm? Can we add a dart in the core at that radius? The answer is often yes, but you need the simulation output to force the conversation.

'We ran the sim, found three steering violations, and fixed two by rotating the ply stack 5 degrees. The third required a core chamfer — took one day to approve.'

— Process engineer, aerospace Tier 1, off the record

The catch is that simulation is only useful when you treat the results as negotiable, not as gospel. A violation flagged at 0.1 mm over-steer might be harmless in a low-load zone. Another at the same magnitude near a bolt hole will fail fatigue. Simulation gives you the map; you still have to read the terrain. Most teams stop at 'red light, stop' and never ask whether that red light applies to their load case.

Document the rationale for future projects

When you resolve a core-AFP conflict, write down the reasoning — not just the final decision. Why did you choose to modify the core instead of re-routing the tow? What was the cost difference? Did the change affect cycle time? That record becomes your shortcut on the next part. I have seen the same conflict reappear across three product generations because the original team left a toolpath file and a blank email.

A simple table works: conflict description, options considered, decision, rationale, date, approver. No need for a database — a shared spreadsheet beats institutional memory every time. The trade-off table from earlier in this blog is a starting template; adapt it to your tolerances and your machine's limits. What hurts most is rediscovering a fix that already existed in a different folder. Documentation is not bureaucracy — it's the difference between solving a problem once and solving it every time a new core ships.