How ESI builds tooling. Plug construction and the surface standard, draft and parting lines, tooling gelcoat, print-through and exotherm control, stiffening, post-cure and the first pull — plus two full chapters on the long-term semi-permanent release systems this yard runs instead of constant waxing: how to apply them, break them in, evaluate them and reset a tool when they fail.
Before you start
This is the tool manufacture course. A mold is not a part — it is a piece of production equipment that has to survive hundreds of cycles and reproduce a surface finish it can never be better than. Everything in the plug ends up in the mold, and everything in the mold ends up in every part you ever pull from it. Two full chapters are given to release systems, because ESI runs long-term semi-permanent coatings rather than constant waxing, and that is a different discipline. Fourteen chapters, 80% on the final to certify.
Plug → mold → part. Every defect passes down the chain and gets no better. Fix it at the plug.
Sealer plus release agent, applied and cured to a procedure. Not a can of wax and a rag.
A mold with a log, a maintenance record and a known release history is worth far more than the same mold without one.
Chapter 1
A one-off part gets a full production mold. The mold cost more than five of the parts.
Three words get used loosely and mean very specific things:
| Term | What it is | Surface |
|---|---|---|
| Plug (pattern) | A full-size positive model of the finished part | The outside surface of the part, finished to a higher standard than the part will ever need |
| Mold (tool) | The negative taken off the plug | The mirror image. It becomes the part surface |
| Part | What you pull out of the mold | Can never be better than the mold surface |
| Approach | Use when | Cost profile |
|---|---|---|
| Moldless / one-off | A single part, no repeat, cosmetic requirement low. Built over a temporary former or direct onto foam. | Cheapest. No asset at the end. |
| Splash off an existing part | You need to reproduce something that already exists and is in good condition | Fast and cheap, but you inherit every defect and every bit of wear in the donor. |
| Soft / short-run tool | A handful of parts, moderate finish requirement | Middle. Will not survive long production. |
| Production mold from a plug | Repeat parts, cosmetic surface, or a part that has to be dimensionally consistent | Highest up front, lowest per part. Becomes a yard asset. |
That is why tooling is built heavier, from different materials, and post-cured, when a part would not be.
What process will run in this tool, and what does the part surface have to be? Answer that first and the rest of the decisions follow:
Chapter 2
The plug moves half a millimetre overnight. The mold is now wrong and nobody knows why.
A plug is a full-size positive of the part. It has one job: to be exactly the right shape with exactly the right surface, and to stay that way long enough to take a mold off it. It does not need to be strong, it needs to be stable.
| Substrate | Good | Watch out for |
|---|---|---|
| MDF / timber, frames and battens | Cheap, fast, easy to shape, good for developed and chined shapes | Moves with moisture. Must be sealed completely, on every face, including the back and the edges. |
| Foam over a frame | Fast to shape compound curves, light | Soft. Needs a hard skin over it before fairing, and many foams are attacked by polyester and styrene — test first. |
| Tooling board | Dimensionally stable, machines beautifully, holds detail | Expensive. Polyurethane board is not usable if the mold will ever see an elevated cure — PU inhibits epoxy cure at temperature. |
| An existing part (splash) | Fastest route to a shape that already exists | You copy its defects, its wear and any distortion it has taken in service. |
| Composite plug | Stable, durable, survives repeated mold pulls | Most work up front. |
Accuracy comes from a reference, not from eye. Use whatever the job justifies:
Chapter 3
The plug looked perfect matt. Under the first gloss coat you could see every ripple.
The rule from Chapter 1 in numbers: the plug surface must be better than you want the part to be, because nothing downstream improves it. A Class A part needs a Class A mold, which needs a plug finished beyond Class A.
These are different problems and people confuse them constantly:
Is the shape right? Long wavelength. You find it with long boards, battens, sweeps and raking light — not with your fingers and not with fine paper.
Is the surface right? Short wavelength. Grit progression and polishing. A surface can be mirror smooth and still be wavy.
A guide coat is a thin contrasting mist over the surface. As you sand with a long board, the guide coat disappears from the high spots and stays in the low spots — so it draws you a map of the surface you cannot otherwise see. Re-apply it at each stage. It is the single most useful technique in fairing and it costs almost nothing.
Raking light — a light low and nearly parallel to the surface — exposes waviness that overhead light completely hides. Sight down the plug from multiple directions and multiple heights. What you see in raking light is what will show in the finished part under sun.
A high-build primer or surfacer over the fairing gives you a uniform, sandable, non-porous film to block flat and then polish. It also seals the fairing compound — important, because most fairing compounds are porous and porosity in a plug becomes pinholes in a mold.
Work the progression up: block the primer flat with the guide coat, refine the grit, then polish to a true gloss. The plug is finished when it is a mirror with no waviness in raking light and no visible sanding scratch at any angle.
Chapter 4
The mold is beautiful. The part will not come out of it, and it never will.
Everything in this chapter has to be decided at the plug. Once the mold is laid up, none of it can be changed.
Draft is the taper on any surface running in the direction the part comes out. With draft, the part immediately separates from the tool as it lifts. With no draft, the part slides against the tool for its full depth, and every millimetre of that is friction, scuffing and risk. With negative draft, the part is locked in.
An undercut is any feature that hooks the part into the mold so it cannot lift straight out — a return lip, a recess, a bolt boss with a groove behind it. You have four options and you pick one deliberately:
| Option | What it means |
|---|---|
| Design it out | Change the part so the undercut does not exist. Always the cheapest answer if the design allows it. |
| Split the mold | Two or more pieces that separate along a parting line, released individually. Adds cost, adds a witness line on the part. |
| Loose piece / insert | A separate piece of the mold that comes out with the part and is then removed from it. Simple, but it has to be found and refitted every cycle. |
| Flexible tool section | A silicone or flexible insert that deforms out of the undercut — the same principle as the nonskid pattern mould in Composites 2 Chapter 12. |
Where a mold splits, the split leaves a line on the part. So the parting line is chosen for three things at once:
The split is created on the plug with a temporary parting board or flange, sealed and released, so the first mold half is laid up against a clean, sharp, well-supported edge.
The flange is the flat land around the mold, and it does far more work than people expect:
Chapter 5
Somebody waxes over the semi-permanent "to be safe." Now the tool has to be stripped.
Traditional mold release is sacrificial: a layer of carnauba paste wax sits on the tool, some of it transfers to every part, and it is reapplied constantly. Semi-permanents work on a completely different principle, and ESI runs them for good reasons.
The solvent-based semi-permanents are moisture-cure reactive polymers. Chem-Trend describe the category plainly: they are "designed to cross-link with humidity or heat and chemically bond to the mold, thus providing an inert, durable polymeric film that is resistant to wear, chemical and thermal attack." Henkel put the practical consequence directly: "Unlike sacrificial waxes or silicones, Frekote semipermanent mold release agents do not transfer to your parts; instead they chemically bond to the mold surface."
That single difference — bonded film versus sacrificial layer — drives everything else in this chapter.
Bridges porosity, micro-porosity and fine scratches so that the release film laid over it is continuous. TR describe their 910 sealer as protecting against styrene attack, maximising the performance of the release, aiding break-in release and sealing micro-porosity. A discontinuous release film is the mechanical root cause of most sticking.
Supplies the low surface-energy interface that the part will not wet out onto and therefore cannot bond to. It is the layer that actually releases, but it can only do that reliably if the sealer gave it a continuous surface to sit on.
Chem-Trend split this further into a four-step system — Clean, Prime, Seal, Topcoat — with a separate primer under the sealer on rougher or more porous tools. Other manufacturers offer single-product systems that claim no separate sealer is needed. Follow the system you have chosen, complete, and do not mix components between brands.
| Semi-permanent | Paste wax | PVA | |
|---|---|---|---|
| How it works | Chemically bonded, cross-linked film on the tool | Sacrificial wax layer | A physical water-soluble film barrier |
| Transfer to the part | Manufacturers state it does not transfer | Transfers on every pull | Transfers — it comes off with the part and is washed off with water |
| Reapplication | Multiple pulls per application, then touch-up coats | Frequently, often every pull or every few pulls | Every pull |
| Labour | Low ongoing, higher setup | High ongoing — buffing every cycle | Moderate, but it must be sprayed and dried every cycle |
| Surface finish | Class A grades available | Good if buffed properly, operator dependent | Leaves its own texture — not for cosmetic surfaces |
| Part contamination | Low — the strongest argument for it | Wax on every part must be removed before bonding or painting | Washes off, but must actually be washed off |
| Temperature capability | High — some rated to 400 °C and above | Limited | Limited |
| Recovery when it fails | Harder — must be stripped with the correct cleaner | Easy — strip and re-wax | Trivial — wash off |
Semi-permanents are not universally better and you should know the trade-offs:
Chapter 6
Four coats went on in twenty minutes because the shop was busy. The first pull stuck.
Semi-permanents fail for one of three reasons, and all three are procedural: the tool was not clean enough, the coats were not allowed to cure, or the film was not continuous. This chapter is about not doing those things.
Henkel identify the cause of a complete release failure as an insufficiently cleaned mold that prevented the release agent bonding. The film has to bond to the tool. Anything between them — old wax, old release, styrene residue, polish, handling grease — is a place the system is not attached.
There is no universal schedule. These are published manufacturer figures for representative products, shown so you understand the shape of a schedule — multiple thin coats, a defined interval between them, and a final cure before the first part.
| Product | Coats (new mold) | Between coats | Final cure before first part |
|---|---|---|---|
| Frekote 770-NC | 2–3 base (up to 4) | 5–10 min | 5–10 min @ 22 °C |
| Frekote 55-NC | 2–3 base; 3–6 total | 5 min | 30 min @ 22 °C, or 5 min @ 100–150 °C |
| Frekote 700-NC | 2–3 base | 5–10 min | 15–20 min @ 22 °C |
| Frekote B-15 sealer | min 2 | 30 min | 24 h @ 23 °C, or bake 60 min @ 95 °C |
| Frekote FMS-100 (composite molds) | 1–2 | 15 min | 20 min @ 20 °C |
| Chemlease 15 Sealer EZ | 1 seasoned / 2 new or green | 15 min | 1 hour |
| TR-900 sealer | — | min 30 min | min 3 hours after the last coat |
| TR-950 / 955 release | 4–5 with sealer (6–7 without) | 10–15 min | 30 min. Alternate direction between coats |
| Marbocote HP7 | 3 | 5–10 min after dry | 20 min RT (5 min @ 60 °C better) |
| Marbocote 227-CEE (wipe) | 3, plus 1–2 on new or porous | 15 min after dry | 30 min @ 20 °C / 15 min @ 60 °C |
| Easy-Lease CR1 | 6 on a new mold | 15 min | 1 hour after the last coat |
Figures above are quoted from published manufacturer technical data sheets and are shown for training. Always work to the current TDS for the exact product on the tool in front of you.
A brand new mold is not the same as a mold in production and it does not get treated the same way:
Put water on the surface. A high contact angle — water standing up in discrete beads — indicates low surface energy, which is what a working release film looks like. If the water spreads out into a continuous sheet with a low contact angle, the surface is wettable, which means it is suitable for adhesive bonding — and therefore not releasing.
Release force climbing pull by pull, parts needing more wedging, squeaking or grabbing at demold, patchy gloss on the part surface, or a dull area appearing on the tool. These are the warnings that come before a stuck part. Log them.
One important caveat on the water break test: it is recommended to be run on a sample coupon of the substrate, not on an actual composite part surface.
| Symptom | Likely cause | Action |
|---|---|---|
| Complete failure to release | Mold not cleaned sufficiently, so the release never bonded | Recover the part, strip the tool completely, clean to the manufacturer procedure, and rebuild the whole system |
| Sticking in one area only | Discontinuous film — a missed area, or porosity the sealer did not bridge | Strip that area back properly and rebuild. Do not just add a coat over the top |
| Pre-release — part lifting during cure | Too much release, or too many coats | Reduce coat count on the next application and record it |
| Transfer or residue on the part | Wrong product, contaminated film, or an uncured film | Check cure schedule and product. Parts affected must be cleaned before bonding or painting |
| Build-up over many cycles | Repeated touch-ups without a reset | Strip back to the tool or tool sealer with the correct cleaner and start the system again |
Chapter 7
Standard part gelcoat goes on a mold. It crazes on the eighth pull.
Tooling gelcoat is a different product from part gelcoat. It is formulated to be harder, tougher, more heat resistant and far more dimensionally stable, because it has to survive hundreds of cure cycles, repeated release chemistry, cleaning, and being wedged at the flange. Part gelcoat on a mold will craze, chalk and fail early.
The bond between the tooling gelcoat and the first laminate behind it is what stops the surface separating in service. That bond depends on laminating at the right cure state.
| Defect | Cause | Prevention |
|---|---|---|
| Crazing after a few pulls | Film too thick, over-catalysed, or the wrong product (part gelcoat on a mold) | Correct product, measured catalyst, film build to specification |
| Pinholes in the mold surface | Air in the spray, pinholes in the plug, or spraying too dry | Fix the plug, set the gun correctly, control the air supply |
| Fisheyes | Silicone, oil or contamination on the plug or in the air line | No silicone in the shop, traps and filters checked, plug not touched after finishing |
| Surface separating from the laminate | Laminated outside the bond window, or over a contaminated gelcoat surface | Stage everything, watch the window, keep the surface clean |
Chapter 8
Six months later the weave of the mold laminate is visible in every part it makes.
Print-through is the pattern of the reinforcement showing through the surface. It happens because resin shrinks as it cures and continues to shrink slowly afterwards, and it shrinks more where there is more resin. A coarse fabric has thick resin-rich pockets at the weave crossovers and thinner areas over the yarns — so the surface pulls down unevenly and the weave appears.
The cruel part is the timing: print-through often is not visible when the mold is new. It develops over weeks and months as post-cure shrinkage continues, and by then the mold is in production.
The skin coat is the buffer between the tooling gelcoat and the structural laminate, and it is the main defence against print-through. It uses a fine reinforcement — surfacing tissue or light mat — laid immediately behind the gelcoat, with the specific aim of putting a layer of uniform resin thickness and no coarse weave right behind the surface.
A mold laminate is thick. Laying it all up at once generates a large amount of heat in a confined thickness, and that heat causes:
Chapter 9
The mold flexes when you pull a part. Every part comes out slightly different.
A mold laminate has one requirement above all others: stiffness. Not strength. A mold that flexes produces parts that are not the same shape twice, distorts under vacuum, and cracks its own surface at the flange.
You can build a mold twice as thick and gain far less than you expect, because stiffness in a panel comes overwhelmingly from depth — how far material sits from the neutral axis. That is why the answer to a flexible mold is almost never "more plies." It is framing and stiffening (Chapter 10), which moves material away from the surface where it does real work.
What the laminate itself has to deliver is a stable, void-free, evenly thick shell with a good bond to the surface and enough substance to carry local loads at the flange and the demold points.
| Layer | Material | Purpose |
|---|---|---|
| Surface | Tooling gelcoat | The part surface. Chapter 7. |
| Skin coat | Surfacing tissue or light mat | Isolate the surface from coarse reinforcement. Chapter 8. |
| Intermediate | Light to medium mat or light stitched fabric | Build gradually without introducing coarse texture close to the surface |
| Structural bulk | Heavier mat, woven roving or stitched biaxial in stages | Thickness and substance |
| Flange and edges | Extra plies, solid, flat | Vacuum sealing surface, bolt land, wedging area, handling |
| Stiffening | Framing bonded on. Chapter 10 | Where the stiffness actually comes from |
Tooling resins are formulated for low shrinkage and dimensional stability, which is a different priority from a part resin. Composites 1 Chapter 3 applies here with one addition: on a mold, shrinkage is not a cosmetic issue, it is a dimensional one. The whole point of a tool is that it does not move.
Chapter 10
The mold gets lifted by one corner. It takes a permanent twist and every part after that is wrong.
Chapter 9 established that stiffness comes from depth. Framing is how you get it — and how you get it wrong is by letting the framing show up in the part surface.
Bond a stiff rib directly onto a thin mold shell and you create a hard line. The shell is stiff where the rib is and flexible either side, it shrinks differently across that line, and sooner or later the rib shows in the part. Prevent it by:
Chapter 11
The mold gets used the day after it is laid up. It distorts on the third part.
Composites 1 Chapter 4 defined gel, cure and post-cure. On a mold, post-cure is not optional. A mold that has not been post-cured will continue curing in service, and it will do that while it is being heated by the exotherm of the parts you are making in it — so it moves, distorts and prints while you are trying to produce with it.
Chapter 6 covered break-in of the release system. The first pulls also tell you about the tool itself, so treat them as a test rather than as production:
| What you see on the first part | What it is telling you |
|---|---|
| Pinholes or pimples on the part surface | Pinholes in the mold surface. Fix them now, before they are in a hundred parts. |
| Print-through of the mold laminate | Skin coat or exotherm control was insufficient. Chapter 8. |
| Hard release, squeaking, heavy wedging | Release system, draft, or an undercut. Diagnose before the next pull. |
| Dull or patchy gloss | Release film not continuous, or the tool surface needs polishing. |
| Dimensional error | Plug error, or distortion in cure. This is the one that may mean a new mold. |
Chapter 12
"How many pulls since the last release coat?" Nobody knows. Nobody wrote it down.
A mold is production equipment. Equipment that is maintained on a schedule lasts; equipment that is maintained when it fails does not.
Mold repair is Composites 2 done to a higher standard, on a surface that will reproduce your work in every part from now on.
This is the deliverable of this chapter. Every ESI tool has a log and it records:
| Field | Why |
|---|---|
| Tool number and description | Identity |
| Build record | Gelcoat product and batch, catalyst, laminate schedule, resin system, post-cure achieved |
| Intended process | Hand layup, infusion, prepreg — and the maximum temperature the tool is rated for |
| Release system | Sealer and release products, date applied, coat count, cure conditions |
| Pull count | Since the last full release application, and lifetime |
| Touch-ups | Date and coats |
| Release problems | Every one. This is the data that finds the pattern. |
| Damage and repairs | What, where, when, how repaired |
| Vacuum test result | For any tool used for bagging or infusion |
Tools do not last forever. The honest signs that a mold is at the end of its life:
At that point the question is whether the plug still exists. Which is why Chapter 11 said keep it.
Chapter 13
A perfectly good polyester mold goes in the oven at 120 degrees. It comes out a different shape.
Everything in this course so far assumed an ambient cure. Prepreg changes the requirement completely, because the tool has to go into the oven with the part and come back out the same shape.
Easy Composites note that curing prepreg requires precisely controlled temperature — typically 100 to 150 °C — with good air circulation and vacuum maintained throughout. Your tool has to live through that, repeatedly.
| Tooling material | Prepreg compatibility | Note |
|---|---|---|
| Tooling prepreg (composite tool) | Fully compatible | The tool is made from the same technology as the part. Very close thermal match. |
| High-temperature epoxy tooling | Fully compatible | Purpose-built high-temperature systems, properly post-cured above the intended cure temperature. |
| Aluminium | Fully compatible | Heats and cools quickly. High thermal expansion — see below. |
| Stainless steel | Fully compatible | Stable, heavy, slow to heat. |
| Toughened glass | Fully compatible | For flat sheets and panels. |
| Vinyl ester tooling systems | Semi-compatible | Easy Composites note this can affect surface finish and that maximum cure temperature must be considered. |
| Polyurethane tooling board | Not compatible | Easy Composites are explicit: at elevated temperature polyurethane inhibits the cure of epoxies. |
Compatibility list quoted from Easy Composites out-of-autoclave prepreg guidance. Confirm against the tooling and prepreg data sheets for the specific products before use.
Every resin has a temperature above which it softens and loses stiffness. A tool taken above that temperature will move, and it will move while it is under vacuum and carrying a part. The tool must be post-cured to a temperature above the cure temperature it will see.
Tool and part expand at different rates as the oven ramps. If they differ a lot, the part is being stretched or compressed against the tool through the whole cycle — which shows up as dimensional error, surface defects, or a part that grips the tool at temperature.
This is why tooling prepreg is the natural answer for prepreg parts: a composite tool made from similar material expands at a similar rate to the part, so the mismatch largely disappears. Metal tools are used where their durability and heat transfer justify managing the mismatch.
Chapter 6 listed maximum service temperatures because this is where they matter. Published ceilings vary enormously across the market — some sealers and releases are rated to 400 °C (750 °F) and one TR product to 900 °F, while Marbocote solvent products are rated to 250 °C and Easy-Lease CR1 has a working ceiling of 175 °C.
Chapter 14
The best mold builder in the yard retires. Everything he knew leaves with him.
Tooling is the highest-leverage work in a composites yard, because a tool multiplies. A one-hour improvement to a repair helps one boat. A one-hour improvement to a tool helps every part that tool ever makes.
Tool building is slow, infrequent and full of judgement calls. That combination is exactly where knowledge disappears: a tech does something clever once, nobody writes it down, and eighteen months later the next tool is built without it. Standard work in the tool room is not bureaucracy — it is the only way the yard accumulates capability instead of repeating a learning curve.
| Metric | What it tells you |
|---|---|
| Pulls per release application, per tool | The number no manufacturer will give you. The whole basis for a recoat standard. |
| Fairing hours per plug | The biggest single labour item in plug work |
| First-pull defect count | How good the tooling process actually is |
| Release failures per tool per year | Which tools have an underlying problem rather than bad luck |
| Time from gelcoat to skin coat | Whether the bond window is being hit consistently |
| Tool lifetime pull count | The real cost per part of a tool |
A repair being hand laid over and over. A part being fabricated from scratch every time. A tool converts hours into a pull.
Hard demold every time, a flange that always leaks, a mold that always needs a touch-up early. Those are design and maintenance problems with findable causes.
Two techs building the same kind of tool differently and getting different results means the yard has no standard yet. Say so.
Plugs thrown out, templates binned, station frames scrapped, pattern moulds lost. Every one of those is a future job starting from zero.
And the other half: flags are aimed at processes, not people. "This tool takes forty minutes to demold" is a flag. Naming whoever built it is not.
Final certification
Thirty-six questions drawn from all fourteen chapters. You need 80% to pass. Put your name on the certificate and your completion is recorded to ESI training automatically.