How to repair mould gel-coat (lots of photos)

So, I’ve been doing the tidy-up on the mould before I infuse it, and I’ve had a bunch of problems with bubbles behind the gel-coat. These are down to a mistake I made whilst building the damn thing. One puts the gel coat on the mould, or it gets the hose again. Well, I put the gel coat into the mould (spraying), and let it set. Then you put a coupling coat of glass on the back. This is pretty thin (tissue) glass – 100gsm chopped strand. Or you do the wrong thing, and try a new product. I tried a coupling veil, which is like breather fabric. I found it was very very hard to wet out, and hard to make it stick to the gel. What’s more, I also started running out of resin. So, half way down the tub I stopped using it and went back to tissue. When you look at the mould now, you can many more bubbles on one side than the other where I swapped from veil back to tissue.

What I’ve found is that I’ve ended up with a lot of repairs to do to the mould. Here’s my technique – it may work for you.

Step 1 – Find out where the hole is

Once you’ve found it, dig it out with a screwdriver – be sure to probe around the edges – it’s quite surprising how far a run is from a simple small blob.




Step 2 – make it concave

One of the things I’ve worked out is that when getting the gel in the hole for the repair, you can’t get it into every pocket a the edge of the gel .If you’re digging out with a screwdriver you won’t get under every edge, and if you’re coming in down from the top with the screwdriver, you may not lever out the weak points as well where adhesion to material and gel isn’t perfect. So, what I do is put a grinding stone on the Dremmel and go around all the edges until the edge of the hole is convex. Then there’s two advantages – one: no overhanging void that the gel can’t get under and leave a small air pocket you’ll be repairing later; two: I’ve found that it occasionally chips out another void that wasn’t visible at all. It only takes a few seconds to do this. Of course kiddies – safety first. This kicks out a lot of dust and is prone to kicking up gel-flakes, so wear a mask and goggles. It’s also loud so i go for ear defenders. I look like the human fly.

Here it is, all dug out. It’s interesting that I found a lot more void when digging it out.


Step 3 – Apply the gel

Gel-coat is weird stuff, and doesn’t fully set in air – it needs a barrier or it remains tacky. There are two ways to do this – first is add a liquid wax solution to the gel coat (typically at 2%). As it sets, the wax migrates to the surface and forms an air-tight seal. It works if you are doing your repair as a one-off or is perfect if you’re spraying a repair. However, I think there are downsides to this for spot repairs:

  1. If you have to build up a repair, you will have to take off the top layer to get rid of the wax (mould cleaner disolves it away, but you’re using expensive chemicals).
  2. If you are building up a repair, it will leave a relatively smooth surface, so not much of a key
  3. The wax is dissolved in styrene (which should totally gas out though) and epoxy and styrene don’t get along well
  4. It is extra faff adding 2% wax to, say, 30g of gel-coat.

So what I do is make up the repair, and seal it with flash tape. It’s specifically designed for resins not to stick to it, as well as being stable at high temperatures. This works quite well for making deeper spot repairs, and the tape gives the gel-coat lots of support if I’m repairing a vertical surface. If you’re using wax, you have to build it up in multiple layers or else it will run. This method is faster (for me).

Step 4 – Flat it off


I don’t have a picture of the large wound I repaired half-way through, so I’m going with this one. If you run your fingers over the repaired gel, quite frequently you can feel it’s slightly proud. If you can feel it, you certainly will see it on a cosmetic part. Worse, if it’s half a millimetre proud or more, it may give you release issues as well. You can see around the repair that I’ve started to flat it off. I find a 120 grade paper on a small random orbital sander such as this Ryobi. It’s great because it also has a extensible pointy nose thing that gets into the corners. Like a wasps ovipositor. Kind of. Maybe. I just like saying ovipositor.
The white speckles you can see are where the gray top coat has been flatted back to the white undercoat. Once you start seeing white, you stop. The white line at the top of the repair is a slight highlight reflecting the light from the spotlight I’m using. It certainly shows there’s a lip there. What’s more, this was two holes so I knew to keep flitting back until I’d seen the two holes again. It takes about 10 mins and some patience to do it with the Ryobi.

I didn’t want to go any more coarse than 120 grit because I’d be digging deeply into the part. 120 makes good progress, but then flats off well. To flat the part, I used an 800 grit disk on a sander like this one which did it quickly and safely . No gouging. From 800 I went straight to 1500, then onto the polishing compounds.


Here’s the finished version of the holes above, or one similar. What you end up seeing is the original holes filled in. once they’re polished (800 -> 1200 -> 1500 polish ->2000 polish -> anti-swirl polish) the new black gel-coat comes out as shiny as the original.






So, here’s the final result, polished as well. You can even see the little loop in the picture in black where i accidentally scratched the gel coat with the Dremmel. it’s in the third photo from the top. This is a very solid repair and will take multiple pulls if necessary





3d printed plastic that is petroleum resistant

So, NinjaTek have released a new very hard filament which is petroleum resistant – could be useful for making my composite tank. I still need to understand if the regs require foam filling, or if flap gates are adequate. I could then completely print the pump mounts, flap gates swirl pot, etc.

This comes on the back of this post about making tank bladders.

rubber infusion for fuel tank bladder?

I’ve been thinking about an internal bladder when I make my fuel tank, and one thing struck me – infusing aramid with viton rubber (very low viscosity) – wouldn’t that get the best of both worlds – really strong bladder and fuel resistance to boot. then sling the entire thing in a CF/Aramid tank.

I can’t see anyone selling VITON compound on its own though – seems to be as a finalised product.

Any thoughts?

Aramid floor tray is in

Here’s the floor tray, all nicely bonded in.

IMG_3334Here’s the floor-tray all nicely bonded in. You can see on the right of the picture how the floor-tray now replaces the cross-member I removed. I put several extra layers in there as well to pass even more force forward (than the 6 layers of 300gsm that’s already in there).

There’s also extra reinforcement, like a lardy-blokes truss, to take engine mounts if I decide to do that.

So Mark – tell us how you did it

IMG_3316 2


Here’s the part as it came out of the mould – all sharp edges and over-sized. Next job was to put it on the car – mark it up and trim it. That’s the boring bit and I don’t have any photos to document.



IMG_3324This is the chassis before the part is bonded to it. It had been blasted before powder coating, and I used the right high-temp non residue leaving breaker-tape to mask off the mating areas. The blasting heaves a fantastic key. I was worried originally that the tape wouldn’t make a brilliant barrier to rust after coating but it’s worked out fine. The reason being, the coating forms a seal against the tape so there’s nothing to get in – no air or moisture so no rust.

I marked the part up after clamping it down, and drilled for rivets. The rivets were just soft-ally headed ones to provide clamping force rather than to be anything structural.




Here I am as Clampy-McClamp-Face. I had every surface clamped before I drilled rivet-holes. If you drill-rivet as you go along the part, it starts to creep and your holes don’t line up. Clamping the entire thing before drilling keeps the accuracy.



IMG_3325Here’s a close-up of the rivet holes – I’m pretty pleased with how accurately the holes line up. Structurally it’s not really important, but attention to detail matters.








One post-trimming, bonded, riveted part in place. I used an epoxy two-part adhesive. It’s the slightly flexible stuff for parts under a lot of vibration.

Once the adhesive set (24 hours for full cure at 20C) it feels rock-solid.


Air in a composite, under the microscope

So, a while back, I was making a trial part for the tub to understand how much resin a 12 layer infusion would take with a 10mm core. I cocked that up, but since then, I’ve also done an excellent infusion for my floor pan, and I wanted to compare the differences in them both for you under the microscope, because I’m that amazing, informative guy.

Pants Infusion – With Lots of Air in

bubblesIn doing the infusion, I inadvertently admitted some air into the infusion, and it ran over the part. Also, I capped the infusion off once complete rather than letting the pump run for ages to try and pull the air out. Now I know better and know it should be under vacuum until it gels off.

So, what you can see here is a scale at 0.1mm per subdividing line. Ignore the 5mm bit. So, these bubbles are anything between 0.1 and 0.5mm wide. Wherever there’s a bubble, there’s a weakness. To the naked eye, they just look like a very fine dot.

Good Infusion, Where I Got It Right

So, this yellow bitch is going on the car. I had infused the resin at 28 degrees, with the mould also at about 30 degrees. I used a brewing mat under the resin to warm it.

no bubble

This time, you can see no air bubbles, and the weave is easy to see. It’s at the same magnification as the part above, but it’s at a different weave. This is 300gsm twill weave (rather than 2/2 twill) but has less threads per twill than above – more tightly woven if you like. I also let the pump run all night to ensure absolutely no spare resin remained in the part. I also have a new technique to ensure there isn’t any air in the original input pipe, when it’s submerged in the epoxy.

Now, I’ve Keyed it!


In laying up the part, I put some strategically placed 1″ strips of peel-ply in where the chassis rails will be when it’s bonded in. Epoxy doesn’t stick to it peel-ply. When I took the part out and tore off the peel ply, I ended up with a nice keyed surface for the adhesive to the chassis. You can see it here. The crappy red fibers are just bits of the peel-ply I can’t get off. It’s incredibly thin nylon but the red threads are only at the edges. I don’t think they will (at all) compromise the quality of the adhesion.


How much will my tub weigh?

So, based on the sketch I made on the white board (below), I need to calculate just how heavy the tub will be. The main reason I need to calculate this is to be sure the 19kg I’ve taken out by chopping out all that steel and removing the ally panels isn’t then replaced by even more carbon.












I made a trial part, consisting of:

  • 4 * 600gsm carbon (2 above the core, 2 below)
  • 2 * 200gsm e-glass (1 above the core, one below)
  • 1 * 300 gsm aramid (on the bottom, facing the tarmac)
  • 1 * 300 gsm carbon (on the top, to look pretty)
  • 1 * 10mm thick closed cell foam for the core (in the middle)

The layup is symmetrical around the core, apart from the aramid on the bottom and the facing carbon on the top. I have discounted the weight of clear gel-coat applied to the finished part (assume 1kg at the end).

The part measured approximately 103mm * 204mm, and weighed 130g. This meant a unit weight of 0.006 g/mm2.

From this, I fed the dimensions of the panels above into my CAD package. Given a surface extruded to 1mm depth, it will tell me the mass of the panel, to a bazillion decimal places.


Part count Unit Weight Total Weight Running Total
Tunnel Side 2 2.37 4.74 4.74
Tunnel Top 1 1.16 1.16 5.9
Back 2 1.0375 2.075 7.975
Base 2 2.7 5.4 13.375

So, if I go for this, the new tub will weigh 6 kg less than the original steel work.


  • 10mm closed cell foam is used uniformly. This won’t be the case – the sides do not need a 10m core – I will probably go for a 3mm core.
  • the base, back and top of the tunnel need to be strong in bending load, the sides need to be strong in lateral load. As such, I can use a thinner (or even no core) for the sides. I think I will save 1kg there.

Tub nearly finished

so, here is the tub, and here are some of the gaps. The gaps aren’t going to be a problem because all the in the corners need to have a radiused edge. I will be doing this with plasticine and the radius ball. At the end of this I should have really neat lined edges.

Then it’s ready for the mould making process. Actually it’s not, first I need to put all the flanges on in order to actually pu the mould  in this, but it’s nearly there.

I also have some of the polypropylene sheet in the tub which is still a little bit flexible. I’ll be getting in behind that with expanding foam to give it a backing layer to stop it flexing. I will need to cover some of the suspension with release sheet first otherwise the expanding form will stick to it and everything will be an unholy mess. I’m also going to have to take the engine and gearbox out now. I don’t want them to get covered in gelcoat when I start spraying.



Testing Carbon Fibre before I commit – Part 2

So, here is the raw results file:

raw data

I’ve kept it there in the spirit of honesty, but the following distilled data into the graph is what really matters:

new graph


So the numbers that matter are those on top of the line – each of those is the number of KG force required to deflect the part by a given distance. You can correlate the colour to the number to the data in the top chart (click through for details)

Finally, there is the density/deflection ratio, which would show the ideal performing part if absolute strength wasn’t the most desired outcome.



There is a trade-off between core and layers, which is what would be expected. What actually surprised me the most was the difference a core makes.

The top line has two variables set – 4 layers of CF (rather than 2), and a 10mm core, rather than the yellow line, which is 4 layers and a 6mm core. The third highest line is the darker blue line, which is 2 layers and a 10mm core.

So, it’s layers over core but again, there’s a trade-off. The red line is 4 layers over a 3mm core, and you can see it yielded really quickly at 3mm with a very low amount of force (well, 98kg of force). Without doing any statistical analysis, I am observing that each 3mm of core seems to give me an extra 100kg of resistive force before yielding. However, I don’t know how far that scales.

Finally, you can see that there are a selection of flat lines near the bottom, and they are parts made without core. The seem to bend a lot and not yield. For my purposes though, they’re not suitable. Parts 6 and 7 (four and two layers), deflected up to 7mm without yielding, but weren’t much use to me.

As the thicker parts started to yield, we could actually hear them crack (quietly). I’m assuming that’s the fibres snapping. As such, the moment it starts to happen the part is compromised.

Testing Carbon fibre before I commit – Part 1


Some carbon fibre flat pieces were made and tested to see how far they bend for a given load. A load cell was used to measure the force.


So, before I go mad and put several hundred pounds of material into my CF tub, I thought it best that I do some clever investigation into just how does a composite structure bend, yield and break under various loads. So, I made the pieces, tested the pieces, and did some graphs, and came to some conclusions.


8 trial pieces were made of various thicknesses – different cores and layups. Each piece had a few things in common though:

samples 1

  • single layer of 300 gsm aramid for intrusion protection (if it were on the car)
  • this was balanced with a 300 gsm piece on carbon on the opposite layer to the core
  • one layer of 600 gsm either side of the core, and one layer of 200 gsm e-glass either side of the core for a little tolerance and flexibility.


samples 2I could then vary the thicknesses of the cores and extra layers of 600 gsm carbon I wanted to use. I opted for the following core thicknesses:

  • 10mm closed cell foam (CCF)
  • 3mm CCF
  • 6mm (2x 3 layers of 3mm CCF)
  • none


In order to test the bending load, I went to see my good friend Simon at Cornering Force. There are only so many people who like it when you call them at 8pm and say “Hey – do you want to stress test and break some carbon fibre?” After marking up each sample, we then put each on a support, resting on the bed of a mill, and pressed down on it with a load-cell. The load-cell had a 25x50mm contact face. The mill allows us to make vertical adjustments in 0.1mm increments easily, and the load-cell is accurate to the gramme.


The pieces were laid out on one sheet, and infused as one piece, and then cut to size afterwards. The vacuum feed was left on post infusion and for the curing cycle. The resin was degassed. This was to ensure the maximum amount of spare resin was removed until the gel stage. After gelling and the initial cure, they were post-cured at 60C for 12 hours, as per the manufactures instructions. This gave me the maximum strength possible for the sample for the least weight. I decided not to vary curing methods (e.g. length of cure) because a 12 hour, 60C cure is what I can achieve for a part as large as this.

The plan was to work out how much a piece would deflect, how it would yield and fail, and just what were the most significant factors in this – is it core thickness, is it amount of cloth? Did I turn around three times widdershins?

This was also a good opportunity to try a couple of techniques given to me by Vic at Scorpion CDL. I am always astounded at the wealth of technique he has and the quality of work done there.

Lessons Learned

I also had a slight infusion issue, which was the temperature of the resin I infused with. I had it at about 10C which was great for it not going off, but made it too viscous, and meant de-gassing it was a pain. It also meant the infusion was quite slow for such a small part. Further checking says 25C to 30C is a better temperature for this. I should have also wrapped more mesh around the input spiral which would have also alleviated this problem, but that was fixing the wrong problem – get the temperature right – let the resin flow.

Analysis and Conclusions

Coming in the next post.