Jur's RV7 Aircraft Factory
4267 hours and counting

The world these days is full of Artificial Intelligence LLMs (large language models) and AI tools that have become available to the everyday user. Knowledge that used to be hidden in libraries and dusty books has now become available and is easily accesible to the average person, which is certainly a positive thing. Most people's knowledge of AI is limitted to the usual chat prompts.

Lately, I have been experimenting with more advanced usage, and I wanted to share my experience and knowledge with other builders as I think it could also benefit their information research.

A word of caution, which is valid for my method as well as for simple prompting using any AI chat engine: Always cross-check the response you receive from the AI model by visiting the referenced websites or sources before actually using the information in the response on your project.

LLM AI tools do "hallucinate". Hallucination is a technical term to indicate that if AI doesn't find a definite answer to your question, it may sometimes invent information based on other - probably non-relevant - sources it identifies. This causes the user to read and rely on false information. Always double-check the responses you receive, certainly if you use them for your aircraft project.

I have been playing with different AI models to see how AI can help me look up information from the web and guide me through the list of manuals that keeps getting longer and longer as I progress in the building process.

For a few months, I had been looking up information through ChatGPT with prompts like, for example: "What is the torque value for an AN924 jam nut on an AN833-6D fitting for 6D fuel lines on my RV7?" It usually comes up with a good answer as it can find the relevant information in torque tables online. It gets more difficult if the information is not publicly available, like with information in proprietary Van's builder plans and manuals.

It would have been useful if AI could actually search my Van's PDF documents and DWG plans as well as other technical install manuals all at once. I discovered the use of "projects" in Google Gemini but soon hit the boundary of how many files you can upload with the free plan of Gemini. Then I discovered Google's Antigravity app. Antigravity is very similar to Gemini, but it is more oriented toward projects and coding assistance and it also uses projects and skills. The main difference seems to be that Antigravity can also work on local folders and the information that resides in them when it searches for information. And the amount of files in the folder does not seem to be limited.

Last days, I have been busy figuring out how to cut the epoxy cloth in such a way that it fits as good as possible in order to avoid a lot of sanding once it's cured. As mentioned in the previous post, I have decided to lay up the glass permanently on the airplane such as it is described in the Vans manual. There are other methods where the windscreen is removed  after laying up, sanding and finalizing it off the airplane and then using sikaflex to attach it such as described in a kitplanes article (Easy slider: part 2) by Larry Larson.
If you go for the Vans method, the layup will have to be more accurate as all sanding can only be done on the airplane, very close to that beautifull PMMA windscreen. 

I decided to make templates just like a tailor would do when making clothes. The Vans method describes starting at 1/2" in the center and gradually increasing the width up to 8 layers by cutting each layer of cloth 1/4 inch wider than the previous. That would bring you to 2 1/2" wide. My masked 3M lines are more or less 3.2" appart. The plans state a distance between 2.5" and 3.5"is fine.
So I have increased the step a bit. Increasing by 13/32" for the first 4 layers, then by 3/8 on the last 4 layers. This will increase the width on each layer by 3/16 on each side, creating a nice step up to 3.2".

The final layer is the most critical one as that one will have to fit as good as possible between your electrical tape masking lines.
The best way I could find was to use plastic sheet which I taped to the windscreen and then used a red sharpie to draw the contours of the tape to the plastic sheet. In the image below, you can see the sheet over the side of the windscreen and upper top fuselage skin.

Below another shot of a plastic sheet to create a template by tracing the exact location of the windscreen fairing

Then I cut out the shape from the plastic and layed it on thick paper and transferred the shape on to the paper. Here you can see the two corner pieces and two curved front pieces.
Then cut out the paper shapes.

Now that all pieces are cut for the final layer. I copied the paper shapes 8 times full size.
Put layer 8 on the side and start sizing down the other paper copies per layer, stepping down 3/8 on each template. Pay attention on which side you need to transition down in width. (eg: on the top of the fariring over the roll bar, you need to transition down to the front but can remain full thickness over the rear as this part will be trimmed down later. You'll have to think about this when trimming down the paper templates. It's a job that requires patience.

All templates are finally done, 72 unique paper templates ready to be transferred to epoxy cloth.

Next day took it all back to the garage and started cutting out the templates from a roll of epoxy weave.

A circular cutting knife works best here. The direction in which you cut the cloth is not important for the windscreen but you'll have learned by now that cutting the cloth at different angles causes different degrees of stability in the cloth and can cause the fibreglass cloth to fray.
I cut out 4 layers and next step will be the actual lay up work.

With the windscreen masked and all surrounding area's protected, it's time for the first step in the epoxy process.

A first glassing job is to make a transition fillet across the windscreen bottom and top forward fuselage skin.
I have first used 80 grit sandpaper to scuff the aluminum and PMMA windscreen thorougly so that the epoxy will attach well and safely to both surfaces.
As the manual also mentions, you want to scuff the windscreen really well and you don't want to have any glossy or shiny area left. 
In the image below, the material is sanded and the surfaces cleaned with degreasing agent.

With this fillet I hope also to cover most of the aluminum clips.

Here is an image of the cleaned PMMA windscreen and you can see there is no gloss left after cleaning. This is now ready to receive the epoxy fibers.
Make sure the masking is done reall well, you don't want to accidentally sand into the windscreen and have your beatifull canopy scratched before it even took it's first flight.
The screws you see are not thightened hard. They just are screwed in without force. The epoxy will cover them and there is no way they can ever come out.

The next 2 pictures show a before and after image. 

On the left image you can see a plastic sheet over the pilot side rear corner which I used to make templates but more about that in the next article.

The gap between the windscreen on the side and the aluminum skin is quite large and it's best to fill this first before laying up glass layers. The better you prepare the surface and the more accurate you cut the epoxy layers, the less work you will have in sanding later on.

I mixed up 2 times about 100 grams of epoxy resin and added black pigment and cotton flox to it. Some people use filler but it would take a huge amount of the high density filler for 100gr of resin.
The black pigment is needed because you can see the epoxy from the inside. and having a different color than the top of the instrument panel wouldn't look nice.
The consistency should be quite dry as you don't want the epoxy to drip down and distort while it's drying.
I applied the epoxy flox with a plastic bag with a tip cut off and then spread it out with a hotel room keycard which I had rounded down on one side approximatly to the radius of the bend on the windscreen.

Here is a forward view of the fillet in place. You may notice there are some gaps in the epoxy. 
I plan to have a second layer with high density filler tomorrow which will fill all these voids.

View from the passenger side. It looks pretty good. It will need minor sanding and a second filling layer but overall, I'm satisfied with the result.

Then added a second layer of 3M electrical tape in red. A different color so I see when I accidentaly would sand through the red layer later, I'll see the black tape underneath.
The peel ply is removed and after some sanding of rough spots most of it looks ready for laying up glass.

There is only one more place which needs some additional attention. The transition on the side is still not thick enough for a smooth transition. This is the case on both sides. 

I will made some more flox and thicken the fill up area to ensure a smooth transition. I added the flox and then used a plastic squeegy against the side skin of the slider to create an even and smooth transition.
You can also spot the rounded tape in the lower rear corner which should allow for a nice rounded corner. 

Some builders make the windscreen epoxy removable and later use sikaflex to put it on. It allows easier sanding than doing it on the airplane.
I was tempted for a while doing it that way but in the end I decided to follow the Vans manual and glass it permanently on the airplane right away. Now it's a matter of being accurate in laying up the glass to avoid excessive sanding later on. 

Work is continuing on painting. Some time back, Brent has painted my wheelpants in white. That already looked awesome.
Yesterday he showed up with the 2 wheelpants finalised using the paint scheme I had given him and I have to say that I'm blown away with the result.

Here is a drawing of the design I had given him.

These are pictures of the both sides of the finished wheel pants.
I'm very satisfied with the color choice I made before and we can now with a peacefull mind continue painting on the other components.

Here is an image of the outboard side

And an image of the inboard side.