I’m searching for/trying to setup a workflow that allows me to calculate time estimations for the cnc machining of metalworking parts by uploading a 2d pdf technical CAD drawing. Basically, I want to be able to upload a 2D drawing of the part (this file has the measurements of the part, ref of the material to be used etc) and the software should be able to give the time estimations for the machining of the part in each of my cnc machines so I can decide upon it. I’m not sure if this is even doable at the moment so I’m asking for clues on the models/tools available to do such a thing. Is there an open source AI model that’s more capable for this kind of stuff? I read about OCR and IDP but to be honest I’m a bit confused over the direction I should go or if this is even doable for an average Joe. This is doable using chatgpt with back and forth conversation but I would need something more direct, without as much interaction and explanation, the idea would be to just upload the file and get the time estimations. Also the 2d drawings are client information, that’s also why I want to host my private AI with n8n instead of using something like chatgpt.
An example of a 2D pdf file could be something like this:
It looks like your topic is missing some important information. Could you provide the following if applicable.
- n8n version:
- Database (default: SQLite):
- n8n EXECUTIONS_PROCESS setting (default: own, main):
- Running n8n via (Docker, npm, n8n cloud, desktop app):
- Operating system:
Hey @itrrmp Welcome to the community 
Just wanted to say this is a very interesting problem! I’ve no clue about CAD, CNC or machining but curious to understand how estimates are calculated and how AI (Vision) models could help?
- Identification of size and materials which could help with job classification?
- Identification of parts and cuts required which can be correlated to a rough, known time estimate?
- Validating drawings and flagging issues/red flags which could assist in estimations?
- Extract the measurements (front,top,sides) from the drawings which could be given to another software to do the estimation?
- Produce rough 3d models of the 2D drawing which could be given to another software to do the estimation?
Theoretically, you could self-host using an open-source vision models but the best are still very much proprietary and definitely not something for the average joe!
I would recommend using Google Gemini which is part of Google Cloud services. Just be sure to stick to the “pro” models as these are stated to not train on user data.
Good luck!
Hi @Jim_Le, really sorry for such a late reply, I just saw it now.
I’m not an expert in CAD or CNC machinning, I’m just working for a factory that specializes on this stuff, and as an IT guy I’m trying to provide automation.
To summarize, our business starts by receiving a PDF from the client with a 2D model drawing of the part (the one I attached before in the post was taken from the web but it’s similar to the ones we receive). Then the comercial and engineering teams analyse the PDF, and do some manual calculations to get an estimate on the time the part would take to manufacture.
This estimations mostly depend on:
- Complexity if the 2D part (dimension tolerances allowed for quality control)
- Type of the material used
- Diameter/measurements of the raw material
- CNC machine used to manufacture the part
- Whether or not it needs heat treatment
Having this time estimation we can then calculate the cost of the manufacturing process of the part, based on our machine’s cost per hour, raw material price, etc.
Getting an estimate from a 2D model (before spending time designing CAM machine toolpaths for a 3D model that takes time and effort) allows us to save time quoting a part that we don’t yet know if we are going to sell.
Regarding the AI models, do you think gemini is superior to openAI’s models on this matter?
This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.