ML-Driven Axial Fan Design Automation
Budget: ₹600 – ₹1,500 INR
Subject: AutoThings – CAD → ML Design Intelligence (Fan Example)
Hi ,
Sharing a quick explanation of the AutoThings idea, using a fan example to make it concrete..
Core idea:
Move from manual CAD + trial‑and‑error to a closed loop:
Parametric CAD → Data → Learning → Optimized CAD
CAD remains central, but geometry becomes machine‑readable and learnable.
Fan example (in brief):
Fan blade modeled in Creo with full parametric control (twist, sweep, chord, spline points)
Geometry parameters exported as data
CFD results (flow, efficiency, noise proxies) added back to the dataset
ML learns how geometry affects performance and noise
Optimized parameters are written back and CAD rebuilds automatically
Designers validate surfaces, manufacturability, and intent
Why it matters:
Same approach applies to compressors, turbines, motor cooling, ducts, structures—any mechanical product where geometry matters.
This is new, exploratory research (limited references, budget‑constrained). If it proves value, it can evolve into reusable IP and potentially a small firm (AutoThings).
Would appreciate your quick thoughts:
Does this make sense from a CAD perspective?
Any concerns or ideas?
Interest in trying a small pilot?
Happy to discuss more over the weekend.
Best,
Rohit
Hi ,
Sharing a quick explanation of the AutoThings idea, using a fan example to make it concrete..
Core idea:
Move from manual CAD + trial‑and‑error to a closed loop:
Parametric CAD → Data → Learning → Optimized CAD
CAD remains central, but geometry becomes machine‑readable and learnable.
Fan example (in brief):
Fan blade modeled in Creo with full parametric control (twist, sweep, chord, spline points)
Geometry parameters exported as data
CFD results (flow, efficiency, noise proxies) added back to the dataset
ML learns how geometry affects performance and noise
Optimized parameters are written back and CAD rebuilds automatically
Designers validate surfaces, manufacturability, and intent
Why it matters:
Same approach applies to compressors, turbines, motor cooling, ducts, structures—any mechanical product where geometry matters.
This is new, exploratory research (limited references, budget‑constrained). If it proves value, it can evolve into reusable IP and potentially a small firm (AutoThings).
Would appreciate your quick thoughts:
Does this make sense from a CAD perspective?
Any concerns or ideas?
Interest in trying a small pilot?
Happy to discuss more over the weekend.
Best,
Rohit