Marco¶
Original OneNote page: Survey and research - Marco
Route: meeting/person note
Categories: people_meetings, product_technical_or_learning
Source¶
- Word export:
C:\Users\adeel\OneDrive\100_Knowledge\203_TextCAD\01_Product_Project_Management\TextCAD_Wiki\inbox\onenoteexport\01_PROJECTS\Gauss Compute\Survey and research - Marco.docx - MHT export:
C:\Users\adeel\OneDrive\100_Knowledge\203_TextCAD\01_Product_Project_Management\TextCAD_Wiki\inbox\onenoteexport\01_PROJECTS\Gauss Compute\Survey and research - Marco.mht
Extracted Notes¶
These notes are extracted from the Word export so the source is visible in the wiki without moving or deleting the original file.
- Marco
- Here’s a clear, polished, and structured version of your interview notes with Marco Losurdo (CFD Expert) — formatted so it reads like a professional research/interview summary and is ready to feed into your AI-in-CFD or CAD-LLM idea documentation:
- Interview Notes – Marco Losurdo
- Role: CFD Expert
- Tools Used: Siemens STAR-CCM+ for ESU (Energy Storage Unit) thermal simulations
-
- Main Issues and Pain Points
- 1.1 Watertight Geometry
- The most frustrating issue: non-watertight CAD models.
- For simulation, geometry must be perfectly closed — no gaps or overlaps.
- Official CAD models contain built-in tolerances (by design intent), which must be manually fixed.
- Engineers often have to extend or merge surfaces and solids to create one continuous volume before meshing.
- 1.2 Naming and Surface Management
- When importing from STEP or STL, all surface names and attributes are lost.
- This makes it harder to assign boundary conditions, materials, or regions consistently.
- Requires manual re-naming and re-assignment each time → repetitive and time-consuming.
-
- Meshing Challenges
- Meshing time heavily depends on the geometry complexity.
- For a full ESU model, meshing can take 12 hours to a full day.
- Possibility of batch meshing:
- STAR-CCM allows creation of macros for automating meshing sequences.
- Still requires manual setup and monitoring — not fully autonomous.
-
- Simulation and Solver Performance
- Typical simulation runs take several days to complete.
- Biggest wish: reduce solver time.
- Geometry preparation is a one-time task.
- Once geometry is finalized, most of the workflow time is consumed by solving.
- Physics-based ML could potentially reduce runtime, but:
- It only works if the geometry and boundary conditions are similar to the training data.
- Otherwise, results are unreliable (“otherwise it’s bullshit”).
-
- Desired Improvements / Wishlist
- Solver Time Reduction
- The top priority for productivity gains.
- Even small reductions in solver iteration time translate into major savings.
- Intelligent Post-Processing
- Ability to define requirements and constraints (e.g., temperature limits, pressure drops).
- Solver should automatically evaluate results against those requirements and generate graded reports (pass/fail or performance score).
- Pre-Processing Enhancements
- Major issues during pre-processing:
- Identifying and fixing contacts and gaps between parts.
- No current robust automatic solution in mainstream CFD tools.
- Contact algorithms rely on manually defined tolerances or parameters, often tuned by trial and error.
-
- Summary of Key Opportunities for AI Integration
- Workflow Stage
- Current Problem
- AI / Automation Opportunity
- Geometry Cleaning
- Manual fixing of tolerances, closing gaps
- AI-based geometry repair or feature recognition
- Surface Naming
- Loss of metadata after import
- Smart surface re-labeling using context (e.g., geometry + prior runs)
- Meshing
- Long, repetitive
- Batch meshing automation with adaptive learning
- Solving
- Days of computation
- Physics-informed neural nets to accelerate convergence
- Post-Processing
- Manual checks vs. requirements
- Automated evaluation + report generation
- ✅ Key Insight:
- Marco emphasized that once geometry is ready, simulation time dominates — hence AI tools that reduce solver or post-processing time would bring the highest real-world value.
- Would you like me to now summarize this into a short “Interview Insight” paragraph (like for a market research section of your business plan or pitch deck)? It would read like: “According to Marco Losurdo, a CFD expert at … the most time-consuming step in CFD is …”
- From https://chatgpt.com/c/68d11b92-4fd4-832a-bb58-42235673c2f3
- Interview with Marco Losurdo
- CFD Expert
- Please structure the notes
- Notes from tht chat
- Interview with MArco
- STartCCM -- ESU thermal simulation
- ISSUES
-
- Most annoying issue
- water tight geometry: no gaps in the model, contact in the model
- connecting stuff from CAD
- official cad has tolerances by design, those must be removed
- have to extend, either one volume or another
-
- naming surfaces
- Everything is lost, when importing from Step or stl
- MESHING TIME
- Depends on what you are meshing
URLs Found¶
- https://chatgpt.com/c/68d11b92-4fd4-832a-bb58-42235673c2f3
Next Curation Action¶
Promote this into a person page, link it to the relevant company/account, and create/update the EspoCRM contact if this relationship is still active.
Sources¶
C:\Users\adeel\OneDrive\100_Knowledge\203_TextCAD\01_Product_Project_Management\TextCAD_Wiki\inbox\onenoteexport\01_PROJECTS\Gauss Compute\Survey and research - Marco.docxC:\Users\adeel\OneDrive\100_Knowledge\203_TextCAD\01_Product_Project_Management\TextCAD_Wiki\inbox\onenoteexport\01_PROJECTS\Gauss Compute\Survey and research - Marco.mht