Fix PINN Code for Pressure Prediction
Budget: $30 – $250 USD
[URGENT - 48 HOURS] Fix PINN Code to Predict Pressure Field at High Eccentricity in Reynolds-Averaged Equations
⸻
✅ Project Description:
We are looking for an experienced developer or researcher to modify an existing Physics-Informed Neural Network (PINN) code to correctly predict the pressure field under high eccentricity conditions (e.g., e = 0.9) for Reynolds-Averaged equations.
⚠️ The current code works well for low eccentricity (e = 0.3), but fails for higher eccentricity (e = 0.9). Specifically, it underestimates the peak pressure — prediction is stuck around 8, while the reference solution (from finite difference method in MATLAB) is much higher.
The following attempts have already failed:
• Increasing sampling points near the pressure gradient;
• Deepening/widening the network;
• Trying other activation functions (Tanh, Sigmoid, SiLU);
• Clipping negative pressure;
• Training progressively from low to high eccentricity.
Now we need you to solve this properly by redesigning the network/loss/sampling to achieve correct predictions at high eccentricity.
⸻
✅ What You Need to Do:
1. Modify and improve the provided PINN code (Python + PyTorch);
2. Adjust the network/loss/sampling to improve peak pressure prediction under high eccentricity;
3. Ensure output pressure fields are physically reasonable and close to the reference;
4. Provide the following deliverables:
⸻
✅ Deliverables:
• ✅ Modified, working main.py code;
• ✅ Prediction plots at high eccentricity (e.g., e = 0.9);
• ✅ Video recording showing your model running (training + output generation);
• ✅ (Optional but preferred) Short explanation of what was changed and why.
⸻
✅ We Will Provide:
• The original PINN code (working for low eccentricity);
• A PDF description of the physical model and current problem;
• An explanation of failed attempts and what needs to be improved;
• Any clarification you need during the job.
⸻
✅ Deadline:
Strict: 48 hours after accepting the project.
Only apply if you are confident and available immediately.
⸻
✅ Requirements:
• Experience with PINNs or scientific ML (especially PDE modeling);
• Ability to work under deadline;
• Able to explain your changes clearly if requested.
⸻
✅ Project Description:
We are looking for an experienced developer or researcher to modify an existing Physics-Informed Neural Network (PINN) code to correctly predict the pressure field under high eccentricity conditions (e.g., e = 0.9) for Reynolds-Averaged equations.
⚠️ The current code works well for low eccentricity (e = 0.3), but fails for higher eccentricity (e = 0.9). Specifically, it underestimates the peak pressure — prediction is stuck around 8, while the reference solution (from finite difference method in MATLAB) is much higher.
The following attempts have already failed:
• Increasing sampling points near the pressure gradient;
• Deepening/widening the network;
• Trying other activation functions (Tanh, Sigmoid, SiLU);
• Clipping negative pressure;
• Training progressively from low to high eccentricity.
Now we need you to solve this properly by redesigning the network/loss/sampling to achieve correct predictions at high eccentricity.
⸻
✅ What You Need to Do:
1. Modify and improve the provided PINN code (Python + PyTorch);
2. Adjust the network/loss/sampling to improve peak pressure prediction under high eccentricity;
3. Ensure output pressure fields are physically reasonable and close to the reference;
4. Provide the following deliverables:
⸻
✅ Deliverables:
• ✅ Modified, working main.py code;
• ✅ Prediction plots at high eccentricity (e.g., e = 0.9);
• ✅ Video recording showing your model running (training + output generation);
• ✅ (Optional but preferred) Short explanation of what was changed and why.
⸻
✅ We Will Provide:
• The original PINN code (working for low eccentricity);
• A PDF description of the physical model and current problem;
• An explanation of failed attempts and what needs to be improved;
• Any clarification you need during the job.
⸻
✅ Deadline:
Strict: 48 hours after accepting the project.
Only apply if you are confident and available immediately.
⸻
✅ Requirements:
• Experience with PINNs or scientific ML (especially PDE modeling);
• Ability to work under deadline;
• Able to explain your changes clearly if requested.