- Prepare Data
- Get airfoil data files.
- Import data into a numpy array and plot airfoils using matplotlib. Check coordinate ordering.
- Generate new airfoil from a given airfoil using random noise.
- Work on high, medium and low variance airfoil generation. Prepare directory structure.
- Work on getting Xfoil working.
- Prepare the data generation pipeline.
- Use Xfoil to generate only valid airfoils.
- Find and save the L/D ratio for generated airfoils.
- Generate H, M, L variance airfoils. Look at plots to decide on noise levels.
- Estimate time to generate a H, M, L data point on average.
- See if data can be generated through multi-processing in parallel.
- Decide H, M, L count depending on time taken and given that I am willing to spend D days on just data generation.
- Need for using parametric representation and smooth surface airfoil generation after adding noise.
- TE and LE at (1, 0) and (0, 0) and setting angle of attack of generated airfoils to 0.
- Work on neural network training pipeline.
- Get data from dataset using Dataset and Dataloader in pytorch.
- Define Neural Network class.
- Define the hyperparameters.
- Complete the training loop.
- Work on checkpoints saving.
- Work on input optimization given the neural network.
- Fix L/D
- Change Loss function.
- Work on getting the entire neural network pipeline working.
- Move code to GCP. Connect to GCP via VSCode and SSH.
- Make sure all code works on GCP.
- Generate actual data.
- Decide where to store the data.
- Train the neural network.
- Run the two optimization processes.