Reputation: 9
Neural Network Type for Time Series Approximation
I am currently trying to decide which type of neural network I should choose to complete the following task:
First imagine the following problem setup:
We are looking at a 2D airfoil with a prescribed motion pattern (traverse as well as periodic rotational/flapping movement). This means the lift and drag produced by the foil will show some kind of harmonic behaviour related to the flapping movement of the foil.
The neural network should be capable of approximating the behaviour of the lift (or drag) of the foil over one period of flapping motion for airfoil geometries which differ from the training data in selected geometrical parameters. Currently I could envision two approaches to solve this (it is mandatory to use some kind of neural network approach)
1) Use a standard feed-forward neural network with the output parameters being the lift / drag at discrete points of the flapping period. Depending on the complexity of the lift / drag curve many output parameters would be needed. Input parameters are geometrical parameters such as chord length / camber / ...
2) Use a recurrent neural network in order to capture the temporal progression. I am not sure if this is needed or meaningful but most time-series related problems are solved with RNNs.
I would be really grateful for any kind of suggestions regarding the problem!
Best Regards,
Bob
Upvotes: 0
Views: 45
Answers (0)
Related Questions
- What is the best activation function to use for time series prediction
- What kind of Neural Networks can I use for forecasting?
- Deep Learning Neural Networks for Time Series Prediction
- Convolutional neural networks for time-series
- neural-network non linear time series Narx model in python
- In Tensorflow, What kind of neural network should I use?
- LSTM NN: forward propagation
- Time Series Prediction via Neural Networks
- Which activation function should I use for this Neural Network?
- Choosing the correct type of neural network