I have a beam loaded with random loads, for each timestep I obtain accelerations and maximum moments on the beam.
My prediction should receive a timeseries of accelerometer data and output the maximum moment at the end of that time. What type of NN should I use?
I've found that LSTMs and GRUs predict on a future time step of the input data (here next accelerometer value) but I need to predict the moment. Note that the simulation lasts for more than 0.5 s.
When training the network both accelerometer and moment data is available but when testing only accelerometer timeseries should be provided to the prediction.
The logic is: after 0.5 s of these displacements I should expect a maximum moment of...