Nonlinear System Identification using ANN

Nonlinear system identification using artificial neural network example

This example file (Example.m) performs system identification using artificial neural network (ANN) on the nonlinear dynamic behaviour of a two dimensional frame. The output of the two dimensional frame was generated using Two Dimensional Frame Nonlinear Dynamic Solver:

https://www.mathworks.com/matlabcentral/fileexchange/111825-two-dimensional-frame-nonlinear-dynamic-solver

The fourteen necessary files of that solver are already uploaded in this submission.

The neural network consist of the following layers:

-Input layer: 1 node for the force at the current step, 4 nodes for the displacement at the previous four steps, and 4 nodes for the base shear at the previous four steps using open-loop feedback.

-Hidden layer: 4 nodes for four inner states.

-Output layer: 1 node for the displacement and 1 node for the base shear.

After training and getting the predicted output, the network was converted to closed-loop network and trained again (closed-loop networks uses predicted feedback from previous step instead of actual feedback). The predicted output from open-loop and closed-loop networks was compared with the actual output in two figures.