This application demonstrates the ability of a neural network to accurately predict future values of a cantilever beam tip position knowing only past and present values of the tip position. Most neural networks are trained based on a relationship between input and output pairs. However, in this application the actual system input is unknown. The neural network uses only past tip position to predict the position of the cantilever beam tip for the next five samples. The output of the neural net structure is then used to calculate predictions of the velocity and acceleration. The predicted position, velocities, and acceleration are then used to calculate the voltage needed to excite piezo-ceramic actuators in order to bring the beam back to its steady state. The controller brings the beam back to a resting position when subjected to impulse and a random input to the beam significantly faster than when the beam has no control.