An Approach on Predicting a Machine’s Effector Vibrations Based on Motor Vibrations Using a Regression Artificial Neural Network

Vibrations are mechanical movements which can offer valuable information about the state of a machine. Being able to simulate and predict this type of signal is of great importance in fault diagnosis and condition monitoring of a machine since vibrations are considered the most used and most efficient way to detect and diagnose a fault. Since simulating vibrations can be a difficult task given the non-linear equations that need to be solved, machine learning algorithms can offer a great solution to extract information from data and predict an outcome close to reality given an input signal. Obviously, the input signal is processed, normalized, and fetched to the neural network. In this paper, a regression neural network is used for predicting the vibrations of an effector of a one-dimensional system which moves in only one direction. The vibrations are measured in the direction of the effector’s movement while the vibrations of the motor are acquired by the motor’s support.