MEASUREMENT AND ANALYSIS OF VIBRATIONS USING ARTIFICIAL NEURAL NETWORK

Abstract

Effective automatic control in maintenance scheduling processes depends on properly designed and implemented computerized system for required measurement and analysis. More over as the vibration measurement has always been considered the best tool for understanding the dynamics of any machine or structure, the measurement and analysis of vibration signals have a significant importance. In this thesis, a system for measurement and analysis of-vibrations using artificial Neural Network, is designed. Digital signal processing techniques are first used to convert collected time domain vibration signals into corresponding frequency domain. Then Neural network based software( more specifically, Error. Back Propagation algorithm supported Multilayer Perceptron model) is used to identify these signals by examining their characteristic frequencies. Implementation of Neural Network in program logic and `. the system's computational properties are also discussed. The promising results demonstrated by the example application, show that Neural Network based systems do have a good potential in automatic maintenance or other process control.