Classification of the revolving machines defects by a neuronal model

• Published in: 2013 International Conference on Control, Decision and Information Technologies (CoDIT) pp. 899 - 903
• Main Authors: Imed, Boufedj, Azzedine, Bouzaouit, Ouafae, Bennis
• Format: Conference Proceeding
• Language: English
• Published: 01.05.2013
• Subjects: classification; conditional maintenance; Digital signal processing; gradient of error; Levenberg-Marquardt; Maintenance engineering; Monitoring; Neurons; neurons networks; Probes; Training; Vectors; vibratory monitoring
• DOI: 10.1109/CoDIT.2013.6689662

The current evolution of technologies and the techniques of monitoring, analyzes defects led the industrialists to make important progress in the field conditional maintenance, which consists an adequate application with the tools and analyzes powerful, by but the vibratory analysis is that which experiences today the most significant developments because of developments in the technologies in the fields of the data processing and the signal treatment. Our objective of this work consists in proposing and to implement under MATLAB an approach for the development of a system for the monitoring of the machines based on a neurons network optimized compared to the hidden layers numbers, with the numbers of neurons in layers hidden with the functions of activations used, with the type of algorithm of training used like data input in order to feed a neurons network. The coefficients of correlation as MSE are exploited and used like output data of the neurons network of the type MLP by using the algorithm type of retro-propagation of the error gradient with Levenberg-Marquardt.