Multi-sensor integration—An automatic feature selection and state identification methodology for tool wear estimation

Abstract

Listen

Translate article

Artificial systems obtain knowledge from the real world through sensors. The nature of the transducer, the location in space and the acquisition time must be considered to define the perceived state of the environment. Example base techniques are able to learn from the system experience. They are specially adequate for a complex multi-sensor system which lacks a formal knowledge background. Data logging from diverse sensors, signal feature extraction, identification methods and evaluation strategies for the results have been integrated in a unique programming environment. Some critical aspects of empirical methods, such as automatic feature generation and selection or compaison of pattern recognition procedures, are approached from a global point of view. Some results obtained monitoring tool machines for the estimation of tool wear validate the presented work.