LAUSR

Abstract This article presents a method of processing and analyzing the measurement signals used in the problem of diagnosing the state of a manipulator's tool. The analysis of the signals was performed within the domain of time and frequency. The signals utilized in the analysis were the mechanical vibrations and the rotation speed of the tool. The database for analysis was obtained in a research environment and it includes the instances of the functioning of the system with tool in good technical state as well as instances with a damaged tool. With the intent at reducing the data, the registered signals are represented with the use of selected features. The preliminary selection of the significant features of the signals is made with the use of the sequential feature selection procedure. The reduced set of features is used for the creation of a tool condition classifier, which has a form of an artificial neural network. The obtained classifier operates on-line on robotized system and generates diagnostic information on the state of the tool.