Knowledge-based signal analysis and case-based condition monitoring of a machine tool

This paper describes an innovative methodology for knowledge-based signal analysis and interpretation. The force sensors introduced in a rotative machine tool are used for the monitoring and supervision of the functional state of the tool and the possible faults in the environment, such as lubrication and correct speed. The monitoring turning processes get the torque data from the force measurement directly for the main spindles and feed axes given by piezo-electric sensors. They are analysed keeping in mind the knowledge of the experts about the semantically weighted patterns in order to monitor the operating conditions of the machine tool. The proposed system consists of two parts: ABSALON (ABStraction AnaLysis ON-line); and LAMDA (Learning Algorithm for Multivariable Data Analysis). The first one is an abstraction or transformation of the signal into features to be interpreted. This is done by the sliding windows methodology. The second is performed by a fuzzy classifier, including a learning procedure from cases defined by expert knowledge. ABSALON transforms the raw signal into a vector whose components are meaningful features to be classified by LAMDA. Industrial experimental results are shown in the paper