Abstract
By grinding process acoustic emission is produced. It was used for characterising the system grinding wheel - grinding machine - workpeace. Frequency distribution till 12 kHz was measured by grinding process, roughness was measured too, by touch instrument. Three attributes were extracted from frequency distribution: total spectrum area (SpArea-P), frequency of the maximum area peak (MaxArea-fm), frequency of the spectrum area central point (AreaCX-fc). Obtained data is base of expert system made by computer program FORS. Methods of artificial intelligence - automatic learning of continuous class in forst order formalism was used for the first time. Expert system has rules of the form IF-THEN-ELSE. It is a base for control computer program. Surface roughness changing is obtained by input attributes (SpArea, MaxAreaX, AreaCX). This rules can help in deeper understanding of grinding mechanism. By this system we can predict surface roughness directly with measuring of acoustic emission. With additional work (appropriate filters, feed back loop) we obtain a system for "on-line" controlling of grinding system.