Monitoring of grinding burn by AE and vibration signals

Abstract

The grinding process is widely used in surface finishing of steel parts and corresponds to one of the last steps in the manufacturing process. Thus, it's essential to have a reliable monitoring of this process. In grinding of metals, the phenomenon of burn is one of the worst faults to be avoided. Therefore, a monitoring system able to identify this phenomenon would be of great importance for the process. Thus, the aim of this work is the monitoring of burn during the grinding process through an intelligent system that uses acoustic emission (AE) and vibration signals as inputs. Tests were performed on a surface grinding machine, workpiece SAE 1020 and aluminum oxide grinding wheel were used. The acquisition of the vibration signals and AE was done by means of an oscilloscope with a sampling rate of 2MHz. By analyzing the frequency spectra of these signals it was possible to determine the frequency bands that best characterized the phenomenon of burn. These bands were used as inputs to an artificial neural networks capable of classifying the surface condition of the part. The results of this study allowed characterizing the surface of the work piece into three groups: No burn, burn and high surface roughness. The selected neural model has produced good results for classifying the three patterns studied.