AE 센서와 신경회로망을 이용한 NAK80 금형강의 자기연마 가공특성 모니터링

Abstract

ABSTRACT

Development with mold industry, the demand of precision parts was increased considerably and the study on precision process technique is progressed actively. NAK80 has many advantages such as machinability, weldability, stability, surface enhancements, high surface hardeness. And NAK80 mold steel have great application potential in clear lens mold, extremely critical diamond finish applications, molds requiring special EDM finish etc. many mold fields. Magnetic abrasive polishing is one of the effective polishing method, it is possible to polishing without surface damage because of that has flexible tool. Parameters were considered such as spindle speed, working gap and current of inductor. For grasp of magnetic abrasive polishing characteristics, monitoring technique is very suitable by AE sensor that has high sensitivity. In this study, the magnetic abrasive polishing experiments with on AE sensor attachment were performed basic research. The acquired signal data from the acoustic emission sensor were analyzed to predict an improved surface roughness of a polished surface in magnetic abrasive polishing process for NAK80 material. A dimensionless coefficient(), which consisted of average AErms signal and standard deviation was defined as a characteristic of the magnetic abrasive polishing and a prediction equations were obtained using linear least square method and second order polynomial least square method. As the analysis of these two prediction equations, linear least square method was more suitable than second order polynomial least square method according to error rate and coefficient of determination. And other analysis of magnetic abrasive polishing was selected neural network. This method will find the minimum number of convergence according to input function such as process parameters, dimensionless coefficient and process parameters with dimensionless coefficient. The learning method was selected back-propagation algorithm. As a result of this study, it was seen that there was very close correlation between the AE signal and the improved surface roughness of magnetic abrasive polishing. And then on-line prediction of the improved surface roughness in a magnetic abrasive polishing part was possible to use the defined dimensionless coefficient.