Tool Diagnosis Method of CNC Machine based on Color Space Conversion and Deep Learning

Abstract

Tool diagnosis system is necessary to prevent an accident or defective product. This paper proposes the tool diagnosis method of CNC machines based on color space conversion and deep learning. To apply the deep learning algorithm, we generated images from the current data of CNC machines by wavelet transform. However, generated images by wavelet transform are difficult to distinguish whether it is normal data image or not. Because generated images by wavelet transform are very similar and there is no outstanding feature. Therefore, we applied color space conversion from RGB image to CIE L*a*b* image. Converted images represent outstanding features whereas generated images by wavelet transform and RGB images do not. And, to make up for imbalanced data, oversampling is applied. Finally, deep learning algorithm is trained to classify the converted images. Experimental results showed that the proposed method can implement the deep learning network for tool diagnosis of CNC machine effectively.