Research On Milling Surface Roughness Detection Method Based On Feature Extraction And Convolutional Neural Network

Surface roughness is one of the important indicators to measure the quality of processed surfaces in the mechanical field.How to improve its detection speed and accuracy in engineering applications has always been a research hotspot.With the advancement of technologies,the detection accuracy of roughness is further improved.In the actual production application,the qualified or not of the product needs to reach the range of technical requirements so the detection of the surface quality of the workpiece requires higher speed.In this context,the surface roughness detection based on the surface image of the workpiece can fully meet the requirements of the actual processing line.According to the national standard,the common value range of surface roughness is classified.The purpose of this paper is to obtain the milling surface of the workpiece through experiments and build surface roughness classification prediction model by image feature extraction and convolutional neural networks.It has greatly improved the detection speed under the premise of meeting the accuracy requirements.According to the actual texture state of the machined milling surface,the surface roughness classification detection method based on image feature extraction adopt the Gray Level Cooccurrence Matrix to perform surface texture feature detection,selects 6-dimensional eigenvalues as image feature vectors and establishes back propagation neural network to classify eigenvectors.Using training set data can achieve 98.96% accuracy,using test set data can reach 91.67% accuracy.The surface roughness classification detection method based on convolutional neural network is an end-to-end image analysis method.The surface image is directly put into the model built by the CNNs.Through multi-level operation and comprehensive processing,the final output is the surface roughness classification prediction.The training set data can achieve 100% accuracy,which can be achieved by using the test set data.99.34% accuracy.The above two types of model have better classification accuracy on image-based surface roughness detection and detection speed is faster,which has practical value in industrial field.