Research On Crankshaft On-line Recognition Method And Application For CNC Machining

Under the background of "Made in China 2025" and "Industry 4.0" strategy,CNC manufacturing equipment develops in the direction of the intelligent integration.The machine vision detection technology is integrated into the CNC work cycle to realize on-line recognition of the workpieces,which can expand the function of CNC machine tools,improve the working efficiency of CNC machine tools,and significantly reduce the production cost.At the same time,it is helpful to improve the technology of intelligent measurement and control of manufacturing equipment and the level of integrated equipment integration technology,it is also helpful to improve the level of intelligent manufacturing equipment and application.Firstly,in view of the problem of the real time and accuracy of on-line recognition of complex crankshaft workpieces,a fast recognition method for crankshaft based on global features was proposed;The improved FDMA skeletonization algorithm and the improved Hu moment invariant algorithm were adopted to get the skeleton of the crankshaft images and the global features of the skeleton,support vector machine(SVM)was used to realize the fast recognition for crankshaft workpieces.Secondly,a recognition method for crankshaft based on local detail features was proposed,and the improved LeNet-5 model of the convolutional neural network was constructed.The DropOut training algorithm was added to prevent overfitting,and the ReLU function was adopted as the activation function of the network to improve the training speed.By removing the C5 layer of the model,the real-time performance of training and recognition was further improved.Thirdly,in order to verify the theoretical method proposed in this paper,the image experimental works about crankshaft recognition of considering the global features and local features were completed.Experimental results show that the recognition method based on skeleton global features has low computational complexity,fast recognition speed and high recognition rate,suitable for recognizing the workpieces with large difference of structure.The optimal parameters of the improved LeNet-5 network were obtained through experiments,and the improved network has good performance for recognizing the local features of the crankshaft images,the recognition rate can reach 90%,and it has fast training speed,suitable for recognizing the workpieces with similar structure.The hierarchical recognition experiment shows that the method of hierarchical recognition can make full use of the above two methods for the advantages of the global characteristics and local characteristics of the crankshaft.When the crankshaft type is more,it can get higher recognition rate than single method.Finally,the system of CNC crankshaft on-line recognition was constructed,the engineering application of the CNC crankshaft on-line recognition method was completed.The applicability of the proposed methods of CNC automatic recognizing the type of workpieces in industry was proved.