A Hybrid Algorithm To Recognize Machining Features Based On Geometric Reasoning And Neural Network

The independent operation of CAD and CAPP system leads to the incompatibility between design information and manufacturing information,which increases the product development cycle and reduces the competitiveness of enterprises.And feature recognition technology is one of the effective ways to realize the communication between design information and manufacturing information.Currently,machining feature recognition algorithms and software systems have low recognition rates for complex machining features,such as intersecting features and composite features.This thesis studies the technical problems existing in feature recognition technologies,and integrating the advantages of geometric reasoning technology and neural network technology,a hybrid algorithm to recognize machining features based on geometric reasoning and neural network is proposed,which provides a new idea for the research of feature recognition technology.And the main research work of this thesis is as follows:(1)This thesis analyzes the development trend of feature recognition technology at home and abroad,and studies the expression of attribute adjacency graph on the part model based on geometric reasoning in feature recognition technology.Currently,aiming at the lack of expression information of part model in attribute adjacency graph,an algorithm to construct attribute adjacency graph is proposed.It enriches the expression of the part model information in attribute adjacency graph to add attributes of feature faces for nodes in attribute adjacency graph.(2)This thesis introduces the classification of machining features.Aiming at the problem that traditional machining feature recognition algorithm is effective for recognizing typical machining features but difficult to recognize intersecting features and complex features,a machining feature recognition algorithm based on geometric reasoning is proposed.The algorithm is based on the feature face to decompose attribute adjacency graph,realizing the separation of the machining features.Meanwhile,the feature segmentation caused by the feature intersection is processed.The relationship between the geometrical topological information of the machining features is analyzed,and the machining feature recognition rules are summarized for realizing to recognize typical machining features,the intersection features and the complex features.(3)In order to solve the difficulty of using geometrical reasoning technology to recognize machining features of geometric parameters and topological information variation,considering the abstraction and qualitative judgment of neural network technology for geometric dimensionality reduction information recognition,a machining feature recognition algorithm based on neural network is proposed.The algorithm is based on the dimensionality reduction algorithm,realizing the mapping from 3D solid space to low dimensional space.The interest views are extracted based on the degree of attention,and the transformation of the machining feature from topological information to the image feature information is realized.The machining feature classifier is used to realize the recognition of the machining features of geometrical variable parameter but the topology being the same,and the machining features of geometrical but the topology variable parameter being the same.(4)Combining the advantages of high accuracy of geometric reasoning technology to geometric matching and low sensitivity of neural network technology to topological variation,a hybrid algorithm to recognize machining features based on geometric reasoning and neural network is proposed,and the UG-NX and the Tensorflow development technology are used to realize the algorithm.Constructing the prototype system for feature recognition,the feasibility and effectiveness of the algorithm are verified by examples.By studying the hybrid algorithm to recognize machining features based on geometric reasoning and neural network,the effectively recognition of complex machining features such as intersecting features and composite features is realized.And the effectiveness of the machining feature recognition algorithm are verified by examples of automatic dimensioning of 3D part models.