关键词: 神经网络, 曲面推理, 反向工程

Abstract: This paper applied the neural network technology to surface reasoning in reverse engineering and established the neural network computation model. One of the main advantages of reasoning solid surface using neural network is that no knowledge about surface is needed, and the limited measured points on the surface will do sufficiently. The neural network can learn the combinatorial relationship between parametric variables, inference function relation and calculate the unknown coordinates. This paper listed the related reasoning cases, including the elementary analytical surfaces and freeform surfaces, proved the feasibility and efficiency of this approach from theory and practical computing cases. The results show that a neural network is an excellent aided analysis means for surface reasoning in reversing engineering and possesses practical use for a solid surface that is complex, incomplete and partially worn-out or damaged.

Key words: neural networks, surface reasoning, reverse engineering