支持设计重用的产品三维模型多关联启发适配关键技术研究

Designing 3D models of complex mechanical products is time-consuming and error-prone. Reusing existing models can improve design efficiency and quality. In front of massive 3D models, the premise and key of reuse is to find the model, meeting the design intent. The designer inputs a model instance, hoping to find multiple suitable models, which are associated with the instance differently and can inspire his design thinking. It is a multi-correlative inspiring model matching process. Studies on the key technologies of multi-correlative inspiring model matching can effectively improve the efficiency and effectiveness of reuse. The main contents of this research project include the methods of cross-scale constraints pyramid representation of 3D models, nearest subspace multi-correlative model matching, efficient relevance feedback and joint response of model sequences, and long-term evolution of matching for enhancing inspiring design thinking. Innovations of the research project include the main scale mapping method based on the feature layers-complexity function, the nearest model subspace decomposition based on joint sparse representation, as well as the joint response of feedback based on the cross cascade network clustering. Furthermore, a prototype system of multi-correlative inspiring 3D model matching will be developed, and it will be tested by the design of a horizontal coordinate boring machine. The completion of this project is of important theoretical and practical value for the effective reuse of existing 3D model and design knowledge.

复杂机械产品的三维模型设计十分耗时且易出错，重用能够有效提高设计效率与质量。面对海量三维模型，设计重用的前提与关键是查找符合设计意图的模型。设计者输入模型实例，希望找到合适的、与实例有不同关联的、能启发其设计思维的多个三维模型，其实质是一个模型的多关联启发适配过程。研究三维模型多关联启发适配关键技术，能够有效提高重用的效率与效果。项目主要研究三维模型跨尺度约束塔式表征、三维模型最近邻子空间多关联适配、模型序列高效相关反馈与联合响应，以及设计思维启发增强的长效适配进化等方法。在基于模型特征层-复杂度映射的主尺度配合方法、基于联合稀疏表示的模型最近邻子空间解析方法、以及基于三维模型交叉级联聚类的联合反馈响应技术方面取得创新。并开发产品三维模型多关联启发适配原型系统，以卧式坐标镗床整机及核心部件的设计重用为对象进行应用验证。本项目的完成，对有效利用现有设计模型与设计知识有着重要的理论及实用价值。

面对海量的三维模型数据，如何方便、准确、快速地获取想要的三维模型,是提高设计效率、缩短产品开发周期的关键之一。现有三维模型相似匹配技术往往只能检索到与模型简单相似的重复模型，难以满足设计者的启发性需求。针对这一问题，本项目研究了一种新的三维模型启发适配方法，重点研究了产品三维模型的旋转不变图像表征方法、三维模型最近邻子空间联合匹配方法、模型匹配序列高效反馈与启发增强技术、三维模型匹配检索原型系统开发及验证等内容。通过项目研究，构建了一个完整的基于二维图像的三维模型匹配检索技术架构，有效解决了现有基于内容的模型匹配检索方法存在的三维模型实例难以获取，准确的草图难以绘制的问题。在三维模型的表征方法、三维模型特征的匹配方法、模型序列的反馈及启发增强技术等方面实现了技术创新，解决了现有模型检索技术缺乏启发性的问题。项目还开发了原型系统并进行了验证，结果表明，本项目所研发的技术可以提高三维模型检索效果，具有重要的理论与工业价值。