基于粗糙概念格理论的多尺度模具知识发现及重用方法研究

This research investigates how to discover and reuse multi-scale knowledge, which is able to help engineers avoid design errors and defects. The research will also investigate problems occur during mold production, and analyse how the knowledge and experience benefit product design. Three aspects are included in this research:. Rough Concept Lattice and Ontology are employed to build the knowledge particle and determine the granularity of knowledge. A maximum entropy model combined with Simulate Annealing Algorithm are used to integrate distributed knowledge in order to refine the knowledge for easy reuse；. Rough Concept Lattice is used to simplify features for engineering requirements, which will be automatically converted into a “matrix of requirements”. The granularity corresponding with the potential solution is calculated through fuzzy equivalent recursive search；.Fuzzy Analytic Hierarchy Process (FAHP) and Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) are used to solve a systematic problem which includes weight distribution calculation, preference definition and similarity matching so as to find the wanted knowledge.. Through this research, knowledge scale is optimized through deleting the redundant knowledge. It greatly improves knowledge discoverability and reusability, which helps the engineers to improve design efficiency.

分析模具产品生产所面临的问题，及经验和知识对产品设计的影响。以破解模具设计过程中出现的“设计错误与缺陷的难以避免”为总体目标，研究多尺度知识的发现及重用方法。包括：. 基于粗糙概念格和本体方法构建知识颗粒，确定知识粒度。以最大熵模型为基础，使用模拟退火算法进行多尺度碎片化知识的融合和提炼，提高知识的有效性和易重用性；. 通过粗糙概念格实现需求特征约简，将复杂需求自动转化为“尺度-粒度”空间的知识需求矩阵，通过模糊等价递归搜索存在解空间的粒度范围；. 通过模糊层次分析法（FAHP）和逼近理想解排序法（TOPSIS）方法，解决综合考虑语义支持度、层级松耦合的权重分布、功能需求偏好特征、相似度校验的知识颗粒匹配的系统性问题，实现基于知识颗粒覆盖集的特征规约与匹配，发现有效知识颗粒。. 通过该研究，删减冗余知识，优化知识规模，提高知识的可发现性和可重用性，大大提高设计效率。

项目首先分析了模具设计过程中设计知识的获取与有效利用的必要性和紧迫性，剖析了模具行业将知识作为设计中的核心要素进行合理重用的两个难点。围绕着问题的解决，有针对性地提出了基于知识工程本体方法的领域知识构建和 基于模拟退火算法的模具知识检索以提高知识的有效性和易重用性等关键技术。本项目的特色与创新之处在于：创新地研究了基于本体的可制造性验证和知识重用产品设计框架，以支持设计和制造知识的共享和重用，并为设计人员在设计过程中提供制造工艺约束的建议和反馈。有效解决了因设计人员和制造工程师之间知识上的不对称而造成的效率低下等问题。同时针对传统的基于产品结构的设计知识发现机制在多尺度环境下灵活性不够的问题，创新地提出基于模拟退火算法（SAA）的注塑模具改型知识（IMMK）系统，优化了模具知识在多尺度环境下的灵活发现和重用，从而对蕴含在大量模具设计实例、设计评审、返修记录中的设计知识进行冗余删减，提高了模具知识的可发现性和可重用性，解决工程变更频繁时模具设计人员经验缺乏难以应付的问题，显著提高了模具产品重新设计过程的准确性和效率。