【作者】 王毅；

【导师】 陈庆新；

【作者基本信息】 广东工业大学 ， 机械电子工程， 2014， 博士

【摘要】 塑件产品的质量主要依赖于注塑模具,由于塑件产品结构多样复杂、一般为单件定制,即使正确应用模具设计制造原则也难以确保一次试模成功,不可避免地需要对模具进行修改(即改模)。而改模涉及模具设计制造过程中的许多环节,要求改模工程师不仅要掌握模具的设计和工艺知识,还包括课本上没有的各种工艺经验性知识。这些经验性知识目前只能通过不断实践,接受教训,逐步积累与提高。为此,许多模具企业已经开始注意收集整理相关的历史记录,以“问题与解决方案”的形式加以文档化,形成电子化的“改模方案”。其主要作用是在改模工程师制订改模方案时能够通过相似问题搜索,查找相似改模案例。这样可以在一定程度上改善改模的效率和效果,但随着案例数量的不断增加,可借鉴的方案越来越多况且有些问题看似相似,解决方案却又完全不同,改模工程师往往望而却步,迷失在成千上万的“案例”之中。走出上述困境的唯一方法是从这些历史方案中归纳出其内在规律,形成真正意义上的“经验性知识”资产。获取上述经验性知识的难点主要在于以下三个方面：首先,改模方案特征属性的不同取值之间存在着复杂的语义关联关系。如何建立面向改模方案的特征属性值之间的复杂关联模型,并从中抽取和归纳出规律,将成为获取改模知识的难点所在。其次,改模经验性知识具有多样性与不确定性,使得归纳总结改模方案中存在的内在规律难以入手。最后,实际应用要求模具知识管理系统必须具备可持续归纳能力。由于新增的改模案例会对原有的数据决策表产生影响,需要不断从当前的改模案例中抽取新的规律,归纳新的知识。因此,如何应用新的改模信息,建立可持续的学习机制,将是又一难点。针对上述应用问题与难点,本文做了如下四个方面的研究工作：(1)基于网状结构的改模知识本体建立和基于SWRL (Semantic Web Rule Language)语言的改模方案表达和现有只考虑利用概念间的上下层次关系描述经验性知识的模型不同,本文不仅考虑概念间的上下层次关系,同时也考虑概念间的复杂的自定义关系,提出了基于描述逻辑的改模经验性知识的表示方法,并借助本体开发工具建立了网状的改模知识本体,从而比较完整地描述了经验性知识概念之间存在的复杂语义关联关系。提出利用语义网络规则语言SWRL构建改模方案,SWRL语言能直接利用OWL文件中的概念(类)和概念之间的关系,实现对原始改模方案的结构化处理,从而解决了改模方案不被机器所理解的问题。(2)基于自定义关系语义相似度算法的构建和“语义坐标”的建立和现有基于层次模型计算语义相似度的方法不同,本文提出了在考虑层次关系和自定义关系的基础上进行语义相似度计算的方法,通过计算把所有概念分为若干大类,从每个大类中筛选出一系列“中心概念”,提出以这些“中心概念”形象地扮演概念的“语义表述坐标”,每条改模方案是各条坐标轴上的语义值向空间某点进行投影映射的结果。这样解决了由于概念的多样性引起每种特征属性的不同取值数量多,当针对这些属性取值进行规则归纳时,将导致大量应用有限的规则出现的问题。从而可以极大地减少所归纳知识中涉及的概念数量,提高规则的简洁程度和适用范围。(3)基于粗糙集的改模方案增量式更新在粗糙集基础上,作者提出了一种规则的增量式获取方法,并首次应用于注塑模改模知识发现。首先对原有的差别矩阵进行改进,设计了基于改进的差别矩阵求核与属性约简增量更新算法；然后针对求属性约简过程中进行析取与合取运算的时候计算量大的问题,引入了分明差别矩阵,简化了属性约简的计算复杂度；最后通过计算规则的精度和覆盖度,并通过设定规则的阈值,对规则进行提取,得到了完备的改模规则集,提高了系统学习的效率(4)增量式更新的改模知识管理系统设计与实现设计了一个增量式改模知识管理系统,实现了改模知识领域本体语义编辑、语义坐标维护、改模方案知识维护、改模方案聚类、基于增量式的改模方案更新等功能。通过上述研究工作,解决了如何表达改模方案特征属性值之间复杂的语义关系,如何获得典型改模方案,如何持续进行增量式更新等问题,为模具制造企业有效地发现改模方案中蕴含的改模知识提供相关理论和方·法。更多还原

【Abstract】 The quality of the plastic parts was depended on the injection mould. The plastic parts with complicated molding surface and structure, most of them were customized, and mould tests of the injection moulds were hard to be successful with the general designing and manufacture principles, so the mould repairs were inevitable in practice. In the process of the mould repair, the mould repair engineer should formulate the reasonable mould repair scheme with the traditional mould designing knowledge, manufacturing knowledge and the experience knowledge which accumulated in the mould repair practicing. The experience knowledge was accumulated and improved by constant practicing and accepting lessons. So many mould enterprises began to collect related historical records which were documented in the form of "problems and solutions" to form electronic mould repair schemes" which was helpful to the engineers find the similar mould repair schemes by searching the similar problem when a new mould repair project was formulated. But with the increasing number of cases, more and more schemes can be referenced and many similar problems have different solutions, so the engineers will be puzzled in lots of cases. The only way to resolve the questions was that the inherent disciplines should be concluded from the history projects and the true " experience knowledge" asset will be obtained. There are three kinds of difficulty to obtain the experience knowledge. Firltly. there are complicated semantic relationships among the different values of the mould repair schemes attributes. How to establish the complex related model for different characteristic parameters of mould repair schemes and how to extract the new are very difficulty. Secondly, the mould repair experience knowledge was variety and uncertainty which led to that the inherent disciplines from the mould repair schemes will be concluded. Lastly, practical application demanded the mould knowledge management system possessed the compacity of inducing rules from the knowledge database constantly. The new mould repair case can affect the original data decision table, and the system should extract the new rules and induce new knowledge from the mould repair cases, so how to use the new mould repair information and establish the sustainable learning mechanism was another difficulty.According to the difficulty, the main studies of the paper had been conducted as following.(1) The establishment of the mould repair knowledge ontology based on the nets structure and the representation of the mould repair scheme based on SWRL (sematic web rule language)By contrast with the model that only using the hierarchy relations among the concepts discribing the experience knowledge, the new mould repair experience knowledge representation model based on description logic was built, and not only the hierarchy relations but also the complex user-defined relations among the concepts were took into account. The nets mould repair knowledge ontology was established. Then the complex semantic relationships among the concepts of the mould repair experience knowledge were described completely. The mould repair schemes were built by SWRL which can use directly the concepts (classes) and the relations between the concepts of OWL, which can carry out that the mould repair schemes can be structured. So the problem that the mould repair schemes can’t be recognized by the computer was solved.(2) The establishment of the algorithm for the similar calculating between the concepts based on the ontology sematic nets model and the construction of the "sematie coordinate". By contrast with the method of similar calculating based on the hierarchy relationships model, a new semantic similarity computation algorithm based on the ontology semantic nets model was built, in which the user-defined relationships among the concepts and the subclass relationships were both involved. A series of "center concepts" which were used to act the "semantic expression coordinate" of the concepts were obtained by the new algorithm. The mould repair scheme was the mapping point from every concept semantic value. It can resolve the problem that lots of rules with limit application were generated when the many different values were concluded. So the number of the concepts for rule induction was reduced by using the method, and the concise degree and the application ranges of the rules were improved.(3) The mould repair scheme incremental updating algorithm based on the rough setA rules incremental updating algorithm based on rough set was built and applied to the repairing knowledge discovery of injection mould.. Firstly, the discernibility matrix was improved, and the incremental updating algorithm for finding the core and attribution reduction based on discernibility matrix were built. Secondly, according to the heavy computation of disjunction and conjunction in the process of attribution reduction, the distinct difference matrix was adopted which can reduced the calculation complexity. Lastly, the complete mould repair rule set was obtained by calculating the rule’s accuracy and the coverage, setting the threshold value, and inducing the rules.(4)The design and implementation of the mould repairing knowledge management system based on incremental updatingThe system was designed to carry out the functions such as semantic editing of the mould repair knowledge domain ontology, maintenance of the semantic coordinate and mould repair schemes knowledge, clustering of the mould repair schemes, incremental updating of the mould repair schemes.Through the above research work, the problems how to express the complex sementic relations among the values of mould repair schemes attributes, how to obtain the typical mould repair scheme, and how to incremental update constantly were solved. The methods and conclusions drawn from the paper will provide an efficient way to discover mould repair knowledge in mould manufacture enterprises.更多还原