Research On Semantic Integration Technology For Heterogeneous Product Models Across Multiple Domains

In collaborative product development, product model is the basis for carrying out acitivities such as design optimization, performance analysis and simuliation. How to integrate the product models from different development phases efficiently and realize sharing and reuse product information fully are the fundamental issues needed to be addressed. With the increasing importance of knowledge in product lifecycle, a common agreement about product information sharing and integration is achieved that it is no longer merely exchanging geometric data, but more semantics hidden behind product data.The development of complex product often involves multiple disciplines, which needs collaboration between partners coming from a variety of companies and support by various design and analysis software tools. Under such environment, some problems arise when semantic integration of product model across multiple domains. First, for most of product knowledge is impicit and software tools encapsulate the constraints and rules inside, there is few engineering semantic knowledge contained in model so that sharing and reuse is realized at relatively low level. Moreover, domain views the same product information in different manner and each system normally has its specific data types and formats with the aim of solving specific needs. Thus, the information managed in various systems is highly heterogeneous, which results in semantic loss and semantic ambiguity during information translation from one application to another. The degree of collaboration is limited. Consequently, it is necessary to study on semantic integration technology for product models across multiple domains, in order to realize product knowledge sharing, exchange and reuse at semantic or conceptual level.This work is supported by the Major State Basic Research Development Program in China (2002CB312106) and Aerospace Science and Technology Innovation Fund Project. Aiming at some problems of semantic integration for product models such as semantics expression, capturing and mapping, our main works are listed as follows:(1) A framework named CPMSIF to support semantic integration between multiple heterogeneous product models across domains is proposed.The CPMSIF framework is classified into concept layer and model layer. The concept layer is composed of the core product ontology and multiple domain-specific ontologies, which support formal semantic representation of product knowledge. The model layer is composed of multiple domain semantic models which can be extended easily. The organization and definition of product ontology, the construction way and content of core product ontology is analyzed detailly. Based on the product ontology in concept layer, a semantics-driven knowledge sharing and reuse methodology among models is presented, by which the low-efficiency of modeling repeatedly can be avoided, and knowledge reuse based on semantic understanding across various domain contexts can be achieved.(2) A collaboration oriented product semantic model with knowledge represented at different levels of abstraction is proposed.Product semantic model serves as basis for semantic integration. For solving the problems of semantics implication and loss occurring in collaboration, semantic model captures knowledge at multi-abstract levels including design intent and engineering semantics by the way of semantics interpretation and extension based on the existing domain model. Utilizing the concepts in ontology as semantics carrier, the interpretation-based semantics make the hidden semantics of collaborative information explicited and achieve the target of semantic enhancement, while extension-based semantics further complement key characteristics and other high-level knowledge related to semantic interpretation object. With the knowledge represented at different levels of abstraction, the semantic model can help collaborators understand the product knowledge consistently and correctly, and heighten the degree of knowledge resue effectively.(3) A semantics capture mechanism based on meta-model semantics lifting and multi-strategy complementing is proposed.Combining pull and push strategies, product semantics at different abstraction levels is captured. By the way of meta-model analysis, semantic enrich and strengthen, the interpretation-based semantics is automatic extracted from the low-level data in existing domain model. By the way of pre-defined complement and knowledge-based automatic reasoning, the extension-based semantics is abtained. (4) An engineering context-dependent semantic mapping technology is proposed.Aiming at the diversity characteristics of semantics sharing and exchange among application models, we establish three kinds of semantic relationships: semantic equivalence, semantic similarity and semantic connection. Utilizing semantic description at different levels, a semantic mapping mechanism is proposed, including mapping based on core product ontology, mapping based multi-level features of concept description, and mapping based on domain rules. At the same time, the engineering context which reflecting the focus of the project objectives is applied to the computing of concept matching, in order to improve the quality and accuracy of ontology mapping and provide solutions for engineering application.