复杂不确定环境下基于模糊Petri网的知识表示与推理方法研究

With the rapid development of information technology, expert system has become bigger and bigger and its complexity and uncertainty have also been increased greatly. The two most important issues of expert systems are the acquisition of domain experts’ professional knowledge and the representation and reasoning of the identified knowledge rules. As a promising tool for knowledge representation and reasoning, fuzzy Petri nets (FPNs) still suffer a couple of deficiencies in constructing complicated knowledge-based systems. First, they are not effective and intuitive enough to allow knowledge engineers to acquire or extract different experience and knowledge from domain experts. Second, the parameters in current FPN models could not accurately represent complicated rule-based expert systems. Third, fuzzy reasoning algorithms of many FPNs are not suitable for concurrent reasoning and the rules in their models cannot be adjusted dynamically according to varying antecedent propositions. Therefore, in this project, we will propose a robust knowledge acquisition method by using linguistic evidential reasoning and hybrid averaging operator, develop hybrid linguistic FPNs based on cloud model theory and by introducing other knowledge parameters, such as global weight and time factor, and present a bidirectional concurrent reasoning algorithm combing Muirhead mean operator and max algebra operations. Finally, based on the above researches, an integrated knowledge representation and reasoning system will be developed, and empirical case studies will be applied to demonstrate the effectiveness and advantages of the proposed model and algorithm. This project will benefit to enrich the artificial intelligence (AI) theoretical system and promote the development of AI technologies and intelligent expert systems.

随着信息技术的迅速发展，专家系统规模越来越大，同时复杂性和不确定性也日益提高。而传统模糊Petri网的知识表示及推理机制在复杂知识系统构建中存在很多不足，不能很好地满足实际应用的需要。具体表现在：不能获取多样和不确定的专家知识、不能完全表征复杂系统知识、模糊推理效率低且不具有动态自适应能力。因此，本课题在国内外模糊Petri网理论及其应用研究的基础上，首先运用语言证据推理方法和混合平均算子，提出更加可靠的知识信息获取策略；其次应用云模型理论同时引入全局权重、时间因子等知识参数提出更加一般化的混合语言推理Petri网模型；然后基于Muirhead平均算子和极大代数运算提出双向并行的知识推理算法。最后，在上述研究的基础上开发出知识表示与推理原型系统，并通过实例验证所提出模型的有效性。本项目的研究成果及相关理论和方法的提出将有助于丰富人工智能理论体系，对人工智能技术的发展具有重要的推动作用。

随着新一代信息技术的快速发展，专家系统作为人工智能应用研究的核心领域其智能化程度及准确性受到了企业界和研究者广泛关注。本项目针对传统模糊Petri网模型的不足与专家知识的复杂性与不确定性，主要开展了以下研究：(1)通过研究专家知识的多样性和不确定性，提出了能够获得更加可靠、全面知识规则参数信息的策略；(2)通过研究知识表示的复杂性和系统性，构建了基于云模型的语言产生式规则，提出了更加一般化的混合语言推理Petri网模型；(3)设计了基于矩阵运算的双向并行推理算法，减小了系统推理结果与实际输出的偏差，提高了模糊Petri模型的泛化能力和推理效率；(4)最后开发了基于混合语言推理Petri网的知识表示与推理原型系统，并通过医院实际诊疗案例进行了实证验证，有效提高了医院的医疗水平和竞争力。本项目完成了预定的研究内容，取得了超过预期的研究成果。在IEEE TKDE等国际知名学术期刊上累计发表SCI论文61篇，其中IEEE汇刊论文12篇、中科院1区论文19篇，ESI高被引论文10篇；在国内重要期刊上发表论文5篇，Springer出版英文专著2部。研究成果在多个医疗机构进行实证研究及推广应用并取得显著成效，获得上海市自然科学奖二等奖1项，上海市科技进步奖三等奖1项，申请国家发明专利5项。项目执行期间，培养博士和硕士研究生12名，邀请20余名国内外学者来校合作交流，项目组成员出国合作交流3人次。此外，通过本项目的研究积累，项目负责人晋升为IEEE学会高级会员，入选全球前2%顶尖科学家（2021年）、全球高被引科学家（2020年）、连续三年入选中国高被引学者。