SIA OpenIR  > 自动化系统研究室
面向复杂制造过程质量的模糊Petri网预测及诊断方法研究
Alternative TitlePrediction and Diagnosis of Fuzzy Petri Nets Oriented to Quality Issues of Complex Manufacturing Processes
袁杰1,2
Department自动化系统研究室
Thesis Advisor史海波
ClassificationTH165
Keyword模糊petri网(Fpn) 质量 预测 诊断 模糊着色petri网(Fcpn) 模糊产生式规则(Fpr)
Call NumberTH165/Y87/2009
Pages108页
Degree Discipline机械电子工程
Degree Name博士
2009-03-25
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract在复杂制造过程中,存在质量异常预测及诊断能力弱、智能化程度低、效率低等问题。如何针对制造过程质量问题特点采用合适的预测与诊断方法,满足日益提高的过程自动化水平的要求,是该领域研究人员面临重要的亟待解决的问题。由于模糊Petri网是模糊集理论与Petri网理论有机结合的一种网络理论,其突出优势在于知识表示、推理和处理模糊信息的能力; 目前,尽管模糊Petri网已有成功的应用案例,但仍存在某些不足,需不断地改进与完善。因此,对模糊Petri网理论方法的研究,有利于提高其知识表示能力、模糊动态推理能力、推理结果可靠性与准确性等,对模糊Petri网理论的广泛应用具有推动作用。本文以制造过程质量问题的预测与诊断为研究和应用背景,对模糊Petri网预测与诊断方法的研究为主线,以研发的系统为辅助分析工具,重点从方法的层面上对模糊Petri网理论进行了研究和探讨。旨在进一步完善模糊Petri网相关理论,并应用于制造过程质量问题的解决,提高过程的质量监控能力、事故预防能力、缩短故障原因查找周期、提高定位准确性及可靠性奠定方法基础。针对制造过程质量预测与诊断问题特点,在广泛阅读相关文献并深入探索的基础上,对模糊Petri网理论方法进行了较深入的研究和探讨,重点解决了以下问题: 1)模糊Petri网自动建模方法:对模糊Petri网理论研究的基础和前提是建立模糊Petri网模型。为解决当前模糊Petri网建模效率低、工作量大、易出错等问题,本文提出了模糊Petri网的自动建模方法。该方法的提出,易于保证知识库与模型库更新的同步和一致,提高了建模效率,避免了建模的人为失误。 2) 模糊Petri网参数确定:模型建立后,为实现可靠有效地推理,需进行相关参数的确定。提出了确定模糊Petri网的初始库所token的方法。通过模糊统计的方法来获得模糊token,减少确定token时的主观臆断性和不一致性,为物理量与模糊token的实时转换提供了技术支持。由于构建符合客观实际的、连续的隶属函数是确定模糊token的前提条件,本文提出采用最小二乘拟合来构造模糊隶属函数方法。该方法简单,拟合能力强,人工干预少。由于变迁阈值影响推理的正确性及可靠性,这里对阈值设定进行了初步探讨。阈值设定越高,预测及诊断的漏报率越高;反之,误报率越高。给出了阈值设定的总代价计算式,阈值选择的目标是使总代价最小。在建立了模糊Petri网模型、确定了相关参数后,便可对异常事件进行预测及诊断推理。 3)模糊Petri网预测方法:对预测模式进行了分类与定义,便于对不同模式下进行预测分析。提出了改进的FPN四种基本推理模型,通过禁止弧的引入,避免了激发过的变迁反复被激发,减少不必要的计算,实现了推理与模型结构的一致性。从而提高了推理效率和基于规则系统的响应能力。 4)模糊Petri网诊断方法:给出了一种模糊Petri网诊断推理方法。该方法充分利用模糊Petri网自身的结构与数学特性这一突出优势,实现了并行推理。以矢量计算方式获得中间库所能力,取代了常规的搜索方式,提高了推理效率。通过引入人机交互的处理策略,减少了模糊Petri网的复杂性及规模。指出在实践中,推理方法的效率、成本及实际的应用效果, 在重要性方面,要远大于方法自身的运算效率。 5)模糊着色Petri网推理方法:在建模复杂大型系统时,为解决模糊Petri网存在模型空间过大,模型数据结构松散等问题,提出了FCPN并行推理方法及FPR与FCPN模型转换算法。提出的FCPN与现有方法的主要区别在以下方面:首先,算法实现变迁的单次激发,避免推理激发变迁的重复计算。其次,某个使能变迁前集库所中token在该变迁激发后并不移除,符合实际推理情况。此外,通过输入/输出关联矩阵计算迭代,实现了并行推理。最后,以一典型制造过程—埋弧自动焊接过程质量问题的预测和诊断为例,来说明模糊Petri网方法的实际应用。通过系统的实现,验证了相应方法是可行的。通过模糊Petri网的预测及诊断推理,便于实现质量异常的分析、预警、处理、过程控制及数字化管理,为生产策略的调整、纠正措施的采取提供了决策依据,加快了系统响应速度。本文研究工作重点围绕模糊Petri网理论方法展开,虽以制造过程质量问题的预测与诊断为研究和应用背景,但并不局限于该领域,是属于具有一般性的共性方法。因此,所开展的方法研究工作具有良好的科研价值和广泛的应用前景。
Other AbstractIn complex manufacturing processes, many quality issues are suffered from weak prediction and diagnosis capacities, low intelligence levels, and low inference efficiency. It is an urgent task to be solved for the researchers of this area on how to select appropriate prediction and diagnosis approaches focusing on the issue characteristics of manufacturing processes, meet the requirements of increasingly automatic levels. Fuzzy Petri nets (FPN) is a net theory combined fuzzy set theory and Petri net theory. The prominent advantages of FPN are its knowledge representation, inference, and fuzzy information processing. Currently, although there are some successful applied cases using FPN, some shortcomings are still existent, which need to be improved and perfected. Therefore, the study on FPN theory is beneficial to improve knowledge representation, enhance dynamic fuzzy reasoning, promote reliability and accuracy of inference, and facilitate its extensive applications of FPN theory. This research and its application background of this dissertation are the quality issues of prediction and diagnosis for complex manufacturing processes. This study focuses on prediction and diagnosis methods of FPN, and takes the developed FPN inference software as the assistant analytic tool.The purpose of this study is to further improve the fuzzy Petri net theory, resolve the quality problems of manufacturing processes, improve quality control capacities for manufacturing production processes, accident prevention capacity and shorten the fault-searching span, and lay a good foundation for improving positioning accuracy and reliability. According to the characters of quality issues for prediction and diagnosis in manufacturing processes, the authors looked up much relative literature and further carry out this study on the FPN theory. This paper focuses on resolving the following key issues: 1) Fuzzy Petri net automatic modeling methods: Modeling FPN is the basic premise of this study. In order to solve the issues of low modeling efficiency, much workload, and man-made errors while modeling fuzzy Petri nets,this paper presents an automatic modeling method of FPN. The proposed method is easy to guarantee the synchronization and consistency of the knowledge base and model base while they updating, improve the modeling efficiency, and avoid human modeling errors. 2) Determine the parameters of FPN: After having built models, the need for the subsequent reasoning study, the required work is determining the relevant parameters of the FPN model. A key parameter of FPN is the fuzzy tokens of initial places. This paper proposes that the fuzzy tokens are determined by fuzzy statistics. This approach reduces the subjectivity and inconsistency, and provides the technical support for the real-time transformation of physical papameters and fuzzy tokens. Because it is a precondition for determining fuzzy tokens to construct practical and consecutive membership functions, this study employs Least Squares to construct membership functions. This approach is of simplicity, powerful fitting, and low human intervention. As far as threshold setting of transitions are concerned, the higher their values are, the higher the missing report rate is; conversely, the higher the false alarm rate is. This study gives the formula of total cost for computing threshold values, and the threshold selection goal is to minimize the total cost. After having established the FPN model as well as determined the relevant parameters, it is feasible to predict and diagnose abnormal events using FPN. 3) Prediction using FPN: Definitions of prediction modes give a better and detailed understanding for prediction, and provide conditions for the prediction mechanism study. This paper proposes four improved basic inference models, to which inhibitor arcs are introduced. The models avoid the defect that fired transitions are fired repeatedly while reasoning, reducing unnecessary computation, realizing the consistency between the reasoning mechanism and its model structures. This approach also improves reasoning efficiency, and increases the response speed of a rule-based system. 4) Diagnosis using FPN: This paper presents a diagnostic reasoning approach using FPN. It takes full advantage of the structural and behavioral properties of a FPN. It can identify middle places by a vector-computational manner rather than the conventional search way, improving inference efficiency. To reduce the complexity and scale of a FPN, man-machine interaction is introduced to it. It is suggested that high efficiency and low costs which an inference method brings in practice, play a more important role than the operational efficiency of the method itself. 5) Fuzzy Colored Petri nets (FCPN) reasoning: To solve the over large model space, loose data structures of net models while modeling a complex and large system using FPN, this study put forward a concurrent reasoning using FCPN. The major differences between the proposed FCPN and the existing ones are the following three ways. Firstly, reasoned transitions can be fired for only once during a reasoning process, reducing much computation. Secondly, the tokens of the antecedent places of a transition are not removed while reasoning, satisfying the real-world reasoning cases. Lastly, this paper takes a typical manufacturing process, i.e., submerged welding process as an example to illustrate the practical application of partial aforementioned methods. The approaches are proofed to be feasible through the software system developed by ourselves, and have a good theoretical research value and application prospects. Through the prediction and diagnostic reasoning using FPN, it is easy to realize the quality abnormity analysis, early warning, treatment and digital realization. They provide a basis for taking corrective measures timely, and adjusting the production strategies. This study focuses on fuzzy petri net theory. Though the research and application background are oriented to the quality issues of manufacturing processes, the research is not limited to this area, and it is a common and universal approach. For this reason, this study has great research value and extensive applied prospect.
Language中文
Contribution Rank1
Document Type学位论文
Identifierhttp://ir.sia.cn/handle/173321/494
Collection自动化系统研究室
Affiliation1.中国科学院沈阳自动化研究所
2.中国科学院研究生院
Recommended Citation
GB/T 7714
袁杰. 面向复杂制造过程质量的模糊Petri网预测及诊断方法研究[D]. 沈阳. 中国科学院沈阳自动化研究所,2009.
Files in This Item:
File Name/Size DocType Version Access License
10001_20061801471901(4277KB) 开放获取--Application Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[袁杰]'s Articles
Baidu academic
Similar articles in Baidu academic
[袁杰]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[袁杰]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.