Font Size: a A A

Study On Theory And Methodology Of Qualitative Compositional Modeling And Simulation

Posted on:2004-04-06Degree:DoctorType:Dissertation
Country:ChinaCandidate:D F WangFull Text:PDF
GTID:1118360092498865Subject:Management Science and Engineering
Abstract/Summary:PDF Full Text Request
This dissertation studies the model construction of qualitative simulation/qualitative reasoning. Qualitative simulation has made fruitful progress both about qualitative description and reasoning techniques since 1980s, and developed applications research among system analysis, diagnosis, conceptual design, qualitative predication, etc,. However, as a new simulation technique, qualitative simulation has encountered some difficulties during its history. The difficulty of formulating qualitative models has blocked the general application of qualitative simulation. Hence, much effort has been spent on developing technique for automated or semi-automated the qualitative model construction process, which becomes a hot-spot of qualitative physics.In this paper, I study the characteristics and modeling process of qualitative simulation, propose the conception of qualitative compositional modeling and realize the corresponding algorithms. From the viewpoint of novel mechanism for model knowledge description, I develop an object-oriented compositional modeling frame; Base on controlling information during modeling, 1 present the relevance-assumption-based algorithms. As to model evaluation, 1 consider qualitative simulation a co-related step as well as model composition during entire qualitative compositional modeling, and implement two simulation algorithms. The main result of this thesis is as follows:First, because compositional modeling resembles the characteristics of object oriented simulation, such as knowledge representation, object depiction and modeling algorithms, I propose the conception of object-oriented (OO) qualitative compositional modeling. By introducing OO method through compositional modeling, I build the formal representation mechanism of object-oriented qualitative compositional modeling, and develop the OO description of qualitative model units. This is one of the main characters of the dissertation.Next, I study the process control of compositional modeling (CM) and a kind of management information about model fragment use, modeling assumption. In this paper, I provide a structural model about model fragment library, and propose hierarchal representation mechanism about model assumption. Base on the utility of modeling assumption during qualitative model construction, I build the mapping mechanism from modeling assumption to modeling decision. As a result of modeling assumption study, I discuss a special kind of modeling assumption, relevance assumption, which defines the relation between model fragments and objects being modeled, and maps concrete model goal to model fragments. I consider CM in the view of relevance reasoning and develop a qualitative modeling algorithm, in which each model fragment is being related to some kinds of modeling goal and task. This algorithm helps to select or compose model fragments partially automated.Finally, this dissertation implements the simulation algorithms of qualitative compositional modeling. Due to the incompleteness of model library and uncertainty of modeling knowledge, qualitative compositional modeling is a process of "Generate and Test" during which the composed model must be modified and improved. There are two separate sub-tasks about CM, model composition and model simulation, which communicate with each other through qualitative differential equations (QDE). Thus, I build the framework of qualitative compositional modeling, which includes both modeling and simulation processes, and accomplish two kinds of simulation algorithm: one is the pure qualitative simulation algorithms (QSIM) and the other is fuzzy qualitative simulation (FQSIM). I regard QSIM as a kind of constraint satisfaction problem (CSP), and improve the qualitative simulation algorithm by constraint logic programming (CLP). Fuzzy qualitative simulation develops conventional QSIM on several facets, such as representation of qualitative value, state transition rule and filtering algorithms. It bridges the gap between pure qualitative and quantitative, and improve the efficien...
Keywords/Search Tags:qualitative simulation, compositional modeling, object-oriented, model fragment, modeling assumption, fuzzy sets, algorithm
PDF Full Text Request
Related items