Font Size: a A A

Evolutionary Multi-objective Fuzzy Modeling And Plug Door Fuzzy Fmeca Analysis Research

Posted on:2014-01-09Degree:MasterType:Thesis
Country:ChinaCandidate:H Y JiFull Text:PDF
GTID:2248330395982527Subject:Mechanical and electrical engineering
Abstract/Summary:PDF Full Text Request
Since the birth of fuzzy theory, it has been widely used in data mining, prediction and control, classification and so on. Fuzzy modeling is an important research direction in fuzzy theory and application, and it is divided into two kinds:approximate and descriptive. The former is focused on precision while the latter pays more attention to interpretability. How to obtain better trade-off between accuracy and interpretability is an important research subject in fuzzy systems, and it is the main research content of this paper. Besides, considering from fuzzy application in engineering, combining fuzzy math with reliability analysis is an effective way, so we also do some study of it.Researches on fuzzy modeling and fuzzy application in engineering were done in the dissertation. The main work is concluded as follows:A method to construct fuzzy systems by considering selection of rules and their antecedents is proposed. First, the C4.5algorithm and the WM algorithm are used to identify initial fuzzy classifiers and Mamdani fuzzy models respectively. Then the fuzzy systems were optimized by NSGA-Ⅱ, one of the MOEAs. Difference from many results of existing research, the proposed method improve the accuracy by considering selection of rules and their antecedents while improving the interpretability by tuning of membership functions. Then fuzzy classifiers or Mamdani fuzzy models with well trade-off between accuary and interpretability can be created. The proposed approach is tested and against recent references on a benchmark problem. The results demonstrate the validity of this method.A method to learn the granularities of fuzzy partition is presented. In this method, the granularities of fuzzy partition and the parameters of membership functions are coded, the input variables and the granularities of fuzzy partition are optimized by NSGA-Ⅱ. To improve the accuracy and the interpretability, a more optimization is done for the model, and then the better trade-off between the accuracy and the interpretability is gotted. The approach is applied in building fuzzy classifiers and Mamdani fuzzy models, tested and against recent references on a benchmark problem. The results show its validity.Researchs are done on fuzzy FMECA, which combines analytic hierarchy process, fuzzy evaluation theory and FMECA. Then we apply the fuzzy FMECA in the reliability analysis for sliding plug door systems in Guangzhou metro line2. The analysis results are accordant with the practical situation, which shows the effectiveness of fuzzy FMECA. In conclusion, this paper makes a deepened discussion about fuzzy from theory and application. The proposed methods are all verified effectively, thus some references can be provided for the future researchs on fuzzy theory and application.
Keywords/Search Tags:fuzzy systems, accuracy, interpretability, multi-objective evolutionaryalgorithm, fuzzy FMECA
PDF Full Text Request
Related items