Font Size: a A A

Research On Multivariate Simulation Result Validation Methods Under Uncertainty

Posted on:2017-04-07Degree:MasterType:Thesis
Country:ChinaCandidate:H Y WangFull Text:PDF
GTID:2272330503987235Subject:Control Science and Engineering
Abstract/Summary:PDF Full Text Request
Uncertainty has widely existed in simulation and it has received more and more attention in the field of simulation. Research on result validation of simulation system under parameter uncertainty is of practical demand. How to validate models with multiple outputs whose parameters contain disparate uncertainty, output is static or dynamic state is of great significance.This paper focuses on the following research issues.Firstly, for static outputs, the paper proposed a result validation method based on an improved Jousselme evidence distance. For problems that Jousselme evidence distance and its modified evidence distance have, the paper presented an improved Jousselme evidence distance. The paper proposed to convert outputs into evidence theory and used the improved Jousselme evidence distance to measure differences to get the validation result.Secondly, for dynamic outputs, the paper proposed a result validation method based on time series error evaluation. The paper improved the phase error evalution algorithm of time series error evaluation to extend its application areas. Aimed at features of dynamic outputs, it used time series error evaluation to convert dynamic outputs into static phase, amplitude and topology errors. Then the improved Jousselme evidence distance is used to obtain the consistent between outputs.Then, the paper proposed to use principal component analysis and interval principal component analysis for multiple outputs in simulation system in simulation result validation. On condition that there existed both aleatory uncertainty and epistemic uncertainty, it introduced interval principal component analysis to reduce the output dimensions and eliminate the correlation between outputs. On condition that there existed only aleatory uncertainty, it introduced principal component analysis.Finally, this paper developed a simulation result validation tool combined with theoretical achievements in this paper. The development process is introduced from main screens, function modules and database design three aspects. And it is applied to a terminal guidance system of a certain vehicle.
Keywords/Search Tags:result validation, uncertainty, evidence distance, time series error evaluation, principal component analysis
PDF Full Text Request
Related items