Research On Confidence Evaluation Method Of Electromagnetic Algorithm

With the development of electromagnetic simulation technology and the increase of simulation requirements,the application range of simulation system is becoming wider and wider,and the degree of complexity is getting higher and higher.As people gradually increase their understanding of the importance of simulation systems,they also raise questions about its correctness.Therefore,it is extremely important to evaluate the reliability of simulation systems.Although statistical methods can reflect the relevance of data to a certain extent,they can't reflect the change of data trend and the difference of data comprehensively.We use the method of confidence sliding window error and Feature Selective Validation(FSV)to evaluate the credibility of the data from multiple aspects of amplitude,trend and error,and give a more comprehensive evaluation result.Verification,Validation and Accreditation(VV&A)theory is the basis for ensuring the credibility of the system.Therefore,before the credibility evaluation,the VV&A theory should be studied first,mainly to analyze the relationship between the theory,Modeling and Simulation,and credibility evaluation.On this basis,combined with the electromagnetic characteristics,VV&A theory is applied to the electromagnetic simulation process to ensure the credibility of the simulation results.On the premise of ensuring credibility,credibility evaluation is carried out.This thesis studies the FSV theory.Using this method to extract the direct current component,lowfrequency component and high-frequency component of the data,it is found that each component can clearly show the data amplitude,trend and more detailed features.The components are combined with each other according to some algorithm to form different measures: Amplitude Difference Measure(ADM),Feature Difference Measure(FDM)and Global Difference Measure(GDM).The evaluation level of the three differences is given according to the FSV evaluation table.Compared with common data evaluation algorithms,FSV method is more suitable for reliability evaluation of electromagnetic algorithm.Secondly,on the basis of the FSV method,the confidence sliding window error is added.For the jump points in the actual measurement or simulation results,different confidence levels and sliding windows of different sizes are used to reduce the impact of the jump points on the data evaluation.At the same time,the FSV method is used to analyze the impact of sliding processing on the data,and the direct current,low-frequency and high-frequency components of the data before and after the smoothing are extracted and compared.It is found that the smoothing processing can not only reduce the influence of noise on the data,but also reduce the influence of data irrelevant to electromagnetic characteristics on the results,and retain the original data features.Finally,the whole process of modeling and simulation,the confidence evaluation process of electromagnetic algorithms,and the applicability of electromagnetic algorithms are studied.On this basis,a variety of benchmark models have been established.The confidence sliding window error and the improved FSV method are used to compare the analytical solutions,similar algorithms and non-similar algorithms,and the credibility is verified from the perspectives of amplitude,trend and error.From the evaluation results,the method shows good results.