Abstract
Existing localization technology on loose particles inside sealed electronic equipment focuses on the selection and parameter optimization of classification algorithms;the features in the specific region has not been considered. Targeting this problem, this paper shifts the research focus to features in the localization dataset and proposes a feature optimization method. First, for the missing values in the localization dataset, the numerical filling method using a variety of statistical functions or directly discarding them are compared to obtain the optimal missing value processing method. Second, different processing methods using z-score standardization, min–max standardization, and row normalization are compared to obtain the best standardization or normalization method. Third, statistic-based and model-based feature selection methods are sutdied and the Pearson correlation coefficient, p-value, and the feature importance of decision trees are used to quantitatively measure the feature performance. On this basis of this measurement, the multi-channel characteristics of the localization technology on loose particles are comprehensively considered, and a multi-channel weighted threshold feature selection method is designed. The feasibility of the new method is verified on the localization dataset. Finally, verification datasets, representing aerospace electronic modules and aerospace power supplies, are established to verify the practicability and universality of the feature optimization method. The results show that the plane and spatial localization accuracy achieved on the localization dataset increased from 87.01% and 83.67% to 95.23% and 95.51%, respectively, after feature optimization, and the plane and spatial localization accuracy achieved on the verification datasets increased from 84.51% and 80.64%% to 90.57% and 89.16%, respectively. This is the highest localization accuracy currently achieved in aerospace engineering. The current study is an important supplement to loose particle detection research and has important practical value for improving the reliability of aerospace systems. This method can also be applied to other localization research on fault sources and acoustic emission source based on machine learning.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.