Research on Material Identification Technology of Sealed Relay Remainder Material Based on Decision Tree
PIND is one of most important tests for sealed electronic components.At present, this test can only detect the presence or absence of remainder and cannot infer the material information of remainders.The material information can provide reference for the subsequent processing decision of remainders, it is of great significance to trace the process steps in which the remainders generated.Currently, the accuracy of the material identification detection is less than 70%, which needs further improvement.Machine learning is used for remainder detection in this paper, which improves the utilization of data information.Based on the orthogonal design theory, the experimental scheme is designed, and the signal data set of the remainder material is established.The filtering and segmentation algorithm are designed to extract the effective signal.Multi-feature combinations of frequency domain and time domain are designed as input parameters.Classification model of remainder material signals is established based on decision tree algorithm.The trained classification model can achieve 85% accuracy of metal and non-metal classification for unlabeled test sets.High-precision identification of the remainder material is successfully realized.
remainder sealed relay identification of remainder material decision tree
Huizhen Yan Xiaohang Cao Zixin Wang Guotao Wang
Harbin Institute of Technology, Harbin 150001 Guizhou Space Appliance Co., Guiyang 550000 Heilongjiang University, Harbin 150008 Harbin Institute of Technology, Harbin 150001;Heilongjiang University, Harbin 150008
国际会议
江苏苏州
英文
237-243
2019-11-04（万方平台首次上网日期，不代表论文的发表时间）