Research On Detection And Classification Method Of Remainders In Aerospace Elctronic Equipments

Aerospace electronic equipments are entire electronic control module or devices of special functions which is composed of various basic components, circuit boards, connectors and wires. They are significant components of national defense weaponry systems and aerospace. The remainders in aerospace electronic equipments have great effect on the stability and reliability of spacecraft. Based on the current problems of detection and classification of remainders in aerospace electonic equipments, this dissertation is going to improve the performance both of remainders detection and materials identification.Focusing on the extremely primitive situation of detection method of remainders in aerospace electonic equipments and the imperious demands, this dissertation employs the application of Particle Impact Noise Detection (PIND) and carries out general detection system of remainders in aerospace electonic equipments. Based on the turntable incentives, simulation model and dynamic model of turntable are established, which for the first time realize the controllable and variable applied stress during the detection of remainders in aerospace electonic equipments. In order to accomplish the detection and materials identification, broadband acoustic emission sensors with high sensitivity are applied to sense collision sound signal which include3variable mode acoustic emission detection channels and one acceleration sensor signal detection channel. And through high-speed data acquisition circuit collecting PIND test data, it is able to realize the automatic detection of remainders in aerospace electonic equipments and constitute the platform for remainder material identification and testing normative research.As aerospace electronic devices are bulky and acoustic emission signal attenuation effects are obvious, optimal model for multi-sensor mounting position is built and three optimal positions are confirmed based on transmitting wave propagation mechanism. The detection method of remainders in aerospace electonic equipments is constituted based on multi-sensor data layers fusion method. Through the homology signal measurement data correlation with isomorphic sensor and assigning data fusion weights, various types of interference signals are effectively suppressed. Background noise removing method based on wavelet packet transformation is carried out. Wavelet packet features full-band signal segment function is applied to effectively eliminates background noise system outside the passb and improve the signal to noise ratio. It is proved in the experimental that remainders over0.5mg are possible to be detected and the accuracy rate is87%. As the signal of remainder material available feature extraction is limited and effective description of the material characteristics of extra material is difficult, based on acoustic emission variable mode sensor signal conditioning circuit, three acoustic emission sensors’ signal and one acceleration sensor signal are detected simultaneously. It is for the first time to comprehensively apply acoustic emission sensor signal and the acceleration sensor signal. Extraction method of remaider material features is carried out based on Hilbert-Huang Transformation (HHT). Coefficient sequence of the solid-mode functions, Hilbert spectral centroid vectors and energy distribution vector are extracted from the perspectives of acoustic emission sensors and acceleration sensor. The comprehensive data of the detection sensors is applied to entirely describe the information of remainder material features, and establish the foundation of remainder material features identification.As the remainder material features identification is complex and effective identification is difficult to achieve, the characteristic quantities fusion method of multi-sensor is carried out, which is able to fuse multi-sensor characteristic quantitie to two feature Matrixes and one characteristic tensor. Through Nonnegative Matrix Factorization (NMF) and Non-negative tensor decomposition, the redundant information is efficiently eliminated. Remainder material features identification method is developed based on the relevance vector machine. Material identification accuracy rate, as the weighting factor with multi-classifier recognition results, can prove that metallic and nonmetallic materials over2mg can be identified and the accuracy rate is85%.