LAUSR

Abstract Due to the complex internal structure and uneven composition material of sealed electronic equipment, traditional acoustic emission source localization methods are unsuitable for the research of locating loose particles inside sealed electronic equipment. Aiming at this problem, the author analyzed the essence of locating loose particles from another angle, considered converting the loose particle localization problem into the multi-classification problem, and sought for solving. Based on the existing loose particle detection system, the multi-channel signal synchronous acquisition circuit was designed to synchronously collect loose particle signals. The two-stage dual-threshold pulse extraction algorithm and the multi-channel pulse-matching algorithm were designed to preprocess the collected loose particle signals. Multiple features with good localization performance were extracted and selected from the processed loose particle signals, and the localization data set was eventually established. The localization performances of different classification algorithms were analyzed and compared by performing them on the localization data set, respectively, then the parameters of better-performed Random Forest were optimized. Finally, the loose particle localization model based on parameter-optimized Random Forest was used to perform localization tests on the aerospace power supply. Test results show that the average localization accuracy is 84.08%, which is the highest localization accuracy achieved in aerospace engineering area of the world currently. It is an important supplement to the loose particle detection research and has important application value for improving the reliability of aerospace system. Theoretically, it can be applied to the research of collision signal localization with similar generation mechanism.