Method of locating unstable acoustic emission sources inside closed cavities based on a multiclassification model

To overcome limitations of traditional acoustic emission (AE) source localization methods, which are affected by the internal structure and composition materials of closed cavities, the authors carefully considered the essence of AE source localization, ingeniously transformed it into a multiclassification (MC) problem in machine learning, and proposed a method of locating unstable AE sources inside closed cavities based on the MC model. The inner space of the closed cavity is divided into multiple closed spaces, each of which is numbered. By alternately placing or generating unstable AE sources in different closed spaces, large numbers of AE signals generated in different closed spaces are captured. Through pulse processing, feature extraction, etc, a dataset containing multiple labels was created, and multiple MC models were trained. Parameter optimization was performed on the outstanding performer and then the optimal MC model was obtained. By integrating the majority voting rule, an AE localization model is constructed. In addition, a new definition of the AE source localization accuracy was proposed. Extensive experimental results demonstrate that the optimal plane and spatial MC model achieve classification accuracies of 89.58% and 84.25%, respectively, while the AE localization model achieves average plane and spatial localization accuracies of 100%. This fully demonstrates the good classification effect and generalization ability of the MC model and the AE localization model, effectively proving the feasibility and practicality of the proposed method. Meanwhile, a comprehensive comparison with existing AE source localization methods has verified the superiority of the proposed method. This study provides valuable references for AE and fault source localization research in related fields.