An intelligent impact load identification and localization method based on autonomic feature extraction and anomaly detection

Abstract

The main problems of impact load identification are the complexity of dynamic data and the establishment of theoretical model. This paper proposed an intelligent impact load identification and localization method based on autonomic feature extraction and anomaly detection. The method firstly extracts features automatically from dynamic response through principal component analysis and trains ensemble learning model to identify and locate impact loads. Then the method sifts abnormal impact samples through mathematical statistics and studies the causes of the impact anomaly to improve the accuracy of identification. An anomaly detection model is also trained. The effectiveness and efficiency of the method are verified by identifying the force peak and location of impact loads acting on a thin-walled cylinder structure. The final error of the proposed method is less than 6.62% and the accuracy of anomaly detection is more than 91%. The result proves that the automatic feature extraction and anomaly detection make the proposed method more intelligent and superior to the existing methods. The method can be applied to various engineering structures and exhibits potential for intelligent structural design and structural health monitoring.