Structural health monitoring for layered structure using an autoencoder

Abstract

In this study, we proposed a structural health monitoring and diagnostic method for layered structures using the Autoencoder (AE). This method belongs to the primary diagnosis one, and its purpose is to identify the location of abnormality quickly after abnormality detection. An abnormality represents a decrease in the stiffness characteristics (spring constant) of the outer wall of a hierarchical structure when it deteriorates or is damaged. The proposed method has the following two features. The one is that the AE learns only the Frequency Response Function (FRF) under normal conditions calculated by the mathematical model of the layered structure. The FRFs of the actual structure under abnormal conditions are not required for the learning data. The other is that the FRFs under abnormal condition computed by the mathematical model are input to the learned AE to obtain a standard of the initial abnormality. The location of abnormality is identified by checking that the abnormality in the actual structure fits the standard of the initial abnormality. First, we considered a three-layered structure as a numerical example and verified the validity of the proposed method. When the method was applied to the three types of abnormal conditions, it was shown that the abnormal diagnosis could be performed correctly. Next, we constructed an experimental model of a three-layered structure, and realized three types of abnormal states similar to the numerical examples. We verified the applicability of the proposed method and showed that correct abnormal diagnosis was possible. As described above, the validity and applicability of the proposed method were clarified.

Keywords

• structural health monitoring
• primary diagnostic method
• frequency response function
• layered structure
• autoencoder