An Experimental-Data-Driven Deep Learning Strategy for Structural Health Monitoring of a Plate in Acoustic Fields

Detecting damages and degradation of aerospace components is crucial to guarantee mission safety and reliability, and to reduce costs and maintenance time. A structural health monitoring architecture, based on a Deep Learning approach, is introduced to analyze the vibration response acquired by accelerometers and investigate the system sensitivity to the loosening of bolts on a double-end constrained plate. Different boundary conditions are realized and a vibration signals database is created by subjecting the plate to several acoustic fields. The system is then trained and tested on the collected experimental data. Results suggest that the proposed approach is promising to detect potential modifications in the plate clamping conditions.