Data-driven methodology to detect and classify structural changes under temperature variations

This paper presents a methodology for the detection and classification of structural changes under different temperature scenarios using a statistical data-driven modelling approach by means of a distributed piezoelectric active sensor network at different actuation phases. An initial baseline pattern for each actuation phase for the healthy structure is built by applying multiway principal component analysis (MPCA) to wavelet approximation coefficients calculated using the discrete wavelet transform (DWT) from ultrasonic signals which are collected during several experiments. In addition, experiments are performed with the structure in different states (simulated damages), pre-processed and projected into the different baseline patterns for each actuator. Some of these projections and squared prediction errors (SPE) are used as input feature vectors to a self-organizing map (SOM), which is trained and validated in order to build a final pattern with the aim of providing an insight into the classified states. The methodology is tested using ultrasonic signals collected from an aluminium plate and a stiffened composite panel. Results show that all the simulated states are successfully classified no matter what the kind of damage or the temperature is in both structures.