An adaptive PCA‐based method for more reliable ultrasonic guided waves SHM: Data‐driven modeling and experimental validation in high attenuating medium

This paper proposes a monitoring method for defect detection and localization in a structure operating under environmental and operational conditions (EOCs) variation. This method is based on a model obtained using Principal Component Analysis (PCA) of the healthy state data. To account for variation in EOCs, more particularly temperature change, the model is updated using a moving window over the collected signals (i.e. the learning is not limited to only a short period in the beginning of the monitoring duration). Defect detection was performed by calculating the Squared Prediction Error (SPE) between the current measured signal and its estimation extracted by the model. Once a defect is detected and since it does not impact the whole signal (i.e. only a part of it), the damage localization is performed by applying the PCA model on a sliding window over the signal. The test of the proposed method was achieved on a pipeline segment. A relatively small defect was created by removing material from inside the pipeline to simulate a minor corrosion. This corrosion-like defect was grown in six steps. The data were collected using Ultrasonic Guided Waves (UGW) technique. Results have shown that the created defect was successfully detected and localized. The proposed method for defect detection method is not limited to UGW, and could be applied to any SHM technique that provides time signals (i.e. transient response) such as acoustic emission.