A novel framework to identify delamination location/size in BFRP pipe based on convolutional neural network (CNN) algorithm hybrid with capacitive sensors

Abstract

Failure detection-based Electrical Potential Change (EPC) is a promising technique. In this article, the internal layers delamination is inspected in basalt fiber-reinforced polymer (BFRP) pipe under long-term fatigue loading (LTFL) of internal pressure effect via an Electrical Capacitance Sensor (ECS) by evaluating the dielectric characteristics of pipe materials and classification between intact and delamination stats. The 3D maps of the capacitance array values and EPC distribution of node potential are tested. The maps can reflect delamination between pipe layers based on the researcher's previous works, however, because the pipes are modeled in 3D, therefore, the bending and twisted effects of the model make these maps not a good choice to accurately detect delamination location/size. Therefore, a new type of convolutional neural network (CNN) algorithm is adopted to train and test the EPC maps to evaluate delamination location/size. The training accuracy of the current technology ($P\%$), recall rate ($R\%$), and F-score ($F\%$) are equal to $95.2\%$, $93.7\%$, and $90.9\%$ respectively, which indicates that the current technology shows identification efficiency and accuracy of the technology. The proposed method results converge with available traditional methods in the literature for assessing the delamination location/size such as the response surface methodology (RSM), and the error band from the diagonal line is less than $4.86$ and $1.14$ degrees for location and size respectively, thus validating the proposed technique's reliability, accuracy, and applicability for the relevant structures.