Optimization of Sensors for Structure Damage Detection Using Deep Learning Approach

An Enhanced Structural Damage Identification Approach Using Statistical Analysis and Optimization

“…A one-dimensional CNN (1D-CNN) is designed to extract essential local features from input data [32,34,35]. This is achieved by applying a small kernel that slides over the data, thereby capturing valuable patterns and information [36].…”

“…The process begins with the input layer, where raw data are received and prepared for the subsequent layers. [32,34,35].…”

“…An essential step in this approach is to assess the performance of the sensors following the concept of 'having the right tools for the job.' We selected five sensors out of the 30 sensors based on our previous findings from literature [32], which showed a relatively lower performance when the signal difference was not used. Including these sensors in the comparison allowed for a more in-depth analysis, and the results are presented in Table V.…”

“…In accordance with the research conducted in [32], we adopted the metrics RSdr and RSopt as indicators. The structure under consideration features n joints, as illustrated in Fig.…”

“…This approach significantly increases the computational complexity. Furthermore, the method introduced in [32] achieves promising detection performance, but exhibits a high degree of sensitivity to sensor selection, demanding meticulous adjustments based on the choice of sensors and detection region.…”