Skip to main content

Table 7 Damage localization, case 2

From: Fully convolutional networks for structural health monitoring through multivariate time series classification

\(\varvec{N}_{\varvec{0}}\) \(\varvec{\lbrace {\mathcal {F}}_{*} \rbrace }\) \(\varvec{A}_{\varvec{l}}\)
1 \(i=8\) 0.226
2 \(i=\,\)(4, 8) 0.722
3 \(i=\,\)(2, 4, 8) 0.774
4 \(i=\,\)(2, 4, 6, 8) 0.906
5 \(i=\,\)(2, 4, 6, 7, 8) 0.865
6 \(i=\,\)(2, 4, 5, 6, 7, 8) 0.937
7 \(i=\,\)(2, 3, 4, 5, 6, 7, 8) 0.899
8 \(i=\,\)(1, 2, 3, 4, 5, 6, 7, 8) 0.993
  1. Accuracy \({\mathbb {A}}_{l}\) of the classifier \({\mathcal {G}}_l\) evaluated on \({\mathbb {D}}^l_{test}\). Different numbers \(N_0\) of input channels \({\mathcal {F}}_{*}\), related to \({u}^{sh}_i\), are employed. Here, \(f_{min}=5\) and \(f_{min}=7\) Hz