Table 1 Precision (p), recall (r), F-score (F), time-to-threshold (T2T), time-to-peak (T2P), and the normalized deviation between the threshold and the average value at the moment of flagging \([(\varepsilon _{\text {threshold}}^{\star } - \overline{\varepsilon _p^{\star }})/\sigma _{\varepsilon ^{\star }}]\), computed from multiple \(1000 t^\star\)-long signals at different Reynolds and wall distances.
From: Conditional POD for predicting extreme events in turbulent flow time signals
\(Re_c\) | \(y^+\) | p | r | F | T2T | T2P | \((\varepsilon _{\text {threshold}}^{\star } - \overline{\varepsilon _p^{\star }})/\sigma _{\varepsilon ^{\star }}\) |
---|---|---|---|---|---|---|---|
2000 | 10 | 0.755 | 0.733 | 0.744 | 0.130 | 0.315 | 1.131 |
3000 | 1 | 0.596 | 0.803 | 0.684 | 0.425 | 1.006 | 1.189 |
3000 | 10 | 0.636 | 0.717 | 0.674 | 0.147 | 0.322 | 1.121 |
3000 | 100 | 0.718 | 0.800 | 0.757 | 0.052 | 0.178 | 0.981 |
5000 | 10 | 0.601 | 0.853 | 0.707 | 0.091 | 0.200 | 1.156 |