References | Model | Data set | Model details |
---|---|---|---|
This paper | MMN (Model 1) | Bondslipnet + SRRC-Net or data set from task A and B | (1) Meta-learning with Bondslipnet (2) Fine-tune with data from target task (3) Multi-tasking learning framework (4) Mahalanobis distance loss function (5) FRN normalization + dropout + L2 (6) Consider the Heissel matrix |
Finn et al. (2017) | MAML (Model 2) | Bondslipnet + SRRC-Net or data set from task A and B | (1) Meta-learning with Bondslipnet (2) Fine-tune with data from target task (3) Using traditional DNN framework |
Amaratunga, (2021) | Fine-tune (Model 3) | Bondslipnet + SRRC-Net or data set from task A and B | (1) Pretrain with Bondslipnet samples (2) Fine-tune with data from target task (3) Traditional DNN |
DNN (Model 4) | Bondslipnet + SRRC-Net or data set from task A and B | DNN framework with Bondslipnet and data from target task | |
DNN (Model 5) | SRRC-Net or data set from task A and B | DNN framework with data from target task | |
CAoB 2010 | Four-linear-stages model (Model 6) | Data from this reference | \(\tau = \left\{ {\begin{array}{*{20}l} {\left( {100f_{t} /d} \right)s} \\ {100f_{t} /3d(s - 0.025d) + 2.5f} \\ { - 2f_{t} /0.51d(s - 0.04d) + 3f} \\ {f_{t} } \\ \end{array} } \right.\begin{array}{*{20}c} {} \\ {} \\ {} \\ {} \\ \end{array} \begin{array}{*{20}c} {0 \le s < 0.025d} \\ {0.025d \le s < 0.04d} \\ {0.04d \le s < 0.55d} \\ {s \ge 0.55d} \\ \end{array}\) |
CAoB 2010 | Four-linear-stages model (Model 7) | SRRC-Net or data set from task A and B | \(\tau = \left\{ {\begin{array}{*{20}l} {\left( {b_{1} f_{t} /a_{1} d} \right)s} \\ {\left( {b_{2} - b_{1} } \right)f_{t} /\left( {a_{2} - a_{1} } \right)d(s - a_{1} d) + b_{1} f} \\ {\left( {b_{3} - b_{2} } \right)f_{t} /\left( {a_{3} - a_{2} } \right)d(s - a_{2} d) + b_{2} f} \\ {b_{3} f_{t} } \\ \end{array} } \right.\begin{array}{*{20}c} {} \\ {} \\ {} \\ {} \\ \end{array} \begin{array}{*{20}c} {0 \le s < a_{1} d} \\ {a_{1} d \le s < a_{2} d} \\ {a_{2} d \le s < a_{3} d} \\ {s \ge a_{3} d} \\ \end{array}\) |
Wu and Zhao ( 2013) | Three-nonlinear-stages model (Model 8) | Data from this reference | \(\begin{gathered} \tau = \frac{{\tau_{\max } \left( {e^{Bs} - e^{Ds} } \right)}}{{\left[ {e^{ - B\ln (B/D)} - e^{ - D\ln (B/D)} } \right]}} \hfill \\ K_{co} = \frac{c}{d},K_{st} = \frac{{A_{st} }}{{nS_{st} d}} \hfill \\ K_{si} = \left( {\frac{d}{25.4}} \right)^{{0.2 - 1.8\frac{{H_{rib} }}{{S_{rib} }}}} \;{\text{or}}\;K_{si} = d^{ - 0.4} \hfill \\ K = \left( {K_{co} + 33K_{st} } \right)K_{si} \hfill \\ \frac{{\tau_{\max } }}{{\sqrt {f_{c} } }} = \frac{2.5}{{1 + 3.1e^{{ - 0.47\left( {K_{co} + 33K_{st} } \right)}} }} \hfill \\ B = \frac{{0.0254 + K_{st} }}{{ - 0.0232 - 8.34K_{st} }} \hfill \\ D = 3\ln \left( {\frac{0.7315 + K}{{5.176 + 0.333K}} - 0.13} \right) - 3.375 \hfill \\ \end{gathered}\) |