Table 2 Overciew of computational costs
From: On the use of neural networks to evaluate performances of shell models for composites
Process | Analyses | Cost |
|---|---|---|
BTD via FE | 32768 linear static analyses | 1 |
Data training generation | 3277 linear static analyses | 0.1 |
Training of NN | 1–3 layers with ten neurons | 0.01 |
BTD via NN | \(2^{15}\) queries to NN | 0.001 |