Issue 67

S. Chinchanikar et alii, Frattura ed Integrità Strutturale, 67 (2023) 176-191; DOI: 10.3221/IGF-ESIS.67.13

rate of 0.01 and 1000, respectively. However, this study considered an ANN model using one hidden layer with ten neurons for further discussion, considering its lowest computational time and better prediction accuracy in comparison to other ANN models. Fig. 16 depicts the training performance of an ANN model using one hidden layer with ten neurons. The optimal validation performance was achieved at epoch 139, with a score of 3.989 x 10 -5 with a prediction accuracy of 0.9975. The average squared error between objectives and outputs, or mean squared error, is used to assess the effectiveness of neural network training. Lesser values are preferable. The correlation between outputs (predicted values) and goals (inputs) is measured by regression (R-squared) values. Figs. 17(a), (b), (c), and (d) illustrate neural network regression graphs with regression coefficients discovered during model training, validation, and testing, as well as for the complete data set. In the developed ANN model, regression coefficient values obtained can be seen as 0.9934 for training data, 0.9928 for test data, and 0.9936 for validation data.

Best Validation Performance is 3.9896e-05 at epoch 139 Train Validation

10 -3

Test Best

10 -4

10 -5

Mean Squared Error (mse)

10 -6

0

100

200

300

400

500

600

639 Epochs

Figure 16: Neural network training performance.

(a)

(b)

(c) (d) Figure 17: Neural network (a) Training, (b) Validation, (c) Test, (d) All data set.

187