Figure 10
The figure shows a single box plot titled “10-Fold Cross-v00alidation Accuracy”. The horizontal axis has no category labels, and the vertical axis on the left is labeled “Accuracy”. The accuracy scale ranges from approximately 0.98 at the bottom to slightly above 0.9950 at the top, with evenly spaced tick marks at an interval of 0.0025. A rectangular box represents the interquartile range of the cross-validation accuracy values. The lower edge of the box is positioned near 0.9875, and the upper edge of the box is positioned near 0.9918. Vertical whiskers extend from the top and bottom of the box. The upper whisker reaches to approximately 0.9955, and the lower whisker extends down to around 0.9835. Below the lower whisker, a single small hollow circular marker is visible near 0.9790, indicating an outlier accuracy value. Grid lines run horizontally across the background, aligned with the accuracy tick marks. Note: All numerical data values are approximated.Boxplots illustrating the distribution of classification accuracy and RUL prediction RMSE obtained from 10-fold cross-validation of the CNN–LSTM digital twin model. The narrow interquartile ranges and minimal outliers indicate high stability and reproducibility across all folds. These results confirm the statistical robustness and generalization capability of the proposed framework for AE-based prognostics of pressure vessel integrity