Improvement of Uncertainty Evaluation on Impact Testing Machine by Implementation of Computer Vision

The previous research on uncertainty evaluation of an impact testing machine has been enhanced through the implementation of computer vision. Data collection involves using cameras to record the analog indicators on the impact testing machine during operation. The indicator readings are taken using computer vision, which predicts the position of the pendulum arm's angle. By using computer vision, the resolution of the indicator has been improved compared to manual readings. In cases where resolution is the dominant source of uncertainty, this study shows that the manual reading resolution of 0.2° can be improved to 0.07° using computer vision. Consequently, the estimated relative uncertainty of absorbed energy can potentially be reduced from 0.88% to 0.36%. These findings highlight the significant role of Artificial Intelligence in improving the evaluation of uncertainty in impact testing results.