Integrating PCA and GVF snakes for high-accuracy metal surface defect detection in robotics automation

Residual stresses (Rs) surface affects the corrosion resistance and fatigue life of machined components. Traditional methods of computing Rs are time-consuming and may damage the parts. Machine learning-based real-time prediction is a feasible method for monitoring residual stresses effectively. Using multi-source data, CNN for feature extraction, PCA for dimensionality reduction, and GPR for precise Rs prediction, this work attempts to develop an intelligent prediction technique. Includes a new approach by making use of GPR in residual stress prediction, CNN in feature extraction, and PCA for reducing data complexity. All dynamic multi-source cuts in cutting force, temperature, power, and noise are utilized to train the model. It performed better compared to the traditional approaches on speed and accuracy, being 99.1% in accuracy, 98.9% in precision, 99.2% in recall, and a Root Mean Error of 0.9%. In this context, these outcomes demonstrate well how well it predicts the residual stress real-time when integrated.MThe proposed approach is highly accurate and reliable for residual stress prediction and has the ability to monitor the surface quality instantaneously in intelligent production systems. It shows notable improvement over the state-of-the-art approaches by combining CNN, PCA, and GPR.

Text 122521_.pdf - Published Version Restricted to Repository staff only Download (1MB) Official URL or Download Paper: https://link.springer.com/article/10.1007/s42979-0...

Actions (login required)

View Item