Phase Space Reconstruction-Based Online Roughness Monitoring in Robotic Milling of Aircraft Skin Edge

Abstract

In robotic milling, monitoring surface roughness in real-time is essential for maintaining the quality of aircraft skin edges. However, existing techniques often struggle with accuracy and robustness due to the complex conditions during milling. This article introduces a novel approach to real-time predict surface roughness in robotic milling of aircraft skin edges. The method begins by dynamically adjusting vibration signals using the frequency response function to maintain prediction accuracy. Subsequently, the phase space reconstruction is employed to map the vibration signals from time domain into phase space diffeomorphic to the robotic system, thereby uncovering the system’s dynamic characteristics. Then, the Gramian angular summation field algorithm is used for feature extraction from the phase space trajectories, followed by classification using a convolutional neural network. A vibration-roughness dataset is constructed based on practical experiments in order to train the proposed model. The predictability of the machining system is demonstrated quantitatively. Furthermore, the model’s practicality is validated through online experiments. The prediction results based on the test set confirm that this approach offers significant improvements in prediction accuracy and robustness over existing methods.