@article{LU2026, 
author = {Haoxiang LU and Yonghong LIU and Xinlei WU and Qiang LI and Pengxin ZHANG and Ziang WANG and Lixin WANG and Maoyuan ZHANG and Fanbo MENG and Renjie JI},
title = {Intelligent online monitoring and processing of electrical discharge and arc compound milling signals},
year = {2026},
journal = {Chinese Journal of Aeronautics},
volume = {39},
number = {6},
keywords = {Deep learning, Electrical Discharge and Arc Compound Milling (EDACM), Electrical Discharge Machining (EDM), Intelligent online monitoring, Online signal processing},
url = {https://www.sciopen.com/article/10.1016/j.cja.2025.103647},
doi = {10.1016/j.cja.2025.103647},
abstract = {Electrical Discharge and Arc Compound Milling (EDACM) presents significant advantages in terms of high efficiency and low cost for machining difficult-to-machine materials such as superalloys, titanium alloys, and ultra-high-strength steels, thereby exhibiting substantial potential for the manufacturing of critical aerospace components. However, its application is constrained by low machining accuracy and poor surface quality, highlighting the necessity for online monitoring and processing of EDACM signals to enable intelligent machining through dynamic adjustment of discharge parameters and enhance both machining accuracy and surface quality. However, the occurrence of atypical discharge waveforms and signal drift during the EDACM process further complicates online monitoring. To address these challenges, this paper proposes the Intelligent Online Monitoring and Processing Method (IOMPM) for EDACM signals, utilizing a hybrid neural network structure that integrates One-Dimensional Convolutional Neural Networks (1D-CNNs) with Long Short-Term Memory (LSTM) networks. An online processing theoretical model is developed to enable automatic identification and real-time calibration of the pulse period start point during the EDACM process, effectively overcoming the limitations of existing methods in terms of real-time performance and classification accuracy. The method constructs a dataset composed exclusively of current signals, incorporating 10% atypical discharge waveforms to improve the model’s adaptability to actual machining conditions. Using this dataset, the hybrid neural network model achieves an accuracy of 98.68% on the test set. Experimental validation confirms both the accuracy and real-time capability of the IOMPM, laying a solid foundation for enhancing both machining accuracy and surface quality, as well as for fulfilling the stringent quality and reliability requirements of aerospace manufacturing.}
}