To achieve an optimal trade-off between processing efficiency and dimensional precision in the assisted electrode electrochemical discharge machining (AE-ECDM) of thermal barrier coating (TBC)-coated superalloys, a 1D convolutional neural network (1D CNN) and gated recurrent unit (GRU) enabled AE-ECDM is proposed, in which processing conditions are tuned dynamically in response to process evolution. The electrical discharge behaviors during distinct processing stages of film cooling holes on TBC-coated superalloys were examined. A 1D CNN-GRU network was constructed, where the 1D CNN extracts local features from voltage signals and the GRU analyzes the temporal dependencies of the features. Subsequently, the performance of the network was evaluated, and a hardware platform was established to conduct validation experiments. The 1D CNN-GRU-enabled AE-ECDM reduces radius deviation by 59.4% and electrode wear length by 41.3% relative to aggressive processing conditions intended for maximum stock removal, and shortens processing duration by 66.1% relative to mild processing conditions optimized for quality control. It demonstrates high efficiency and quality in the processing of film cooling holes on TBC-coated superalloys. Furthermore, the 1D CNN-GRU network manifests competence in the manufacturing of abradable sealing coatings (ASCs) and double-walled structural components.
Ning et al. (Wed,) studied this question.