A Multitasking Environment for Real-Time Monitoring of Discharging Activity During SACE Process Using LSTM

Real-time control of SACE gas film stability is crucial, as it significantly impacts micromachining repeata-bility and quality in this technology. Gas film stability and discharging activity are interconnected, and monitoring real-time parameters like mean discharge current and energy, which serve as indicators of gas film stability, is the first step in this effort. An intelligent algorithm deployed on a dSPACE platform uses LSTM for online discharge activity monitoring, identifying discharges and calculating indicators. Maintaining a short enough sampling time for prompt discharge detection presents overrun errors. Therefore, a real-time multitasking environment with a 1.6e-5 seconds sample time is executed. A more complex LSTM enhances detection accuracy but ex-tends execution time, potentially resulting in more unprocessed data loss. The research examines the real-time model with various algorithm feed batch sizes and LSTM complexities, particularly the number of hidden units. An example of a 2-hidden-unit LSTM demonstrates promising 90.45% accuracy, processing data every 264 milliseconds with a 131-millisecond batch (approximately 0.5 processing ratio), indicating superior performance. In the future, exploring LSTM hyperparameter optimization and real-time model parameter tuning is recommended to enhance accuracy and processing ratio.