Feedback-control based laser micromachining with real-time plasma detection and neural networks.

This research aims to develop a feedback control laser processing system using artificial intelligence. Real-time online measurement of hole depth for blind hole processing is challenging. In this study, electric fields were used to detect laser-induced plasma signals and a neural network was used to predict the hole depth to overcome this difficulty. This electric field was generated by a pair of electrode plates placed above the drilled surface to provide hundreds of volts of DC voltage. During laser processing, laser-induced plasma interferes with the electric field and generates a detection signal. This signal is sent back to the neural network to estimate the hole depth. This study also trained a neural network to predict the machining parameters required for a given hole depth and diameter. If the error in the hole size after processing exceeds the threshold value, the feedback control system will perform immediate compensation processing to reduce the error. This avoids the difficulty of relocating the material. Therefore, this study integrated laser processing, online laser-induced plasma detection, and neural network-based parameter estimation to form an advanced feedback control laser machining system that can solve the problems of processing parameter prediction, real-time hole depth measurement, and compensation processing relocation encountered during the processing process. The experimental results demonstrate that the proposed method can effectively compensate for processing and reduce errors. [ABSTRACT FROM AUTHOR]