Hole Depth Prediction in a Femtosecond Laser Drilling Process Using Deep Learning.

Area of Science:

• Manufacturing Engineering
• Optical Engineering
• Artificial Intelligence

Background:

• Measuring drilled hole depth during high-aspect ratio laser drilling is challenging and time-consuming.
• Accurate depth measurement is crucial for process control and quality assurance in laser machining.

Purpose of the Study:

• To develop a non-contact method for estimating drilled hole depth using 2D images.
• To leverage deep learning for precise depth prediction in high-aspect ratio laser drilling.

Main Methods:

• Utilized a deep learning methodology to predict hole depth from captured 2D hole images.
• Optimized laser drilling parameters (power, cycles) and image acquisition conditions (brightness, exposure, gamma).
• Employed an interferometer to extract contrast data for image analysis.

Main Results:

• Achieved a prediction precision of within 5 μm for holes up to 100 μm in depth.
• Identified optimal microscope exposure duration and gamma values for accurate hole form forecasting.
• Demonstrated the feasibility of using 2D imaging and deep learning for in-situ depth estimation.

Conclusions:

• The proposed deep learning approach offers an efficient and accurate solution for estimating drilled hole depth.
• This method can facilitate real-time monitoring and control of high-aspect ratio laser drilling processes.
• The findings contribute to advancements in automated laser manufacturing and quality inspection.