Vision-based smart laser manufacturing via advanced machine learning and control approaches
Laser machining is a versatile non-contact fabrication method that finds application across many industries and materials. However, the flexibility of the process, which allows for numerous variations in wavelength, pulse energy, and pulse length, makes optimizing the process difficult. Small changes during manufacturing, such as variations in laser power or beam shape, can lead to subpar product quality, resulting in higher costs and waste. To achieve sustainable and intelligent laser manufacturing, a set of defect modeling, real-time quality evaluation, and performance control methodologies are needed. The highly complex light-matter interactions in laser machining make it challenging to build accurate physics-based models. Data-driven and learning-based approaches, however, provide a promising way to optimize productivity in real-time, reduce costs and waste, and develop advanced software-sensing schemes and control architectures. This research aims to devise learning-based data-driven techniques for modeling laser machining processes and defining quantitative quality indicators to optimize productivity in real-time. The research will leverage active quality regulation, using images captured by cameras that directly observe the cutting process to create virtual sensors for control systems that regulate quality and productivity in real-time.
Back to Current Students