Artificial intelligence technology provides a new idea to solve the bottleneck problem in the current laser processing field, and has a significant supplementary effect on laser micro/nano processing. It performs tasks through statistical techniques and numerical algorithms, without explicit program instructions, and can realize prediction, optimization of processing parameters and construction of complex dynamic models. For example, neural networks is used to model the physical process of laser-matter interaction without the physical model. Machine learning-based classification technology has significantly accelerated the intelligent process of defect detection in laser processing, and significant innovation has been achieved in the traditional laser processing process control method that relies on pre-defined processing paths.

Progress With artificial intelligence technology development, laser industrial manufacturing intelligence has become an important trend. Machine learning, as one of the main artificial intelligence technologies, is widely used in related fields, promoting a significant breakthrough and demonstrating specific potential to drive the next generation of ultrafast laser processing technology. This study therefore reviews key machine learning applications in various processes in the laser-micro and nano-processing field. This includes laser processing parameter optimization and process window prediction, real-time monitoring and control of the processing process, processing results prediction, and the investigation of auxiliary physical mechanisms. It summarizes and anticipates future applications and currently available improvement solutions where machine learning and laser processing are expected to intersect.

Conclusions and Prospects Firstly, intelligent laser micro/nano machining is an interdisciplinary direction, encompassing the two major fields of laser micro/nano processing and artificial intelligence. In recent years, the technological advancement of laser micro/nano processing has been relatively slow compared to that of artificial intelligence. Therefore, to promote the development of laser micro/nano processing research, it’s a good choice to continuously keep up with the advancements in artificial intelligence. Secondly, because the neural network approach is data-driven, model generalizability and accuracy are directly related to the amount of available data. The model generation method enables data augmentation and the random generation of new data that is extremely similar to the training data, thus artificially expanding the dataset. Finally, although deep learning is an extremely useful machine learning model, its “black box” nature prevents users from effectively controlling the neural network, hence, this has created doubts including skepticism regarding the “black box” concept. Consequently, artificial intelligence technology represented by machine learning technology has begun to empower all stages of laser micro/nano processing. In the future, it is expected to bring innovation to the process of laser micro/nano processing and form a new model of “artificial intelligence-driven scientific research”.

Since laser was invented, laser processing has completely transformed technologies such as cutting, cleaning, drilling, engraving, ablation, additive manufacturing, and welding. Laser technologies are used in a wide range of applications including aerospace, automotive, electronics, batteries, medical, 3D printing, semiconductors, sensors, and solar energy. With the development of technology, laser micro/nano processing is developing in the direction of finer (from millimeter to micron or even nanometer), more efficient (processing speed of hundreds of square millimeters per second), conformal (three-dimensional), and random (multiple materials) . However, due to the complexity of the physical mechanism of the interaction between laser and matter, its further development is restricted, accurate modeling cannot be achieved, and there are still obvious deficiencies in processing quality and conformal processing.