The world of industrial laser processes is on the cusp of a revolution. Machine learning has taken center stage, enabling researchers at Empa’s Advanced Materials Processing laboratory in Thun to simplify complex laser-based techniques. The goal? To make these processes more affordable, efficient, and accessible for industries such as automotive and aviation, where precision is paramount.

Additive manufacturing (3D printing) using lasers is one such process that has been optimized using machine learning. Researchers Giulio Masinelli and Chang Rajani focused on the powder bed fusion (PBF) method, which involves melting metal powder in exactly the right spots to create a final component. Before production begins, however, a series of preliminary tests is typically required to determine the optimal settings for parameters such as scanning speed and laser power.

The two researchers used machine learning to reduce these experiments by around two-thirds while maintaining product quality. They “taught” their algorithm to recognize when the laser was in conduction or keyhole mode (where metal is melted or vaporized, respectively) using optical data from sensors incorporated in the laser machines. Based on this information, the algorithm determined the settings for the next test run.

This breakthrough has far-reaching implications. “We hope that our algorithm will enable non-experts to use PBF devices,” says Masinelli. Integration into the firmware of laser welding machines by device manufacturers would be all it takes to make machine learning-driven 3D printing accessible to a wider audience.

The researchers have also explored real-time optimization of laser welding processes using special computer chips called field-programmable gate arrays (FPGAs). These FPGAs enable the evaluation and decision-making process to occur in near real-time, even for complex tasks such as observing and controlling laser parameters.

Empa’s Masinelli and Rajani are confident that machine learning and artificial intelligence can contribute significantly more to the field of laser processing of metals. They will continue to develop their algorithms and models, expanding their area of application through collaboration with research and industry partners.

The future looks bright for industrial laser processes, thanks to the power of machine learning.

The discovery of a new phenomenon that allows researchers to control solid objects in liquids using ultrasound waves has sent shockwaves throughout the scientific community. The technique, which concentrates solid particles suspended in a liquid by inducing spin through the use of high-frequency sound waves, holds immense promise for biomedical testing and drug development.

At the heart of this innovation is the work of Chuyi Chen, an assistant professor of mechanical and aerospace engineering at North Carolina State University, who explains that the process involves creating ultrasound waves on the surface of a piezoelectric substrate. This causes the fluid inside the droplet to stream in a circle, while the surface tension of the droplet prevents it from spreading out into a flat sheet.

As a result, particles within the droplet are driven by a combination of forces – from the ultrasound waves, the spinning droplet itself, and the fluid moving within the droplet. This movement creates a helical pattern as particles corkscrew through the liquid to come together at a central point.

“This is a novel way of concentrating solid particles in a liquid solution,” Chen notes, “which can be extremely useful.” For instance, by making it easier for sensors to detect relevant materials within cells, this technique could significantly improve biomedical assays.

However, to fully harness this phenomenon and develop technologies that make use of it, researchers need to understand the underlying physics. This is precisely what Chen’s team has achieved in their groundbreaking research.

“This paper lays out in detail the physics responsible for controlling particles inside the droplet,” Chen says. “Now that we understand the forces involved, we can make informed decisions and engineer technologies to concentrate particles in a liquid sample in a controlled way.”

One key aspect of these findings is that researchers can manipulate several parameters – including surface tension, droplet radius, and ultrasound wave amplitude – to influence particle movement within the droplet. This gives them multiple mechanisms for fine-tuning rotation and behavior.

In addition to its potential utility in biomedical research, this new technique also holds promise for exploring fundamental physics questions related to rotating systems. By creating miniaturized tornado-like vortex flows or studying Coriolis-driven transport on a small scale, researchers can gain valuable insights into these phenomena without the need for extensive resources.

The work behind this discovery was supported by grants from the National Institutes of Health and the National Science Foundation. As Chen notes, “This research opens up new avenues for exploring complex physics questions in a compact and relatively inexpensive way.”

The world of holography has taken a significant leap forward with the development of a one-pixel camera that can record three-dimensional movies. This innovative technology, pioneered by researchers at Kobe University, has opened doors to holographic video microscopy, allowing for minimally invasive, three-dimensional biological observation.

Holograms are no longer just used as fun safety stickers on credit cards and electronic products; they have real scientific applications in sensors and microscopy. Traditionally, holograms required a laser for recording, but recent advancements have made it possible to record them with ambient light or light emanating from a sample. Two main techniques have emerged: “FINCH,” which uses a 2D image sensor to record movies in visible light, but is limited by an unobstructed view; and “OSH,” which employs a one-pixel sensor to record through scattering media and outside the visual spectrum, but can only capture motionless objects.

To overcome the limitations of OSH, Kobe University researcher Yoneda Naru and his team created a high-speed setup using a digital micromirror device. This device operates at 22 kHz, a significant improvement over previous devices with a refresh rate of 60 Hz. With this breakthrough, the researchers can now record 3D images of moving objects and construct a microscope that captures holographic movies through light-scattering objects.

The results of their proof-of-concept experiments were published in the journal Optics Express, demonstrating the potential for this technology to revolutionize holographic video recording. Although the current frame rate is relatively low, Yoneda and his team showed that, in theory, they can achieve a standard screen frame rate of 30 Hz through compression techniques like sparse sampling.

The applications of this technology are vast, with potential uses in minimally invasive biological observation, medical research, and beyond. While there are still obstacles to overcome, the researchers are now exploring ways to optimize patterns for projecting onto samples and using deep-learning algorithms to transform raw data into high-quality images.

This groundbreaking research was funded by several organizations, including the Kawanishi Memorial ShinMaywa Education Foundation and the Japan Society for the Promotion of Science. The collaboration between researchers from Kobe University and Universitat Jaume I has paved the way for further innovation in holographic video recording.