American scientists made an important step in automating research and development: they created a self-directed lab capable of conducting chemical experiments in real-time mode and collect data ten times faster vs traditional methods. The new technology changes the very principle of discovering new materials: from a slow and labor-intensive process to dynamic continuous work, when the system manages the experiments by itself, analyzes the results and undergoes through constant learning.
The researchers from the North Carolina University presented the development combining the capabilities of robotics and artificial intelligence. The preceding approach based on waiting for the completion of every chemical reaction before starting a new experiment was replaced with the so-called dynamic flows. Inside such a lab chemical mixtures are constantly changing, and the sensors are monitoring the changes in real-time mode. “Instead of one data point after 10 seconds of reaction we now have 20 every half a second. This is like switching from one photo to a full-fledged video about the reaction evolvement”, Professor Milad Abolhasani, the head of the research team, explains.
According to him, such model allows for collecting bigger volumes of data forming a detailed picture of what is going on. Owing to this, the system is faster in identifying high-potential materials for applications in such spheres as clean energy, electronics and environmentally friendly chemical processes.
Machine learning algorithms play the key role in this process: they analyze the in-coming data and determine the combinations of substances worth further testing. The more data the system receives, the faster and more accurate its forecasts are.
“We can identify the materials-candidates in several weeks or even several days, and not for several years, as before developing this system. Simultaneously we can reduce the costs and the environmental impact. This approach still has even more potential for further development”, Abolhasani summarized.
