Florida State University develops accurate chemical analysis tool

Chemists at Florida State University have developed a new tool that uses machine learning and robotics to identify the chemical composition of dried salt solutions from images. This tool can achieve 99% accuracy, which could revolutionize chemical analysis. By combining robotics with artificial intelligence, the researchers created a cost-effective and efficient method for chemical testing. The findings are detailed in the journal Digital Discovery. Co-author Oliver Steinbock emphasized the potential of using large databases and AI to analyze chemical compositions from photographs. The tool could speed up and reduce the cost of chemical analysis. It could be useful in various fields, including space exploration, law enforcement, and home testing. The team built on previous research that identified salt stains using machine learning, expanding their study from 7,500 samples to over 23,000 with the help of a robotic system called the Robotic Drop Imager (RODI). This robot prepares samples rapidly, allowing researchers to collect more data. The accuracy of their machine learning program improved significantly with additional images, allowing it to distinguish between different salt compositions and concentrations with high precision. Steinbock noted that traditional methods of chemical analysis require expensive equipment and significant expertise. This new method, relying on simple photo capture, could be a game changer, especially for space missions where weight is a concern. Developed for NASA, the project aims to enable rovers to analyze chemical compositions directly on other planets, without needing to return samples to Earth. The tool also has potential for use in laboratories, hospitals, and law enforcement to conduct quick and effective testing with only small sample amounts. The research highlights the transformative impact of AI on scientific discovery, making advanced analysis more accessible and opening new possibilities across various fields.