With the kick-off meeting on June 3 and 4, 2025, at the University of Salerno, the work on the research project 'Automated, Adaptive and Uncertainty-Aware Smart Measurements using Machine Learning' (A3SmartML), which started in May and is planned for a duration of three years, officially began. The goal of the project is to develop intelligent measurement strategies to significantly increase the efficiency of multidimensional measurements. A3SmartML combines international expertise from several national metrology institutes, universities, and companies under the coordination of the Physikalisch-Technische Bundesanstalt (PTB) in Germany and is funded by the European Partnership on Metrology (EPM).
Multidimensional measurements are key technologies in metrology and enable applications such as material characterization through imaging at the nanoscale and quality control of semiconductors through photoconductive mapping with megapixel resolution. However, these processes are often characterized by long measurement times, which leads to problems in practical applications, as time is often limited in large production facilities or during quality controls in production lines. These challenges highlight the need for faster, more efficient, and smarter measurement methods. The improvement of these technologies is also in line with the European strategy for sustainability and digitization within the framework of Industry 4.0 and the European Green Deal.
Combined approaches enable the improvement of the efficiency of multidimensional measurements
To increase the efficiency of multidimensional measurements, this project develops intelligent measurement strategies by combining tools from machine learning, compressed sensing, regularization, and Bayesian statistics. In addition to improving measurement methods, the combination of machine learning with established approaches from statistics and methods such as compressed sensing offers new potential for the application of machine learning in metrology. In the context of the considered approaches of machine learning, reliable methods for uncertainty assessment are developed to enable control of the measurement considering the uncertainty.
The newly developed approaches are applied in measuring instruments for scanning hyperspectral imaging as well as photoconductive mapping of semiconductors. The goal is to develop experimental prototypes and demonstrate the efficiency of smart measurements in these areas. To demonstrate relevance for practical application, the methods will be implemented in several use cases.
To simplify the application and dissemination of the developed methods, extensive guidelines, documented software, and reference datasets will be provided. Long-term impact will be achieved through generic, automated, and adaptive measurement routines that are applicable to a wide range of multidimensional measurement methods. This way, companies in the manufacturing industry can utilize the project results to improve their quality controls and thereby reduce costs and waste in the sense of more sustainable production.
The WZL of RWTH Aachen plays a central role in this project in the development of methods for uncertainty assessment of intelligent measurement methods and coordinates the activities for disseminating the project results to relevant groups from science, industry, and in the field of university teaching in work package 5 'Creating Impact'.
The project (24DIT03 A3SmartML) is funded by the European Partnership on Metrology and co-financed by the European Union's Horizon Europe Research and Innovation Programme as well as the participating countries.
Contact:
