When modelling and data analysis becomes part of the measuring
Introduction to the European Metrology Network for Mathematics and Statistics
Introduction to the European Metrology Network (EMN) for Mathematics and Statistics
Algorithms based on mathematics and statistics are behind many advanced metrology applications. New and emerging fields in particular can gain a lot by using simulations and data analysis. “Mathematical simulations are still needed for many classical physics applications, where the equations, however, are well defined. Bigger challenges are posed by the demands for new emerging metrology fields and technologies, like medical imaging, or environmental monitoring.” says Markus Bär, acting Chair of the MATHMET network.
Digitalisation is a huge part of the transformation that many industries are now undergoing, increasing the demand for metrological expertise. To meet these challenges efficiently the MATHMET network wants to foster collaboration between scientists in Europe as well as raise their visibility with industry partners. “Generally, the mathematical part is not easy to develop and often it is not accessible to others. We want to change that. The main objective is to have a single reference point for information and services for our stakeholders, like software tools and training courses. With the help of improved modelling and statistics, we can help people how to analyse data better and achieve higher returns for the measurement efforts,” Markus says.
Advanced technologies often require algorithms because they work with lots of sensors and a large amount of data. Often, it is not possible to directly extract the information you are looking for. Markus gives an example, “In nanometrology, where you look at very small length scales, it’s often not possible to image directly with a microscope. Instead, scientists use indirect information collected from scattering experiments. They then need to employ mathematical models to extract the right information.” says Markus. For cases where high-resolution microscopy is available, huge amounts of data are obtained from experiments. Then, modern data compression techniques developed by applied mathematicians become mandatory.
Metrologists have to keep in mind that data processing is an equally dangerous source of errors as the physical measurement itself. As Markus points out, “Practitioners sometimes tend to underestimate or overlook these errors. Furthermore, there are a lot of new algorithms in the fields of compressed, machine learning, or artificial intelligence that have started to enter the software behind modern measurement devices. One crucial question is how to test these algorithms properly?” Sebastian Heidenreich, the coordinator of a joint network project (JNP) supporting MAHTMET, explains that such issues will become more important in the near future. He foresees that data science and mathematics departments are growing, both in industry and in scientific institutes. Better data analysis will often provide an advantage, where the measurement hardware is already well developed. Simulations of virtual measurement devices, also known as “digital twins” will be key to develop new technology. Facilitating this change will be an important part of the MATHMET network, whose members represent the mentioned metrological, scientific and technological areas and shall provide input for guidelines as well as software and reference data sets.