Automated, adaptive and uncertainty-aware smart measurements using machine learning

The European Union has identified six key enabling technologies to help achieve an interconnected, digitalised, resilient and healthier European society. Advancements in these technologies are important to support both EU industrial strategy and the European Green Deal and promote the advent of the circular economy envisioned in Industry 5.0.  For quality controls or in production lines many of these areas, including advanced materials, nanotechnology and advanced manufacturing, rely on a sequence of measurements, often at different scales. However, such measurements can often take days or only provide limited resolution, making them unsuitable for large-scale facilities where reliability or stringent production times are required.

This project will address these problems by using statistics, machine learning and compressed sensing techniques to develop ‘smart measurement’ solutions with in-built validated measurement uncertainties. The new techniques will be applied to areas which are foreseen to have the greatest impact including the use of photocurrent mapping for semiconductors, and hyperspectral imaging for nanoscale material characterisation. Along with industrial partners, the project will also produce experimental prototypes to demonstrate the effectiveness of the new methodology and generate software, good practice guides and data sets to help disseminate the results.

The work is anticipated to improve the reliability, efficiency and speed of production and quality control processes in a wide range of areas, including telecommunications, electronics, green energy devices and the semiconductor industries. The latter of which generates over €50 billion sales per year in Europe.