The aim is to widen the applicability of the currently existing microwave sensor systems and develop new innovative sensor system solutions.Objectives
- Develop a widely applicable and high-performance estimation framework for inverse problems in industry by combining physical modelling with machine learning and big data techniques
- Validate and assess the performance of the estimation framework by optimised measurement campaigns in industrial settings
During the last 10 years, we have focused our efforts on microwave measurement systems for industrial use within the context of Chase, with in-kind contributions from related projects funded by SSF and VR. In addition, AstraZeneca has funded us directly for the development of a state-of-the-art sensor system for batch processing of pharmaceutical products. In collaboration with Food Radar Systems, we have developed a novel system for continuous processing of food products, and this system is now introduced on the global market. For medical applications, we have on-going projects with Medfield Diagnostics (determining if a stroke is of bleeding type or clot type) and Micropos Medical (positioning of the prostate during radiation therapy). Each of these sensor systems is tailored for a particular application.
Microwave sensor systems used in industry today typically estimate a single parameter based on a measurement with one or possibly two sensors. Thus, a rather simple model of the system is often sufficient and, then, the sought quantity can be derived analytically. However, for most applications, it is of great interest to extract much more information than a single parameter. Thus, it is necessary to use broad frequency-band measurements, many microwave sensors/antennas, and instruments with large dynamic-range and good phase-synchronisation. As a consequence, reliable and accurate estimates of the quantities of interest in real time require research that explores new directions. In the framework of ChaseOn, we intend to combine physics-based signal processing with cutting-edge techniques within machine learning and big data.
The output of the project is (i) a new category of algorithms and (ii) their software implementation that exploit techniques that have been successfully applied in other areas. This work makes it possible to guide the solution of the inverse problem in a robust manner for a broad range of application scenarios. Thus, the hardware of a particular sensor system can be used for a substantially wider range of applications by virtue of flexible, accurate, and reliable software processing solutions of the measurement data.
Chalmers, Food Radar, Keysight, Medfield, Saab, UniqueSec, VGR
Professor Thomas Rylander