Resonant Microwave Sensors for Industrial Applications

Licentiate thesis, 2015

This thesis presents developments in the area of microwave-based sensor systems for industrial applications. The work is motivated by a need for advanced measurement techniques in industry in general, and the process industry in particular. Microwaves have several advantages for many measurement problems encountered in process industries, including the possibilities for non-destructive and non-invasive measurements in-line. Furthermore, microwaves have good penetrability in many biological and chemical substances and interact strongly with water, which enables sensitive moisture measurements. The sensor systems developed in this thesis are based on microwave resonances, which is a well-suited technique for measurements inside hollow metallic enclosures such as vessels and pipes. These structures are commonly used in process industries for storage, transportation and processing of materials and they form natural environments for hosting resonant electromagnetic modes at microwave frequencies. A main novelty is that multiple resonant modes are used simultaneously to improve the measurement performance. Detailed electromagnetic modelling based on the finite element method is employed, with emphasis on eigenvalue problem formulations for resonant systems. Two specific process-industrial applications are considered, namely the monitoring of pharmaceutical manufacturing processes and the detection of undesirable objects in different material mixtures. In the first application, a fluidised-bed process for coating and drying of pharmaceutical particles is monitored by measurements of the complex resonance frequencies of cavity-modes inside the process vessel. Experimental results from a real fluidised-bed process demonstrate the usefulness of this measurement technique. For the second application, a resonant microwave sensor is developed for measurements on low-permittivity materials, such as dilute powders, that are flowing in metal pipes. The problem of detecting undesired dielectric objects in the material flow is particularly studied. Two detection algorithms based on the likelihood-ratio test are investigated for this purpose, where the first is based on measured resonance frequencies and the second uses uncalibrated S-parameter data. The sensor and detection algorithms are evaluated based on simulated and measured data, and favourable detection properties are observed.