Microwave measurement techniques for industrial process monitoring and quality control

Doctoral thesis, 2018

Process monitoring and quality control by sensor measurements are essential for the automatisation and optimisation of many industrial manufacturing processes. This thesis is concerned with microwave sensing, which is a measurement modality with potential to improve the in-line sensing capabilities in several industries. Two process-industrial measurement problems are considered that involve the estimation and detection of permittivity variations for granular media in a fluidised or flowing state. For these problems, we present microwave measurement techniques based on resonant cavity sensors, accounting for the electromagnetic design and modelling of the sensor, signal processing algorithms, and experimental evaluation in relevant industrial settings. These measurement techniques make simultaneous use of multiple resonant modes with spatial diversity to improve the measurement capabilities. Furthermore, we exploit model-based signal processing algorithms where knowledge of the underlying physics is utilised for improved estimation and detection.

The first problem is to monitor the internal state of a pharmaceutical fluidised bed process used for film-coating and drying of particles. The metal vessel that confines the process is here treated as a cavity resonator and the complex resonant frequency of eight different cavity modes are measured using a network analyser. Based on the resonant frequencies, we estimate parameters in a low-order model for the spatial permittivity distribution inside the vessel, which can be related to process states such as the liquid and solid content of the particles in different regions.

The second measurement problem is an aspect of quality control, namely the detection of undesirable objects in flowing granular materials. We present measurement techniques based on resonant cavity sensors that are capable to detect the presence of small dielectric objects embedded in a flowing granular material. Detection algorithms that exploit the statistics of the noise caused by material density fluctuations and the characteristic signatures caused by an object passage event, are evaluated based on experiments which lead to quantitative assessments of the detection performance.