Model-based Filtering of Interfering Signals in Ultrasonic Transit-Time Measurements

Flow measurement has long been a typical task of measurement technology. Various effects are used for this purpose, such as the Coriolis force, pressure differences, induction or acoustic transmission.

An energy-efficient principle for flow measurement in pipes is the use of ultrasound based methods. If the ultrasound is not scattered or reflected due to the homogeneity of the medium, no evaluation via the Doppler effect can be used. In this case the ultrasonic transit-time method is used.

For this purpose the transit-time difference between signals upstream and downstream is used. Conventional signal processing algorithms estimate the transit-time difference using the Hilbert transform, which is characterized by high noise robustness and fine time resolution. However, the estimated transit-time difference is strongly influenced by additive noise in the same frequency band.

In a new type of process, the ultrasonic wave is to be excited into the medium via the pipe wall. This results in a propagation of the sound in the medium and a propagation in the pipe wall as structure-borne sound. Due to the multipath propagation, an additive superposition of signals is measured in the receiver. The signals which pass through the medium carry a measuring effect which is present in the form of a time shift of the signals. Signals that are transmitted purely via the pipe wall do not carry a measuring effect, but are in the same frequency band. Due to the superposition in the same time and frequency range, this is an inverse problem.

The aim of the research is the extraction of the signal components carrying the measuring effect by filtering the possible structure-borne sound signals via regularizations. The additional boundary conditions that have to be met for the identification task are to be obtained from a physical understanding of the measurement system and the measurement process. For separation, time-frequency analysis techniques suitable for non-stationary filtering are to be investigated. Due to the process conditions, the methods must be robust against temperature fluctuations, density and viscosity changes of the medium.

Wave propagation during flow measurements