Magnetic resonance velocimetry in high-speed turbulent flows: sources of measurement errors and a new approach for higher accuracy

Abstract

This study focuses on the measurement accuracy of Magnetic Resonance Velocimetry (MRV) in high-speed turbulent flows. One of the most prominent errors in MRV is the displacement error, which describes the misregistration of spatial coordinates and velocity components in moving fluids. Displacement errors are particularly critical for experiments with high flow velocity and high spatial resolution. The degree of displacement error also depends on the sequence structure of the MRV technique. In this study, two MRV sequence types are examined regarding their measurement capabilities in highspeed turbulent flows: a conventional MRV sequence based on the popular "4D FLOW" technique, and a newly developed sequence, named "SYNC SPI". Compared to conventional MRV, SYNC SPI is designed for high measurement accuracy, and not for imaging speed, which limits its application to statistically stationary flows. Both sequence types are evaluated in a flow experiment with a converging-diverging nozzle. Time-averaged results are presented for velocities up to 12 m/s at the throat. Supported by Particle Imaging Velocimetry, it is shown that SYNC SPI is capable of acquiring accurate velocity data in these highly turbulent flows. In contrast, the data from the conventional MRV sequence exhibits substantial displacement errors with a maximum displacement of 21 mm. The long acquisition time is the main disadvantage of the SYNC SPI sequence. Therefore, it is examined if undersampling and non-linear reconstruction, known as Compressed Sensing, can be utilized to make data acquisition more efficient. In the presented measurements, Compressed Sensing is successfully applied to shorten the acquisition time by up to 70% with almost no reduction in measurement accuracy.

Measurement in opaque flows: a review of measurement techniques for dispersed multiphase flows

“…It also vastly improves the image quality due to the fact that the signals are now emitted in a shorter time. Common doping agents are MgCl 2 [113], gadolinium-based [114], or CuSO 4 [115].…”

“…There is a continuous drive towards faster acquisition times, by new hardware and acquisition strategies, to better investigate unsteady phenomena such as turbulent flows. For instance, the use of compressed sensing allows a dramatic reduction in the acquisition time [115]. In a recent study, a '4D' data set of a cardiovascular flow could be obtained within 2 min [114].…”

“…Typical resolutions of MRI-derived flow fields are of the order of 0.2-1.0 mm (see, for example, Table 1 of [115]). Note that each voxel here contains one velocity estimate.…”

“…These arise from the fact that the linear gradients and read-out are not simultaneous-the fluid travels between various encoding events. Recent work has highlighted these errors and provides an acquisition sequence that strongly reduces them [115]. There are still open questions, such as the effect of turbulence or the presence of a vapour phase on the accuracy of the velocity results.…”

“…The right-hand side figure shows that in regions with strong gradients the MRV data is smeared out compared to the reference PIV data. Reproduced from John et al [115] with permission ©2020 Springer a b Fig. 10 a An X-ray computed tomography facility to investigate a cavitating venturi.…”