Forward-scattering particle-image velocimentry

NASA Tech Briefs, Dec 1999 by Ovryn, Ben, Khaydarov, John D

Using only one line of sight, one can measure three-dimensional velocity.

Forward-scattering particle-image velocinetry (FSP[V) is a technique for measuring velocities in microscopic flow fields as thin liquid films. As in other particle-image-velocimetry (PIV) techniques, the velocity of a flow is computed from temporally changing images of suitably illuminated, neutral])" buoyant particles entrained in the flow. Unlike other PIV techniques, (1) particles are back-illuminated with coherent or partially coherent light, (2) a microscope is used to resolve forward scattering of light from the particles, and (3) all three components of velocity, can be measured from only one line of sight, without readjustment of optics.

Figure 1 illustrates a laboratory setup for FSPIV, Illumination is provided by a microscope-illumination source equipped uith a wavelength-selective filter. Partial coherence is obtained by clos, ing the aperture on the illumination source. Microscopic sphericat seed particles of known diameter are suspended in the liquid. A transmitted-light microscope is aimed through the cell back toward the source and is focused to a suitable depth within the cell. Forward-scattered light from a particle in the field of view of the microscope is brought to focus on the microscope image plane to obtain a magnified image of the particle. A charge-coupled device (CCD) is mounted at the image plane to acquire the image. The CCD output is digitized, then processed as described below to extract information about the three-dimensional velocity of the particle.

Coherent illumination is needed because it makes the forward scattering of light (including such effects a: diffi-action from particles and phase changes in microscope lenses) amenable to analvsis. Highly coherent illumination like that provided by a laser results in speckle noise; partially coherent illumination is preferred in this application because it can Nield diffraction patterns adeqUate (bi, analysis while generating much less speckle noise.

Figure 2 schematically illustrates how images are processed to extract velocities. First, a frame of image data is acquired at a known sampling time. Next, the image of each particle is separated from other particles. The scattering pattern for each particle is analyzed to determine the distance of the particle, relative to the focal plane, along the line of sight. This aspect of the analysis involves comparisons of the observed scattering pattern. with scattering patterns that have been computed theoretically and/or recorded experimentally for known positions relative to the focal plane. The comparisons and the interpolations between known positions can be performed by neural-network software that has been trained on the known scattering patterns. Then the component of velocity of each particle along the line of sight is calculated as the distance between two positions determined in this way for two sampling instants, divided by the nine between the instants.

The components of velocity in the plane perpendicular to the line of sight are determined by tracking particle-image centroids as they move across the field of view between successive frames. The tracking of particle-image centroids as a function of time is an established PIV practice; techniques for tracking particle-image centroids as functions of time have been reported in a number of previous articles in NASA Tech Briefs. An added benefit of FSPIV is that the scattering is centro-sym metric so that the centroid is found with high accuracy.

This work was done Ben Ovryn of NYMA, Inc., and. John D. Khaydarov of Ohio Aerospace Institute for Glenn Research Center. For further infin-mation, access the Technical Support Package MSP) free on-line at www.nasatech.com under the Physical Sciences category.

Inquiries concerning rights for the commercial use of this invention should be addressed to NASA Glenn Research Center Commercial Technololgy Office, Attn:Steve Fedor, Mail Stop 4-8, 21000 Brookpark Road, Cleveland, Ohio 44135. Refer to LFW-16403.

Since 1982, Microway's products and technical support have helped users get more done for less money. Starting with the concept that PCs could use more numeric power, we built a product line and customer base that is now worldwide, The motherboards and workstations we design today use Pentiumand Alpha-based processors that deliver 20,000 times the throughput of the 8087s we started with in 1982.

Microway has been building Linux Beowulf clusters since 1997. Our users employ either PVM, or MPI to manage communications between processors in clusters from 8 to 200 Pentium or Alpha CPUs. We design systems using 21264 dual Alpha motherboards, UP2000 dual motherboards, 21164-LX single Alpha motherboards, or Pentium processors for all price points.

Our expertise in parallel processing dates back to the mid-80s when we were Inmos's largest customer for transputers. Following this, Microway built small supercomputers that featured up to 20 Intel i860 RISC processors. The 667-MHz dual Alpha motherboard, which we currently feature in our high-end workstations, delivers 2.6 gigaflops of throughput. If you have an application that is a big-time number-cruncher or a DSP application that needs 64 bits of precision, you should consider our solutions.