Summer Scholarship Programme
Project Summary
[EPCC home] [SSP home] [2001 projects] [2000 projects] [1999 projects] [1998 projects] [1997 projects] [1996 projects] [1995 projects] [1994 projects] [1993 projects]

Particle Image Velocimetry Analysis

Student

Riel Carol, Ecole des mines de Saint-Etienne, France

Supervisors

William Easson, Dept of Mechanical Engineering, University of Edinburgh

Gavin Pringle, EPCC

Particle Image Velocimetry is an experimental technique for fluid flow measurement. It involves three main steps: illuminating a seeded fluid flow with a stroboscopic laser light sheet; photographing the flow using suitable shutter timing; and analysing the digital image. Most flow measurement systems give a single point measurement (eg hot wires, or propeller flowmeters). The unique capability of the PIV system is that the digital photograph may be divided up into a number of small areas, which are independently analysed, and the returned vectors represent a complete record of the velocity over the field of view at an instant in time.

Until recently 'wet' photography was the preferred medium for PIV due to its high resolution. The wet film had to be developed and digitised; and this was a major bottle neck in analysis of flow. However there are now cameras with up to 4Mbyte of pixels available (eg the Kodak Magaplus), and the digitised images are captured in a matter of tenths of seconds. The analysis then takes c. 30 minutes per image on a serial machine.

The purpose of this project is to parallelise the analysis. The programme must read a bitmap, divide it into appropriate small images, and carry out 2D transforms and peak-finding algorithms to provide an array of vector values. These are returned with ancillary data, such as signal to noise ratios. The programme must also read in the parameters for analysing the data, and print out all of the information in an appropriate format.

The bulk of the project is relatively straight forward, in that the problem is naturally adapted to parallel architectures. The procedures carried out in each processor are independent, thought during the partition of data, the programme will have to cope with non-integer division of data fields.

On successful completion of the main part of the project there is scope for extension. The adjacent (x and y) vector values should be compared with each other for similarity, and rogue vectors deleted and replaced. Ideally the programme would also operate as a batch processor.

The final report for this project is available here.
Webpage maintained by mario@epcc.ed.ac.uk