ABSTRACT
Particle Image Velocimetry (PIV) is now a well-established, non-intrusive measurement technique which provides instantaneous snapshots of velocity field maps, over a two dimensional region within a flow. Advances in pulsed laser technology, digital image processing and data storage are now allowing PIV to be used in a time-resolved mode where velocity maps can be produced at kHz rates, This important advance makes PIV much more powerful as a tool for the refinement of CFD code predictions, It is now important, however, that the inherent random errors associated with time-resolved PlV measurements are quantified before comparisons are attempted,
One of the major challenges in PIV is increasing the measurement precision, Of particular interest are the errors associated with the post-processing of seeding particle images ranging from digitisation of the images to the location of the correlation peak centre. PIV accuracy has been enhanced by the introduction of sub-pixel interpolation where, typically, a Gaussian profile is fitted to the correlation peak and the centre estimated to an uncertainty of about 0 . 1 pixels, Inherent in a PIV measurement is a random error which occurs due to the random positioning of seeding particle images within a chosen interrogation region, In a time-resolved PIV experiment subsequent snapshots of the same interrogation region will produce changes in the measured velocity over time (which are of interest) and over space, The latter is due to the fact that at each snapshot the particle images are in a different spatial position within the interrogation region. In this way we need to establish the magnitude of the rms variation in velocity due to particle image position, which is a random error, before we can have confidence in deductions made from time varying statistics due to turbulent or unsteady flow. The rms variation in velocity due to particle image position is increased when velocity gradients exist across the interrogation region as described by Adrian ( 1 997).
