Flow Measurement and Instrumentation 12 (2001) 25–28 www.elsevier.com/locate/flowmeasinst Tidal flow local mean velocities in noisy signals Junfei Yin a,* , Peter M. Lloyd b , Roger A. Falconer c a School of Mechanical Engineering, Cranfield University, Cranfield, Bedfordshire, MK43 0AL, UK b Tessella Support Services plc, Cavendish Avenue, Birchwood Park, Warrington WA3 6FT, UK c Division of Civil Engineering, School of Engineering, The University of Wales, Cardiff CF2 3TB, UK Received 28 March 2000; received in revised form 23 May 2000; accepted 27 May 2000 Abstract A new method was developed to calculate the local mean velocity of unsteady tidal flow from the experimental data using Sontek-ADV and Ultrasonic meters with noisy signals. The method is mainly to analyse a time series of velocity signals in order to obtain a time series of a Moving-Averaged velocity. The paper presents the details of the method. The processing shows acceptable results.  2001 Elsevier Science Ltd. All rights reserved. Keywords: Mean velocity; Unsteady flow; Tidal flow; Signal processing; Flow measurement 1. Introduction In the measurement of unsteady turbulent flows, instantaneous velocity can be expressed as local mean value and turbulent velocity fluctuations. These turbulent velocity fluctuations can be considered as a random pro- cess, described with a Gaussian probability distribution. The local mean values of the flow velocities are often used to visualise the flow field and develop the numerical models. But noises often occur in measurement and reduce the measurement accuracy, In some circum- stances, experimental data have been taken, but some noises are unavoidable. In past experimental studies [1,2], the tidal flows with tidal periods from 120 to 480 s in a large research facility UKCRF were measured using Sontek-ADV and ultrasonic meter-Minilab, which can measure 3-dimensional velocity components at the sam- pling rate of 24 Hz. Seeding has been used for the Son- tek-ADV at a very weak flow in order to keep a reason- able signal-to-noise ratio (SNR), exceeded 3 db, which makes a highly coherent coefficient. In the present cir- cumstances, the noises occur during measurement and the SNR could not be made satisfactory and bad signals * Corresponding author. Tel.: + 44-01234-750-111. E-mail address: j.yin@cranfield.ac.uk (J. Yin). 0955-5986/01/$ - see front matter  2001 Elsevier Science Ltd. All rights reserved. PII:S0955-5986(00)00039-X were sampled, typically shown in Fig. 1 by using ultra- sonic meter-Minilab. Danes [3] has investigated the use of a Moving-Average filtering operation to obtain the local mean velocity with good SNR. The current method was proposed to process those kinds of ‘poor’ signals in order to obtain a reasonable time series of a Moving- Averaged velocity distribution. Fig. 1. The data probability distribution for the Minilab.