**Project Coordinator : **
EDF DTG Grenoble ** Project Manager at GIPSA-lab :**
Cornel IOANA

**Project realized thanks to the support of :** Contrat Industriel (EDF)

**Start date : ** 2011/02/01

**Duration :** 36 mounths

The flow measurement based on acoustic scintillation is an interesting technique to evaluate the velocity of water flowing in a conduit. It uses the natural turbulences embedded in the flow which is measured with a three transmitter-receiver pairs placed on the sides of the water column. The received signals contains are random (because of the turbulence) but they are more or less identical, excepting the delay which correspond to the flowing velocities in vertical and horizontal directions.

The transmitters send pulses and, for each pulse, a measurement of the velocities is done. The integration of the values obtained for pulses conducts to an average value of the flowing velocity. The disposal of such system in a section covering all flow area of interest allows us the evaluation of the spatial distribution of the flowing velocity.

This system, proposed by ASL AQFlow (http://www.aqflow.com/index.htm) has many benefits, with respect of the classical methods based on mills. Among these, we can mention the non-invasively of the method, the lack of any rotating part (high reliability), efficiency, etc.

The system starts to be exploited, by EDF, since 2007. The experience and the results proved the efficiency of the techniques and, in the some time, highlight several operational problems related to the design and to the exploitation of this system.

This document is a synthesis of the encountered problems and it proposes some research directions, aimed to solve the operational problems. These directions are mainly based on advanced signal processing tools that could be interfaced with the current version of the system. The expected result is the efficiency increasing of the system and, in the some time, the exploitation cost reductions. These two operational purposes could conduct to the integration of this system in the norms related to hydraulic equipments (CEI 60041). **Structural problems. Solutions**

These problems are related to the design aspects required by the real implementation of flow measurement in hydro-power systems. Basically, several pairs of three transmitter-receivers systems are disposed on a frame that will be placed in the section that we want to measure. The realization of such system is very expensive while a large number of systems are necessary. Additionally, such approach will not be adapted to sections of different sizes which is currently the case of EDF’s facilities. Alternatively, EDF proposes the system disposal on a lifting carrier that could be setup easily, according to the section specifications. The realization of such system requires the existence of fixing elements such as metallic beams. These necessary elements modify the system behaviour while they are responsible of vibrations and of reflexions (if the transmitter-receiver configuration is close to the beam). The artefacts created by the beam affects the signal processing technique and several improvements are possible.

A first direction consists in modelling the perturbation due to the vibrations created by the beam. The conventional modelling of the vibrations in term of spectrum seems to be insufficient while the vibration spectral structure changes in time because of the time-varying interaction between the beam and the water flowing. Naturally, the vibration evaluation must be adaptive in time which leads to the non-stationary analysis of the beam vibrations. In addition, the physical modelling of the interactions between beam vibration and transmitted signal brings additional information about how the beam effect must be compensated at the receiver level.

Concerning the reflexion of the signal at the beam, this problem could be solved by an equalisation-like technique that will separate the direct propagation from the reflexions.

Depending on the configuration of the hydro-power plant, the measuring sections could or not be preceded by a metallic grid that avoids the intrusion of undesired objects toward the turbine. This constitutive element creates a turbulence that could be useful for the measurement process. In other cases, when this grid is not available, only the natural turbulence is used. In all cases, the modelling of ultrasonic signal’s propagation in the turbulent environment between the transducers can enhance the signal processing algorithms by providing a priori information about the deformations that the transmitted signals might be affected. Indeed, the instantaneous phase deformation could be used to improve the processing of the received signals and such method has been recently developed by the GIPSA-lab teams.**Application-context problems. Solutions**

The application of the scintillation-based measurement system in real context gave rise to several problems that affect the signal processing stage. One of these problems is the lateral motion of the water flow, i.e. the movement of the water in parallel direction with the transmitter-receiver line. This problem decreases the performances of the flow velocity estimation because of the spectral changes between transmitted and received signal, introduced by the Doppler effect. This effect is illustrated in the next figure.

*Figure 1. Illustration of the Doppler effect induced by the lateral motion of the water flow*

As indicated by this figure, the lateral motion of the water is materialized by two velocity vectors (the two pink arrows). In term of spectral representation, this motion can be described as a dilatation/contraction and a translation of the spectrums. A current PhD work is going to be finalised in GIPSA-lab which proposes an analytical description of this effect.

Thus, analysing the spectrum deformation in time (non-stationary signal analysis) could help us estimating the influence of the lateral motion in term of spectral content of the received signals. This estimation can serve to compensate this motion effect in order to bring the spectral content in the useful bandwidth defined by the existing filters.

Another problem that appears in real applications is the presence of air bubbles or of solid objects in the water flow. These objects produce the reflexions of the transmitted waves in other directions than the receiver which is materialized by the attenuation of the received signal. But, as it has been shown by the studies driven by GIPSA-lab, the attenuation of the signal is not total and the received signal, even if it is weak, keeps its phase content. Based on these studies, one possible solution to tackle this problem is to analyse the instantaneous phases corresponding to successive pulses. Once the instantaneous phase identified, the amplitude of the received signal could be corrected by an adaptive amplification. The purpose is to ensure that the amplitude is at the same level before the velocity estimation stage, providing the best condition for the correlation-based method.

In conclusion the time-varying analysis of the instantaneous phase of the successive pulses could help us solving the problems encountered in real application of the system. **Technical limits**

The signal processing algorithms are based on the correlation of the received signals, the transmitted signal being a set of pulses. This principle is illustrated in the figure 2.

*Figure 2. Illustration of the scintillation-based measurement of the flow velocity*

As illustrated by this figure, the estimation of the time-delay between the arrivals at each receiver, delta t, is crucial for an accurate measure of the flow velocity. This estimation can be affected by the variation of the turbulence parameters, by the noise and interferences, of the flow parameters, etc. All these phenomena induce the technical limits of the method. In order to improve the current performances in terms of the estimation accuracy as well as to extend the operational capabilities, the concept of wide-band signal processing can be a central element. This concept consists in transmitting signals with high duration-bandwidth product and the process based on match filtering. The possible processing diagram is illustrated in the figure 3.

*Figure 3. The concept of wide-band processing in the context of scintillation-based measurement of flow velocity *

At the transmitter level, the wide-band concept requires the synthesis and the transmission of a signal having a large duration-bandwidth product. Linear or hyperbolic frequency modulations, Phase shift keying (PSK) or Frequency shift keying (FSK) signals are few examples of signals having a large duration-bandwidth product. The property of such signals, widely used in radar, sonar or communications systems, is the pulse compression which ensures high resolution capabilities as well as the robustness face to perturbation. This property is used at the receiver level where the received signals are correlated with the transmitted ones. The pulse compression conducts to well defined peaks that improve the time-delay estimation (figure 3). The results obtained, between 2008-2010, by GIPSA-lab in collaboration with EDF R&D proved the high potential of this concept.

The adaptation of the wide-band processing concept in the context of the scintillation-based measurement of flow velocity requires the physical modelling of the perturbations of the transmitted signal.

We expect that this new processing strategy can overcome, partially or totally, the problems mentioned in the previous sections.

GIPSA-lab, 11 rue des Mathématiques, Grenoble Campus BP46, F-38402 SAINT MARTIN D'HERES CEDEX - 33 (0)4 76 82 71 31