WO2014009366A1

WO2014009366A1 - Method of detecting the blockage of a heat exchanger

Info

Publication number: WO2014009366A1
Application number: PCT/EP2013/064480
Authority: WO
Inventors: Nicolas Paul; Mabrouka EL-GUEDRI; Didier BOLDO; Alexandre Girard; Lorenzo Audibert
Original assignee: Electricite De France
Priority date: 2012-07-09
Filing date: 2013-07-09
Publication date: 2014-01-16
Also published as: FR2993090A1; FR2993090B1

Abstract

The invention relates to a method of detecting various impairments, for example blockage, of a heat exchanger, in which a multifrequency eddy current probe is passed around a tube of said exchanger and said probe is used to measure a measurement signal dependent on the environment in which the probe is located, in which, to identify the parts of this signal corresponding to the traversing of the probe past the spacer plates, - various components of the measurement signal are combined (S1) to determine a detection signal, the combination thus effected being chosen so as to minimize the energy of the detection signal, - detection peaks are detected (S2) on this signal, - from among these peaks (S3) are selected those which exhibit a regular spacing, the peaks thus selected corresponding to the traversing of the probe past the spacer plates.

Description

Method for detecting clogging of a heat exchanger

GENERAL TECHNICAL FIELD

The present invention relates generally to the field of tube inspection of a tube heat exchanger. More specifically, it relates to a method for detecting various degradations, for example clogging, of a tube heat exchanger in which a multifrequency eddy current probe is circulated in a tube of said exchanger and is measured therewith. a measurement signal depending on the environment in which the probe is located.

BACKGROUND OF THE INVENTION

A steam generator is generally composed of a bundle of tubes in which circulates the hot fluid, and around which circulates the fluid to

15 warm up. For example, in the case of a steam generator of a PWR nuclear power plant, the steam generators are heat exchangers that use the energy of the primary circuit resulting from the nuclear reaction to transform the circuit water secondary steam that will power the turbine and thus produce electricity.

The steam generator brings the secondary fluid from a state of liquid water to a vapor state just at the saturation limit, using the heat of the primary water. This circulates in tubes around which the secondary water circulates. The output of the steam generator is the highest point in temperature and pressure of the secondary circuit.

The exchange surface, physically separating the two circuits, thus consists of a tubular bundle, composed of 3500 to 5600 tubes, according to the model, in which the primary water carried at high temperature (320 ° C.) circulates and high pressure (155bars).

These tubes of the steam generator are held by plates

30 spacers arranged generally perpendicular to the tubes which pass through the passages of the spacer plates.

In order to let the fluid that vaporizes, the passages of these spacer plates are folished, that is to say whose shape has lobes around the tubes. As the water changes from the liquid state to the vapor state, it deposits all the contents it contained. If the deposits of matter are in the lobes, they reduce the free passage: it is the clogging, which is therefore the gradual filling, by deposits, holes for the passage of the water / steam mixture.

FIG. 1 schematically illustrates a folate passage in a spacer plate 10, through which a tube 11 passes. The lobes 12a and 12b allow the water to pass through the spacer plate 10 along the tube 11, thus allowing the flow of the water in the steam generator. A deposit 13 is visible at the level of the lobe 12b, clogging said lobe 12b.

Clogging leads to changes in the flow of water in the steam generator, and thus to promote the appearance of excessive vibration of the tubes, as well as induce significant mechanical forces on the internal structures of the generators. steam. This degradation therefore has effects on both the safety and the performance of the installations. It is therefore essential to

15 know the nature and evolution of this degradation.

Currently, the only non-destructive examination system that can access all steam generator tubes is the eddy current axial probe (SAX probe). The eddy currents appear in a conductive material when the magnetic flux is varied in the vicinity. We

20 is thus circulated in a tube of said exchanger a multifrequency eddy current probe and measured therewith a measurement signal depending on the environment in which the probe is located, which can extract information about anomalies in the heat exchanger.

A variation of the magnetic induction, typically by a coil in

Which circulates an alternating current, generates eddy currents, the induced variation of the magnetic field is detected. Typically, the voltage difference generated by the impedance variation of the coil is measured.

The use of the measurement signals of this eddy current probe does not induce an extension of the stopping of the steam generator, since this

The eddy current probe is already used during slice shutdowns, in particular to inspect the integrity of the tubes of the steam generator.

This eddy current probe, initially intended for the detection of damage to the tubes, is also sensitive to clogging. Moreover, the interpretation of this signal is currently carried out manually by specialized operators, which is very long, of the order of about a week of treatment for the analysis of a single steam generator. In addition, the intervention of an operator to take measurements from an analysis software often gives rise to a bias that is difficult to quantify.

An important problem is to extract the useful signal, i.e. the parts of the multifrequency measurement signal corresponding to the passage at the spacer plates 10 of the eddy current probe acquiring said measurement signal. Indeed, the variability of the speed of the eddy current probe, as well as inaccuracies as to the real geometry of the steam generator, makes the extraction of these parts unreliable on the sole basis of an estimate of their position by spatial considerations.

In addition, the measurement signal is not calibrated and is noisy, so that its operation can be difficult.

The document "Automated analysis of a rotating probe multi-frequency eddy

15 current data from steam generator tubes ", by P. Xiang et al., International Journal of Applied Electromagnetics and Mechanics, Vol. 12, p. 151-164, provides an algorithm for automated analysis of multi-frequency eddy probe data. in the context of nuclear power station steam generator tubes.

This analysis is based on the construction, from one-dimensional data, of a two-dimensional image representative of the surface of a tube, taking advantage of the rotation of the probe and interpolating. A Sobel filter is then applied to the resulting image to identify intensity changes in it, and changes above a threshold are

25 selected.

Such an approach requires a rotary probe, and, based on images, does not allow precise detection of parts of the signal corresponding to the passage of the probe at the spacer plates, due in particular to the presence of noise, which can be important.

The document "Estimation of mixing parameters for cancellation of discretized eddy current using time and frequency domain techniques", by CK Sword et al., Journal of Nondestructive Evaluation, Vol. 5, No. 1, pages 27-35, describes a method in which the components of a signal of an eddy current probe are combined for the purpose of suppressing carrier signals from unwanted plate. The method in question therefore does not identify the parts of the signal corresponding to the passage of the probe at the spacer plates.

5 PRESENTATION OF THE INVENTION

The purpose of the invention is to overcome at least one of these defects, all of them preferentially, by proposing a method that automates the extraction of the portions of the signal emitted by a SAX probe corresponding to the passage of the spacer plates of a tube heat exchanger. preferentially a PWR nuclear steam generator, without making use of considerations on the geometry of the steam generator, while providing a calibrated and denoised signal.

For this purpose, there is provided a method for detecting spacer plates of a heat exchanger, in which a multifrequency eddy current probe is circulated in a tube of said exchanger and a measurement signal according to the invention is measured therewith. the environment in which the probe is located.

To identify the parts of this signal corresponding to the passage of the probe at the level of the spacer plates,

Different components of the measurement signal are combined to determine a detection signal, the combination thus performed being chosen to minimize the energy of the detection signal,

detecting peaks on this signal,

from these peaks, those which have a regular spacing are selected, the selected peaks corresponding to the passage of the probe at the level of the spacer plates.

The invention is advantageously completed by the following features, taken alone or in any of their technically possible combination:

The combination of the different components of the measurement signal is an adaptive linear combination, the values of the weighting coefficients used for this combination being determined, for a sample of the detection signal, by minimizing the variance of the detection signal on a window observing the measurement signal around this sample; different components of the measurement signal combined to determine the detection signal are real and imaginary components of the measurement signal;

different components of the measurement signal combined to determine the detection signal are components of different frequencies;

for detecting detection peaks on the detection signal, the detection signal is compared with a detection threshold, said threshold being a function of the minimum value that the standard deviation of the detection signal takes on a part of the corresponding signal. a number of samples smaller than the number of theoretical samples between two consecutive spacer plates;

prior to the step of selecting the peaks, the detected peaks are limited to the peaks corresponding to a local maximum of the detection signal over a range of the detection signal corresponding to a number of samples on each side of the local maximum less than or equal to the minimum of the following three values:

15 - 0.5 times a theoretical number of samples between two spacer plates,

0.8 times a theoretical number of samples between the spacer plate of the hot leg closest to the cold leg of said heat exchanger and the spacer plate of the cold leg closest to the hot leg of said heat exchanger,

20 - 0.8 times a theoretical number of samples between a first of the spacer plates and another previous plate in the direction of flow of the probe.

to select the peaks of the detection signal having a regular spacing and corresponding to the passage of the spacer plates, it is determined at

At least one subset of peaks minimizing the difference between the number of samples of the detection signal between two successive peaks of said subset and a median value of the number of samples of the detection signal between two successive peaks;

the measurement signal corresponds to the measurement of an eddy current detector probe acquiring said measurement signal during the return of the probe in a back-and-forth displacement of said probe in a tube of a heat exchanger;

the method comprises a preliminary calibration of the measurement signal, comprising:

the identification of the parts of the measurement signal corresponding to known defects of the tube of the heat exchanger, the estimation of the phase and the amplitude of the parts of the measurement signal corresponding to the known defects,

determining a transformation to be applied to the measurement signal from said parts of the measurement signal corresponding to the known defects for

5 calibrate the measurement signal,

- the application of the transformation to the measurement signal;

the method further comprises a subsequent verification step in which the positions of the peaks of the detection signal corresponding to the passage of the spacer plates are compared with known data on the geometry of the heat exchanger;

the method comprises a final verification step in which the opposite of the detection signal is taken, the steps of detecting and selecting the positions of the peaks of the detection signal corresponding to the passage of the spacer plates are applied and verifying that the deviations between the positions of the peaks on

The detection signal and on its opposite correspond to the edge-to-edge distance of a spacer plate.

The invention also relates to a computer program product comprising program code instructions for performing process processing steps according to one of the preceding claims, when said

20 program is run on a computer.

PRESENTATION OF FIGURES

Other features, objects and advantages of the invention will emerge from the description which follows, which is purely illustrative and nonlimiting, and which should be read with reference to the appended drawings among which:

- Figure 1, already commented schematically illustrates a folate passage in a spacer plate, in which passes a tube, according to a current configuration of a steam generator;

FIG. 2 is a block diagram illustrating the calibration of the measurement signal in the context of the invention;

FIG. 3 is a schematic diagram of the method according to the invention.

DETAILED DESCRIPTION A multifrequency eddy current probe is circulated in a tube 11 of said exchanger and a measurement signal is measured therewith which is a function of the environment in which the probe is located. The measurement signal is therefore acquired by the eddy current axial probe during its displacement in the tube 11 of the steam generator. This measurement signal is multifrequency, and is in complex form, with real and imaginary components. Within the framework of this description, we will consider a signal with complex components according to three frequencies coming from three acquisition channels, of which two in differential mode denoted z ^ n] and z ₃ [n] and one in absolute mode noted z _A [not],. The frequencies are for example in the range 100-600 kHz.

Since the probe makes a round trip in the tube, the signals corresponding to the phase of displacement where the probe is pulled by the mechanism that moves it, typically during its return, will preferably be selected. Indeed, it results in a better mechanical stability of the probe, and consequently a higher speed

15 regular of it.

The selection of the part of the signal of the probe corresponding to the return of the probe is based on the detection of a significant drop in the voltage measured at the moment when the probe switches to traverse the tube in the direction of the return. The selection of a measurement signal limited to one direction of the probe allows

It also limits the amount of data to be backed up and processed.

Given the sensitivity of the acquisition of eddy current signals to the acquisition conditions, the tilting of the probe and in order to standardize the analyzed signals, a preliminary calibration step is preferably implemented.

It is a question of calibrating the signals with respect to reference signals, called calibration signals. These correspond to defects implanted on the calibration tube, and whose characteristics (amplitude and phase) are known. The calibration (step S0) is carried out in the same manner regardless of the acquisition frequency considered, with, with reference to FIG. 2, for each

Controlled tube:

identification (step S20) of the parts of the measurement signal corresponding to known defects of the tube of the heat exchanger,

estimation (step S21) of the phase and the amplitude of the parts of the measurement signal corresponding to the known defects, determination (step S22) of a transformation to be applied to the measurement signal from said parts of the measurement signal corresponding to the known defects for calibrating the measurement signal,

- application (step S23) of the transformation to the measurement signal.

5 Let z be _gross (f) = x + jy (where j is the imaginary number) from an acquisition at frequency f, and

the complex impedance measured for a given calibration defect at the same frequency.

A calibration defect is characterized by two parameters A ₀ and θ ₀ which are respectively the amplitude and the phase of the vector max-min (vector i od 'maximum amplitude on the representation according to Lissajous). These are the characteristics that the calibration defect must have after calibration.

Before calibration, these two parameters are respectively worth A and Θ. The calibration of the signal z then consists of a rotation of angle δφ = θ ₀ -θ and a homothety of parameter h = A ₀ / A as described by the following equation:

15 = hΣoe ^{} 5} ^

In fact, the geometrical configuration of the tube or the heat exchanger is not used to detect the zone of the fault, but only to the characteristics of the signal (phase and amplitude). This provides an independent calibration of the probe handling or signal sampling problems, which may bias the judgment of the position of the part of the signal corresponding to the defect.

Once the measurement signal has been calibrated, it remains to determine the parts of the multifrequency measurement signal corresponding to the passage at the level of the spacer plates 10 of the eddy current detector probe acquiring said measurement signal in the tube of the heat exchanger .

The measurement signal is very noisy and can not be used directly to detect the spacer plates 10. In order to facilitate this detection, a detection signal, denoted d [n], is constructed by linear combination of components of the frequency measurement signal. different to minimize the energy of the detection signal. Thus, the detection signal then takes a minimum value between the spacer plates 10, while taking a large value when passing the spacer plates 10.

The signals available at the input of the algorithm are the three complex calibrated signals z _A [n], z ^ n] and z ₃ [n] corresponding to the return of the probe in the tube. Only the signals corresponding to the differential mode (z1 and z3), more sensitive to the passage of the probe by a plate, are used in this part. We denote x ₍ [n] and [n] respectively the real and imaginary components of the signal ¾ [n].

At first, all the available measurements are filtered by a high-pass filter in order to eliminate the small variations of the signals due to the horizontal movement of the probe in the tube. For each available signal, a high-pass filter (for example a Butterworth filter) is applied, for example a reduced cutoff frequency of 0.01. The signals resulting from the filtering of the signals z ₁ and z ₃ are denoted z _1f and z _3f . The real and imaginary parts of these signals are therefore noted:

zi _f [n] = xi _f [n] + jy _1f [n]

This first filtering operation makes it possible to construct signals having a baseline of constant average and zero. To then build a detection signal that is as small as possible between the plates, we start by looking for an observation window expressed in number of signal samples which limits the impact of the passage of a spacer plate 10.

For this purpose, information is available on the average velocity v of the probe, as well as on the sampling frequency f _e . Knowing the speed of the probe and its sampling frequency, each distance can be converted into a number of signal samples according to the

has formula:

M - distance x

^JV samples

V

For example, an observation window corresponding to a length of M samples smaller than the number of theoretical samples 25 may be _obtained between two successive spacer plates Ν _ρθ , _ρθ , for example

M = ^Npe ' ^pe

2

This search is performed for example on the signal X _3f .x _3f having a high variance when passing the probe at a spacer plate 10. It is sufficient to find such a window to find the portion of the signal length M where the variance of x _3f is minimal. Noting m ₀ the first index of 30 this window we have:

However, the signal portions corresponding to the passage through the spacer plates 10 are not known, since this is precisely what we seek to determine. Thus, if there were to be a spacer plate 10 in the observation window, this could greatly impact the rest of the process.

In order to limit the impact of the passage of a spacer plate 10 in the calculation, a sufficiently large window is preferably selected. For example, the observation window may correspond to a theoretical number of samples of the measurement signal representing the distance between consecutive spacer plates along the tube 11, that is to say of size Μ = Ν _ρθ , _ρθ .

In order to limit the impact of the passage by a spacer plate 10, it is also possible to choose the set of signal indices n such that | x _3f [n] | is below a certain threshold. This threshold can be, for example, the standard deviation of the signal x _3f .

Different components of the measurement signal are then combined to determine a detection signal, the combination thus effected being chosen to minimize the energy of the detection signal. The combination of the different components of the measurement signal is an adaptive linear combination, the values of the weighting coefficients used for this combination being determined, for a sample of the detection signal, by minimizing the variance of the detection signal on the window of observation of the measurement signal around this sample.

Preferably, the different components of the measurement signal combined to determine the detection signal are real and imaginary components of the measurement signal. Likewise, different components of the measurement signal combined to determine the detection signal are different frequency components.

We thus look for weighting coefficients α, β and γ such that x _3f [n] _* ax _1f [n] + .y _3f [n] + yy _1f [n]

in order to build the detection signal d [n] defined by

] = X ₃ f [n] - ax _1f [n] - .y _3f [n] - yy _1f [n]

In addition, to manage the strong non-stationarity of the measurement signal components, the coefficients α, β and γ are updated for each sample of signal. More precisely, the triplet {α [η], β [η], γ [η]} chosen at an instant n is the one that minimizes the power of the reconstruction error on a signal window centered on the nth sample: [k] - "if [k] - βγ ₃ f [k] - yyi / [k]

The optimization of {α [η], β [η], γ [η]} is performed using the least squares algorithm. Note that we can change the number of coefficients.

A detection signal d [n] is thus obtained by adaptive linear combination of real and imaginary components of the measurement signal.

Such a construction of the detection signal has several advantages:

an automatic estimation of the optimal coefficient of combination of the frequencies (i.e. search for the coefficient minimizing the energy of the off-plate signal);

the combination can be done between all the components (real and imaginary parts) of the different frequencies;

the combination coefficients are not constant along the tube, but vary in order to adapt to the tube. Thus, the variations of the state of the tube are managed.

Once the detection signal has been constructed, it remains to detect peaks of the detection signal likely to correspond to the passage of spacer plates 10, compared with a detection threshold. Indeed, the signal of

The detection thus created contains different peaks, which correspond to the passage of the probe by the different plates, and not only the spacer plates 10, and other structural elements.

These signal peaks can be identified by keeping only the signal peaks greater than a detection threshold s.

Preferably, the detection threshold for detecting the peaks of the detection signal is a function of the minimum value that the standard deviation σ of the detection signal has on a part of the signal corresponding to a number of samples less than the number of samples. theoretical samples between two consecutive spacer plates 10.

The standard deviation σ can be determined as follows:

The detection threshold must then be a compromise between the risk of false detections (a threshold that is too low results in the detection of numerous peaks that do not correspond to the passage of a spacer plate 10by the probe) and the risk of not detecting a spacer plate. 10. It may for example be between 5 and 15 σ, preferably 10 σ.

It is possible that the passage of the probe through a spacer plate will result in several passages of the signal above the detection threshold. To keep only one peak per plate, then to identify the peaks corresponding to the spacer plates 10, the following test is performed on the successive pairs of peaks: if n1 and n1 are two indices of successive peaks, we say that ni is a "secondary peak" of n1 if d [n1]> d [n2] and if the number of samples between n1 and n1 is much smaller than the theoretical minimum distance between two plate peaks.

Thus, prior to the step of selecting the peaks, the peaks detected are limited to the peaks corresponding to a local maximum of the detection signal over a range of the detection signal corresponding to a number of samples on each side of the maximum local less than or equal to at least the following three values:

0.5 times a theoretical number of samples between two spacer plates 0.8 times a theoretical number of samples between the spacer plate of the hot branch closest to the cold branch of said heat exchanger and the spacer plate of the cold branch the closest to the hot branch of said heat exchanger,

0.8 times a theoretical number of samples between a first of the spacer plates and another previous plate in the direction of flow of the probe.

The secondary peaks being suppressed, it then remains to select peaks of the detection signal corresponding to the passage of the spacer plates 10 by a determination of the peaks of the signal having a regular spacing.

The number of theoretical samples between two plates can be calculated from a knowledge of the distance between two spacer plates. Ideally, if the probe has a known and constant velocity v, then the number of samples between two plates is exactly proportional to the distance between plates:

distance between plates xf _e

N _pe , pe ^{= ~}

In this ideal case, it would be sufficient to identify the indices of the corresponding samples to keep only the peaks whose indices are exactly spaced from N _P e, eg. In practice, the speed of the probe is not exactly constant or precisely known and can vary along the tube. The difference between two peaks corresponding to spacer plates 10 is therefore not exactly equal to N _P e, p _e and it may be different from one pair of spacer plates 10 to another. This average gap between spacer plates 10 can however be estimated.

To do this, the peaks of the detection signal corresponding to the passage of the spacer plates 10 are selected by selecting at least one subset of peaks minimizing the difference between the number of samples of the detection signal between two peaks. successive of said subset and a median value of the number of samples of the detection signal between two peaks

15 successive.

A tube of a heat exchanger extends along the latter through two parts, generally called cold leg and hot leg, depending on the direction of flow of the fluids carrying out the heat exchange, separated by a plate-free structure spacer 10 designated as the hanger. Therefore, such a structure breaks the regularity of spacing of the spacer plates. It suffices to call the following method for one branch, then for the other, to select all the peaks corresponding to the spacer plates 10.

Let i _{k be} the index of the k-th peak detected, A [k] = i _{k + 1} -i _k the difference between the successive peaks k and k + 1, and N _pe , eg, the number of k Average samples between two spacer plates 10. As the spacer plates are the majority plates along the tube, the majority of the values taken by Δ [k] is located around N _pe , eg, real. An estimate of N _pe , pe, real is thus provided by the median value of the set of values taken by A [k]:

Npe.pe.réi = median {A [k]}

It then remains to find, among the set I of peak indices, the subset lp _ic of indices corresponding to the number of spacer plates NbPe, such that the gaps between two successive indices are always roughly equal to

N pe, pe, real ·

The indices of the samples of the signal corresponding to the passage of the spacer plates 10 are then obtained.

Once the set of indices corresponding to the passage of the probe through the spacer plates 10 has been estimated, it is then possible to proceed to a subsequent verification step in which the positions of the peaks of the detection signal corresponding to the passage of the spacer plates 10 are compared. with known data on the geometry of the heat exchanger. Two criteria are used: the length of the spacer plates 10 and the length of the hanger.

The first verification criterion concerns the length of the hanger: for each tube, the number of samples between the last plate of the hot branch and the last plate of the cold branch is compared with the known distance according to the plans. A margin is allowed, to take into account the uncertainty on the speed of the probe.

The second validation criterion makes it possible to check the length of all the detected plates. The detection algorithm having made it possible to estimate the indices of the spacer plates 10 uses the positive peaks of the detection signal. However the passage of the probe by a spacer plate 10 generates on the signal at least one positive peak and one negative peak (the two peaks correspond to the passages of the probe by the plate edges).

A new estimate of the indexes of the spacer plates 10 can thus be obtained by detecting this time the negative peaks of the detection signal, without taking into account the first peaks detected. In practice, it is sufficient to recall the peak detection algorithm by taking the opposite of the detection signal (d [n] becomes -d [n]). The method may thus comprise a final verification step in which the opposite of the detection signal is taken, the detection S2 and S3 selection steps are applied to the positions of the peaks of the detection signal corresponding to the passage of the spacer plates 10 and verifies that the differences between the positions of the peaks on the detection signal and on its opposite correspond to the distance edge-to-edge of a spacer plate 10.

We then have, for each branch, two sets of estimated indices. If these two sets of indices are correct, the difference between the two estimates of the same spacer plate 10 must be of the order of magnitude of the size of a spacer plate 10. Taking into account the inaccuracies on the speed of the probe and plate edge effects, a margin is allowed to validate the estimate.

Once the identification of the parts of this signal corresponding to the passage of the probe at the level of the spacer plates 10, it suffices to analyze the said parts of the signal, for example by comparing the variation of the intensity or the power of the detection or measurement signal with a reference model and / or a calibrated model, to determine an anomaly at the passage of a spacer plate 10 may for example correspond to a clogging.

The invention also relates to a computer program product comprising program code instructions recorded on a medium usable in a computer for performing process processing steps according to one of the preceding claims, when said program is executed. on a computer.

In fact, the measurement signal is transmitted from the eddy current probe to a memory for storage therein for processing. This processing of the measurement signal to which the present invention relates is implemented by a processing unit provided with a computer, typically a computer provided with display and communication means, by which it acquires the measurement signal and transmits the results of the implementation of the process.

Claims

claims

Process for detecting degradations of a tube heat exchanger, in which a multifrequency eddy current probe is circulated in a tube (1 1) of said exchanger and a measurement signal which is function of the environment is measured therewith in which the probe is located, characterized in that to identify the parts of this signal corresponding to the passage of the probe at the level of spacer plates (10),

- combining (S1) different components of the measurement signal to determine a detection signal, the combination thus performed being chosen to minimize the energy of the detection signal,

detecting (S2) on this signal detection peaks,

between these peaks, those which have a regular spacing are selected (S3), the selected peaks corresponding to the passage of the probe at the level of the spacer plates (10).

A method according to claim 1, wherein the combination of the different components of the measurement signal is an adaptive linear combination, the values of the weighting coefficients used for this combination being determined, for a sample of the detection signal, by minimizing the variance of the detection signal on an observation window of the measurement signal around this sample.

Method according to one of the preceding claims, wherein different components of the measuring signal combined to determine the detection signal are real and imaginary components of the measurement signal.

Method according to one of the preceding claims, in which different components of the measuring signal combined to determine the detection signal are components of different frequencies.

Method according to one of the preceding claims, wherein, in order to detect detection peaks on the detection signal, the detection signal is compared with a detection threshold, said threshold being a function of the the minimum value that the standard deviation of the detection signal takes on a portion of the signal corresponding to a number of samples smaller than the number of theoretical samples between two consecutive spacer plates (10).

Method according to one of the preceding claims, wherein prior to the peak selection step (S3), the detected peaks are limited to the peaks corresponding to a local maximum of the detection signal over a range of the detection signal corresponding to a number of samples on each side of the local maximum less than or equal to the minimum of the following three values:

0.5 times a theoretical number of samples between two spacer plates,

Method according to one of the preceding claims, in which to select the peaks of the detection signal having a regular spacing and corresponding to the passage of the spacer plates (10), at least one subset of peaks minimizing the difference between the number is determined. samples of the detection signal between two successive peaks of said subset and a median value of the number of samples of the detection signal between two successive peaks.

Method according to one of the preceding claims, in which the measurement signal corresponds to the measurement of an eddy current detector probe acquiring said measurement signal during the return of the probe in a back-and-forth displacement of said probe in the tube (1 1) of a heat exchanger.

9. Method according to one of the preceding claims, comprising a preliminary calibration (S0) of the measurement signal, comprising:

the identification (S20) of the parts of the measurement signal corresponding to known defects of the tube (11) of the heat exchanger,

the estimation (S21) of the phase and the amplitude of the parts of the measurement signal corresponding to the known defects,

determining (S22) a transformation to be applied to the measurement signal from said parts of the measurement signal corresponding to the known defects for calibrating the measurement signal,

the application (S23) of the transformation to the measurement signal.

A method according to any one of the preceding claims, further comprising a subsequent verification step in which the positions of the peaks of the detection signal corresponding to the passage of the spacer plates (10) are compared with known data on the geometry of the 'heat exchanger.

1 1. Method according to any one of the preceding claims, comprising a final verification step (S4) in which the opposite of the detection signal is taken, the detection (S2) and selection (S3) steps of the peak positions are applied. the detection signal corresponding to the passage of the spacer plates (10) and it is verified that the differences between the positions of the peaks on the detection signal and on its opposite correspond to the edge-to-edge distance of a spacer plate (10).

A computer program product comprising program code instructions recorded on a usable medium in a computer for performing process processing steps according to one of the preceding claims, when said program is executed on a computer.