Corrosion control system for metallic structures using ultrasonic guided waves

Abstract

A system for monitoring the corrosion of a metallic structure using guided ultrasonic waves, employing a multi-element magnetostrictive probe forming an array of independent transmitting / receiving transducers on either side of an area to be inspected in the structure, comprising: switching means (70) for sequentially controlling the excitation of each transmitting transducer in the transmitting transducer array (Ei = 1, ..., Ne), and acquisition means (80) for sequentially controlling the acquisition of the detection signal delivered by each receiving transducer in the receiving transducer array (Rj = 1, ...Nr), in response to the signal generated by each transmitting transducer, means for recording a detection matrix corresponding to all acquired detection signals, means for processing said matrix by a tomographic reconstruction algorithm intended to provide a distribution of the thickness variation at any point in the area to be inspected. Figure to be published with the abbreviation: Fig. 5.

Description

Title of the invention: System for controlling corrosion in metallic structures using ultrasonic guided waves

[0001] The invention relates to a non-destructive testing system, using guided ultrasonic waves, designed for detecting defects, particularly corrosion defects, in metallic structures. The invention also relates to a non-destructive testing method implemented by this system.

[0002] The invention is preferentially applicable to the corrosion control of thin-walled metallic structures, such as pipelines, pipes, tanks, pressure vessels, etc., and, more generally, all industrial structures undergoing degradation resulting from their operating conditions. In operation, these structures can be subjected to significant mechanical stresses and chemical attacks, which promote the development of defects. Given the need to keep these structures operational for as long as possible to improve their profitability, safety issues constitute a major industrial challenge in many sectors, such as energy, air transport, petrochemicals, and nuclear power.Thus, non-destructive testing (NDT) methods play an important role in addressing these safety challenges, particularly those employing ultrasonic techniques, notably for assessing corrosion defects in thin-walled metallic structures, which is one of the most damaging damage phenomena for the aforementioned sectors.

[0003] The corrosion of metallic structures is a complex phenomenon that is highly dependent on operating and environmental conditions. It can occur in various forms such as pitting, fatigue cracking, or thickness loss. We are particularly interested here in this last type of corrosion defect.

[0004] Among non-destructive ultrasonic testing techniques, those using guided ultrasonic waves appear particularly promising, not only for the detection and characterization of corrosion defects, but more generally for the implementation of Structural Health Monitoring (SHM) systems. Indeed, due to its propensity to propagate over long distances with little dissipation, this type of wave appears particularly well-suited for inspecting installations made up of large, thin structures such as tubes or plates, or those with hard-to-reach areas. Industrial inspections using guided waves thus most often focus on Plate geometries, preferably of fixed thickness, or cylindrical geometries, preferably of invariable cross-section, define the geometry of the waveguide in which the guided waves propagate. For plate geometries, there are two different families of guided waves depending on their polarization: SH (horizontal Shear) modes and Lamb modes. The first family of modes corresponds to horizontal transverse waves whose polarization is perpendicular to the direction of propagation. The second family of modes is divided into symmetric (S) and antisymmetric (A) modes, according to the symmetries of the displacement field within the thickness of the waveguide. We distinguish between symmetric (Sn) and antisymmetric (An) modes, whose deformations are respectively symmetric and antisymmetric within the thickness of the waveguide. The index n is an integer that corresponds to the order of the propagation mode.For cylindrical geometries, guided modes propagate along the principal axis of the structure and throughout the thickness of the cross-section. These modes can be either torsional, compressional, or bending modes.

[0005] During a structure inspection by guided waves, the inspection consists of emitting an ultrasonic wave along a given direction via transducers on board the structure, the ultrasonic wave propagating and interacting with any disturbances present on the structure being inspected.

[0006] The principle of using magnetostrictively guided waves to control a structure such as a conduit is well known. It relies on the use of magnetostrictive transducers, preferably with a magnetostrictive material patch as an interface material to generate the guided waves in the structure. The principle of the magnetostrictive transducer is based on the magnetostriction effect. Magnetostriction is the property of certain materials to undergo a geometric change (contraction, expansion, bending, torsion, etc.) when subjected to the influence of a magnetic field. Metallic alloys, and in particular ferromagnetic compounds, are magnetostrictive materials.

[0007] The advantage of using magnetostrictive probes as an alternative to the piezoelectric probes frequently used lies in the fact that they offer the choice of suitable guided wave modes and in their ease of integration due to their small size compared to piezoelectric transducers for the type of application targeted by the present invention.

[0008] However, a number of difficulties were observed when testing the performance of existing equipment on the market. In particular, a lack of performance in sensitivity and fault resolution was noted in the results for equipment using magnetostrictive transducers.

[0009] Furthermore, the use of ultrasonic guided waves remains complicated to Implementation, particularly in the interpretation of results, is crucial. In particular, due to the multimodal nature of guided waves, the measured signal is most often composed of several signals emanating from the contributions of different propagation modes. Adding to this multimodal nature is the phenomenon of guided mode dispersion, which results in a temporal spreading of the signals from the different modes depending on the distance between the transmitter and the disturbance. The combination of these two phenomena gives rise to complex time-domain signals, the interpretation of which is a major challenge. Regardless of the characteristics of the controlled structure, the complexity of these signals depends on several parameters, the main ones being the excitation signal used (frequency and bandwidth), the transducer(s) used for transmission and reception, and the choice of the emitted mode(s).

[0010] In terms of analysis and processing of the signals measured by the transducers, the application of an inverse transformation on the set of acquisitions makes it possible to reconstruct - that is to say by forming an image - the properties of the medium and to determine the characteristics of the defects.

[0011] In particular, ultrasonic guided wave tomography offers strong potential for characterizing corrosion defects in thin-walled structures, such as plates, pipes, etc. Several reconstruction methods exist in the literature.

[0012] Among these, we can mention the probabilistic method known as RAPID (an acronym for the English expression "Reconstruction Algorithm for Probabilistic Inspection of Damage"), which is described in patent document CN102928511. This method proves advantageous for structural health monitoring by proposing the use of piezoelectric transducers permanently mounted on the structure to be monitored. It offers great flexibility in the choice of the transducer array geometry, simplicity of implementation, and allows the detection and localization of defects with few sensors. However, it suffers from artifacts due to the crude summation of the collected signals and, above all, it does not allow for direct quantification of the thickness loss, to the detriment of the accuracy of characterizing corrosion defects in the monitored structure.

[0013] Therefore, there is a need for a non-destructive testing system using ultrasonic guided waves produced by magnetostrictive emission, which allows for better selectivity of the guided mode adapted to the corrosion defects to be characterized, better spatial resolution, and a small footprint for better integration into the structure being inspected. There is also a need to improve reconstruction techniques, allowing for greater ease of interpretation, particularly for more precise characterization of corrosion defects, in terms of depth. of extent and shape.

[0014] To this end, the invention relates, according to a first object, to a corrosion control system for a metallic structure using guided ultrasonic waves produced by magnetostrictive effect, comprising, on either side of an area to be inspected of said structure: - at least one transmitting probe comprising at least one strip of a magnetostrictive material, this strip having two opposite faces, one of which is directed inwards and adapted to be made integral with the structure to be monitored, and the other is directed outwards and on which an excitation coil is placed to create in the vicinity of the strip a variable magnetic field in relation to an electrical excitation signal applied to the terminals of the coil, and - at least one receiving probe comprising at least one strip of a magnetostrictive material, this strip having two opposite faces, one of which is directed inwards and is adapted to be made integral with the structure to be monitored, and the other is directed outwards and on which is placed a detection coil to deliver an electrical detection signal related to the variation of the magnetic field in the strip as a function of a received ultrasonic wave, said system being characterized in that said excitation / detection coil comprises a plurality of solenoids arranged on the strip, forming at least one network of independent transmit / receive transducers, said system comprising: - means of excitation by applying an alternating excitation voltage to an emitting transducer at a predetermined excitation frequency, - switching means for sequentially controlling, in time, the excitation of each transmitting transducer in the transmitting transducer network by said excitation means, - Acquisition means to sequentially control in time the acquisition of the detection signal delivered by each receiving transducer of the receiving transducer network, in response to the signal generated by each transmitting transducer, - means for recording a detection matrix corresponding to all acquired detection signals, and - means for processing the detection matrix by means of a tomographic reconstruction algorithm designed to implement a reconstruction of a two- or three-dimensional map of the area to be inspected, providing a distribution of the thickness variation of the structure at any point in the area to be inspected.

[0015] Advantageously, said solenoids are rectangular in shape, the width of which is defined so as to generate and detect, for a selected excitation frequency, a specific guided mode adapted to corrosion control, comprising at least the SH mode.

[0016] Advantageously, the transmitting transducer array and the receiving transducer array are arranged so as to at least partially surround the area to be inspected.

[0017] According to one embodiment, the receiving transducer network is arranged substantially parallel to the transmitting transducer network.

[0018] According to one embodiment, said structure being a tubular structure, the transmitting and receiving transducer networks extend along at least part of the circumference of the tubular conduit, in a plane transverse to the axis of the tubular structure.

[0019] As an alternative or in combination with the previous embodiment, the transmitting and receiving transducer networks extend parallel to the axis of the tubular structure, diametrically opposite each other, over the entire length of the area to be inspected.

[0020] The invention also relates to a method for controlling corrosion in a metallic structure by guided ultrasonic waves produced by magnetostrictive effect, comprising the following steps: - A control system is provided as described above, - the excitation of each transducer in the emission transducer network is sequentially controlled in time by said excitation means, - The acquisition of the detection signal delivered by each transducer in the receiving transducer network is sequentially controlled in time, in response to the signal generated by each transmitting transducer. - the detection matrix corresponding to the set of acquired detection signals is recorded, - the said detection matrix is ​​exploited by means of a suitable tomographic reconstruction algorithm to reconstruct a two or three dimension map of the area to be inspected, providing a distribution of the variation in thickness of the structure at every point of the area to be inspected.

[0021] Advantageously, a prior selection of specific guided mode and excitation frequency is made, adapted to corrosion control.

[0022] Advantageously, the time-of-flight variation for said specific guided mode related to the thickness variation along the wave path between each pair of transmit / receive transducers is determined, the average velocity variation for said guided mode is deduced from this, and the corresponding average thickness variation between each pair of transmit / receive transducers is determined by projecting this average velocity variation onto the dispersion curve of said specific guided mode expressing the group velocity as a function of the excitation frequency-thickness product of said structure, the result of determining the average thickness variation along the guided wave paths for each pair of transmit / receive transducers constituting a projection matrix.

[0023] Advantageously, said mapping is obtained by backprojecting the set of projections of the projection matrix in the form of an ellipse connecting the transmitting / receiving transducers of each pair of transmitting / receiving transducers and by summing the contributions of each pair of transmitting / receiving transducers.

[0024] Advantageously, the back projection step is preceded by a projection filtering step.

[0025] Preferably, the filtering step is carried out for each projection in the Fourier domain via multiplication by a filter function, for example of the ramp filter type.

[0026] Other features and advantages of the invention will become clearer from the following description, which is by way of example and not limitation, with reference to the accompanying drawings in which:

[0027] [Fig.l] schematically illustrates a single-element magnetostrictive probe;

[0028] [Fig.2] illustrates an example of the operation of a generation and reception of ultrasonic guided waves by multi-element magnetostrictive transmit / receive probes as implemented in the control system of the invention;

[0029] [Fig.3] is a schematic representation of the detection matrix constructed according to the principle of multi-element acquisition;

[0030] [Fig.4] is an example of a diagram of the dispersion curves expressed by the variation of group speed as a function of the frequency-thickness product for a steel plate, illustrating the principle of selectivity of the guided mode of the probes;

[0031] [Fig.5] is a synoptic diagram of the control system for monitoring corrosion in a metallic structure, following an example embodiment where said metallic structure is a rectangular steel plate type structure;

[0032] [Fig.6] describes a flowchart of the operation of the control system of the [Fig.5];

[0033] [Fig.7] is a diagram illustrating an example of rectilinear wavefront trajectories between pairs of transmit / receive transducers of the control system of the invention;

[0034] [Fig.8] is a diagram of a steel plate illustrating the principle of determining the thickness variation in the case of a single transmit / receive transducer;

[0035] [Fig.9] is a diagram of the dispersion curves of guided waves in the plate in steel of the [Fig.8], expressed by the group speed as a function of the product plate thickness frequency;

[0036] [Fig. 10] describes a detailed flowchart of the post-processing algorithm implemented in the control system of the invention;

[0037] [Fig. 11] schematically illustrates the plate of [Fig.8], instrumented with two parallel arrays of transmitting / receiving transducers and an example of projection obtained resulting from the calculation of the average thickness variation for each receiving transducer, for a given excited transmitting transducer;

[0038] [Fig. 12] is a diagram illustrating the elliptic distribution function between a transmitting transducer and a receiving transducer, used for backprojection

[0039] [Fig. 13] is a diagram illustrating the principle of reconstruction of the defect by the sum of the elliptical rays;

[0040] [Fig. 14a] is a diagram illustrating the results of the tomographic reconstruction in the case of the example of [Fig. 11];

[0041] [Fig. 14b] is a cross-sectional view along the y-axis, and passing through the center of the defect, of the map illustrated in [Fig. 14a].

[0042] A specific embodiment of a single-element magnetostrictive probe for emitting / receiving ultrasonic guided waves is described with reference to [Fig.1].

[0043] The probe 1 consists of a patch 2 formed from a strip of magnetostrictive material. This strip has two opposite faces, one of which is directed inwards and adapted to be bonded to the structure to be monitored, for example by gluing, and the other is directed outwards. A solenoid 3 is arranged on this other face. The geometric characteristics of the solenoid (wire diameter, number of turns, and, in the case of a rectangular solenoid, the width and length of the coil) are defined according to the structure to be monitored. Optionally, a ferromagnetic core 4 of the ferrite type is placed inside the solenoid 3 to increase its efficiency.

[0044] A device for generating and receiving guided ultrasonic waves in a structure to be inspected by means of single-element magnetostrictive transmitting / receiving probes as described with reference to [Fig. 1], operates as follows. The single-element magnetostrictive transmitting / receiving probes are first arranged at a distance from each other, on either side of an area of ​​the structure to be inspected.

[0045] The magnetostrictive material strip must be polarized, either beforehand by a magnet, or continuously with a magnet or electromagnet placed on the strip, so as to create a static magnetic field Bo, called the polarization field. A voltage generator delivers a short pulse (amplitude, frequency and pulse width to be determined), while a power amplifier transforms and amplifies it. The current is used to excite the solenoid and creates a variable (or alternating) magnetic field (BAC), which interacts with the polarized magnetostrictive strip, generating ultrasound (high-frequency mechanical vibrations) within the strip's thickness (following the resulting field of B0 + BAC). The strip, attached to the structure being monitored, transmits ultrasound (as guided waves) into that structure.

[0046] Guided waves propagating inside the structure's wall arrive at the single-element magnetostrictive receiving probe. The arrow in [Fig. 1] shows the direction of propagation of the ultrasonic waves relative to the solenoid mounted on the strip. This probe converts the mechanical vibrations into an electrical signal. This signal, after passing through a signal conditioner, is digitized by a digitizing device and then processed in a post-processing chain.

[0047] However, while such probes have already been used for detecting corrosion defects in metallic structures, it turns out that detection is limited to defects of a few hundred cm². There is also a significant dead zone, making the area near the single-element transmitting / receiving probe uncontrollable. Moreover, the sensitivity of the detection is greatly affected by the environment surrounding the structure being inspected. Finally, the evaluations of defects using these probes remain general and qualitative, and do not allow for the sizing of the defects.

[0048] As we will now see in more detail, the control system of the invention remains based on the generation and reception of ultrasonic guided waves in the structure to be controlled, but by means of multi-element magnetostrictive emission / reception probes, as opposed to the single-element probes described previously.

[0049] A multi-element magnetostrictive probe comprises, on the transmitting and receiving sides, a strip of magnetostrictive material intended to be placed on the structure to be monitored, a permanent magnet that can be placed on the strip to ensure the stability of the static field over time, and a plurality of solenoids arranged on the strip, for example, but not limited to, at regular intervals, forming an array of independent transmitting / receiving transducers. It should be noted that the strip may consist of a common strip for all the solenoids, or of a plurality of individual strips associated with each solenoid. By way of non-limiting example, the array may be linear (all elements on a single line). Alternatively, the transducer array may be non-linear (at irregular intervals), composed of scattered elements arranged around the area to be monitored.

[0050] The number of solenoids and the interval between each solenoid are defined according to the expected spatial resolution. The relative arrangement of the magnetic fields Bq (static) and BAC (alternating) can be adapted depending on the presence or absence of permanent magnets. The operating principle of multi-element probes in a network of transmitting / receiving transducers, as well as an example of their possible arrangements on the structure to be monitored, are illustrated with reference to [Fig. 2]. According to this example, the structure to be monitored consists of a tubular structure 50, such as a pipe, conduit, pipeline, etc., having a diameter d and a wall thickness e. This example is given by way of illustration and is not limiting, and more generally, the system of the invention can be applied to all types of thin structures such as plates, shells, tubes, tanks, etc.

[0051] Several configurations are possible depending on the required resolution and the degree of accessibility to the structure to be inspected. Thus, according to the example in [Fig. 2], an emitting probe E is placed at one end of an inspection zone Z of the tubular structure 50. The magnetostrictive material strip 2E of the emitting probe E is arranged so as to extend along at least part of the circumference of the tubular structure, in a plane transverse to the axis of the tubular structure, and the plurality of solenoids 3E arranged on the strip 2E, for example at constant intervals, form a linear array of emitting transducers Ei=i„.„Ne of the emitting probe E. A receiving probe R is placed at a second end of the inspection zone of the tubular structure 50, at a distance from the emitting probe E.The magnetostrictive material band 2R of the receiving probe R is also arranged to extend along at least part of the circumference of the tubular structure 50, in a plane transverse to the axis of the tubular structure. The plurality of solenoids 3R arranged on the band 2R, for example at constant intervals, form the receiving transducer array Rj=i,...Nr of the receiving probe R, arranged substantially parallel to the linear transmitting transducer array. The respective transmitting / receiving transducer arrays thus each form a circular transducer array surrounding the structure to be monitored. The guided waves propagate through the thickness of the wall of the tubular structure between the transmitting probe E and the receiving probe R.

[0052] As an alternative or in combination with the previous configuration, the transmitting probe E and the receiving probe R can be arranged on the tubular structure to be controlled, so that the transmitting transducer arrays Ei=i,...jNe and receiving transducer arrays Rj=i,...Nr extend parallel to the axis of the tubular structure, diametrically opposite each other, over the entire length of the area to be inspected Z of the tubular structure.

[0053] According to the multi-element acquisition principle of the control system of the invention, each transmitting transducer Ei=i„.„Ne of the transmitting probe E is controlled in turn to emit an ultrasonic signal towards the receiving transducers R j=i,...Nr of the receiving probe R, which are adapted to each provide, in response to the The signal emitted by each transmitting transducer is a time-domain detection signal. This yields a detection matrix M composed of NexNr time-domain signals measured as illustrated in [Fig. 3]. It is on this detection matrix that the algorithmic reconstruction will be performed to represent the desired result, namely the quantification of corrosion by the variation in thickness and typically by the loss of thickness, in the walls of the structure where the guided waves propagate. This reconstruction will take the form of a two- or three-dimensional map, readable and easily interpretable, as will be described in more detail later.

[0054] The advantage of using multi-element magnetostrictive probes lies in the fact that they offer better selectivity of guided modes, particularly for SH modes. These modes, which propagate parallel to the structure's wall, are advantageously very sensitive to thickness variations. They also have the advantage of being relatively insensitive to the environment, both internal and external, of the structure being monitored. Thus, the aim is to make the transmitting / receiving probes selective in order to favor a guided mode, called the monitoring mode, which is sensitive to thickness variations. In this case, the guided mode in question is the horizontal shear SH mode. However, there are different orders n for this guided mode.

[0055] On the one hand, the accuracy of the proposed reconstruction method relies essentially, as will be seen in more detail later, on the accuracy of determining the time of flight of a wave packet propagating through the structure. To estimate it correctly, the wave packet of the monitoring mode must be well isolated (temporally) from the other wave packets present in the structure. Ideally, the probe should excite this mode alone, at a precise frequency, that is, at a single operating point. This point must be located in a frequency range where the mode exhibits high sensitivity to thickness variations, which translates into a steep slope in the group velocity (wave packet velocity) as a function of the frequency x thickness product.

[0056] On the other hand, the resolution of the method (i.e., its ability to correctly quantify a detected defect) is limited by the wavelength of the mode in question. This must be very small compared to the dimensions of the defects to be quantified. Consequently, the excitation frequencies sought must be relatively high.

[0057] As previously stated, SH modes are very sensitive to thickness variations and, moreover, advantageously exhibit similar frequency-dependent behavior, which facilitates the selection of the operating point, even at high frequencies. It should be noted that Sn and An modes are as sensitive as the SHn modes. However, the difference is that these modes exhibit erratic behavior at high frequencies, unlike SHn modes. This makes it difficult to identify an optimal operating point, which is why SHn modes are preferred.

[0058] The selection of the operating point, and therefore the selectivity of the transmit / receive probe with respect to the most suitable SH mode, is achieved simply by setting the dimensions of the solenoid, preferably rectangular. These dimensions are directly related to the wavelength Lm by the relation: λ = ln / 2, where l is the width of the coil. Note that the solenoid could, however, have a shape other than rectangular.

[0059] It is then necessary to determine this operating point. To do this, a first step is to calculate the group velocity dispersion curves to identify the guided modes likely to propagate in the structure to be monitored following excitation at a given center frequency. Then, depending on the size (diameter Dd) of the defects to be quantified and the distance (D) between the transmitting / receiving probes, the upper limit of the system's wavelength, denoted Lm_High, is set, respecting the condition on the Fresnel zone, i.e., Lm_High = Dd*2 / D. Note that the principle of the method is based on a straight-ray approach. This is a high-frequency approximation. That is to say, wave propagation is considered to occur along straight lines between the transmitting and receiving probes, including through the defect. The effects of diffraction, refraction, and wave conversion are neglected.However, this is only valid when the defect is smooth (slightly abrupt thickness variation) and its size is much larger than the wavelength Lm and much larger than the width of the first Fresnel zone Lf = (Lm^D) A(1 / 2). The lower limit of the wavelength, denoted Lm_Low, is also fixed. It is imposed by the wave attenuation as a function of distance and the generation / detection amplitude limit. Therefore, the wavelength Lm must be chosen within the following interval: Lm_Low < Lm < Lm_High.

[0060] Preferably, a wavelength slightly shorter than Lm_High is chosen and which satisfies the following criteria: - The operating point must be located in a fairly dispersive area on the group speed curve, to ensure good sensitivity to thickness variation. - The operating point must be away from the cutoff frequency. - Since the probe is not perfect, other modes present in the excitation frequency range may be generated. To avoid potential overlap between these modes and the chosen mode, the corresponding operating point must be properly isolated from other modes in terms of group speed.

[0061] The curve of the group velocity fixed by the chosen wavelength is then plotted according to the expression: Cgi - Ct^U Cpl,

[0062] with Cpl the phase velocity corresponding to the chosen wavelength defined as follows: Cpl = Lm*f = 2*Z* / and Ct the velocity of the transverse waves in the material of the structure.

[0063] The points of intersection of this curve with the dispersion curves of the guided modes constitute the possible operating points with the probe of width l. Each operating point is characterized by a distinct excitation frequency. Thus, by choosing a suitable excitation signal (with a sufficiently narrow spectral width), the selection of the operating point can be ensured.

[0064] By way of example, with reference to [Fig. 8], consider a 50" steel plate with dimensions 2100x2100x25 mm, exhibiting a 51" rectangular defect with dimensions 300x300 mm and a depth of 1 mm (i.e., a 4% thickness loss in the defect area). The minimum distance between the transmitting transducer E and the receiving transducer R is 2000 mm. To obtain sufficient resolution for quantifying the defect, the wavelength must satisfy:

[0065] Lm_High < Dd - 300 mm and Lm_High < Dd^HD = sqrtÇl)*300A2Z2000 ~ 64mm

[0066] Following this upper limit of the wavelength, we can define the admissible zone for the choice of the operating point in the group velocity dispersion curves, said zone being delimited by the group velocity curve fixed by the upper limit of the wavelength, referenced Cgi (Lm_High) on the [Fig.4], illustrating the diagram of the group velocity dispersion curves of guided waves as a function of the dimensionless frequency, i.e. expressed according to the frequency x thickness product.

[0067] From this diagram, it can be identified that the SH1 mode is a good candidate, particularly around the dimensionless frequency of 2.5 MHz·mm (shaded area). Thus, for a 25 mm thick plate, the center frequency of the corresponding excitation signal is set at f = 100 kHz. This operating point is indeed located in a sufficiently dispersive region, relatively isolated from the group velocity of the other modes, and quite far from the cutoff frequencies. The corresponding wavelength is estimated at Lm = 40 mm. Consequently, the width "l" of the rectangular solenoid of each transmit / receive transducer is set at l - Lm / h = 20 mm.

[0068] The magnetostrictive emission / reception probes of the control system of the invention are therefore adapted to generate and detect a specific guided mode, suitable for corrosion control. This significantly facilitates post-processing and improves the accuracy of measurements.

[0069] We will now describe, with reference to [Fig. 5], a block diagram of the control system according to the invention for monitoring corrosion in a metallic structure to be inspected, in the case of an embodiment where said metallic structure 50' is a rectangular steel plate structure, which has a corrosion defect 51', for example located substantially in its central part. The control system comprises an emitting probe E and a receiving probe R as described above, arranged on either side of the area to be inspected of the structure 50'. In this embodiment, the emitting / receiving probes form two parallel linear arrays of emitting / receiving transducers Ei=i,.....Nr, completely or partially surrounding the area to be inspected.

[0070] The number of transducers and the interval between each transducer are defined according to the expected spatial resolution.

[0071] The control system includes means for generating an excitation signal 60, capable of delivering an excitation signal at the operating frequency defined according to the principles set forth above. This excitation signal is intended to excite the solenoid of the transmitting transducers Ei=i ...jNede the transmitting probe E via a power amplifier AL. In accordance with the invention, switching means 70 are provided to sequentially control, in time, the excitation of each transducer in the transmitting transducer network by the excitation means 60. These switching means 70 comprise an IxN channel demultiplexer, N being equal to the number Ne of transmitting transducers, capable of sequentially connecting / disconnecting the transmitting transducers E^i, Nede the transmitting probe E to the excitation means 60.

[0072] On the receiving side, the receiving transducers Rj^ Nrde of the receiving probe R are connected to acquisition means 80 capable of sequentially acquiring, via a signal amplifier A2, the detection signal delivered by each transducer in the receiving transducer array, in response to the signal generated by each transmitting transducer. The connection / disconnection of the receiving transducers Rri. Nrde of the receiving probe R to the acquisition means 80 is performed sequentially using switching means 90, comprising an Nxl channel multiplexer, N being equal to the number Nr of receiving transducers.

[0073] The excitation means 60, the acquisition means 80, and the switching means 70, 90 are connected via communication buses to a computer system 100 comprising a human-machine interface and a processing unit. The computer system 100 is adapted to record the acquired detection signals. The processing unit is adapted to execute a tomographic reconstruction algorithm for implementing a two-dimensional image reconstruction Dimensions of the inspected area, providing a distribution of the structure's thickness loss at every point within the inspected area. This thickness map resulting from the tomographic reconstruction is intended to be displayed via the human-machine interface.

[0074] Figure 6 describes a detailed flowchart of the operation of the system [Fig. 5] control. When the measurement is triggered, the first step E0 is the configuration of the system parameters. This step consists of defining all the input data, namely: the geometric and material properties of the structure to be monitored, the geometry and location of the transmit / receive transducer arrays, their characteristics, the excitation signal, and the guided mode chosen for the monitoring. The dispersion curves are calculated here. In a subsequent step E1, the connection to the computer system via the various communication buses can be verified. Then, based on these system input parameters, the excitation means 60 generate the excitation signal in a step E2. This signal can then be amplified by the amplifier A1 to a power level necessary to drive the transmitting magnetostrictive probe E in order to obtain a satisfactory signal-to-noise ratio.

[0075] The system is used for transmission tomography. Measurements are performed in a loop for each pair of transmitting transducers Ei=i ...jNe / receiving transducers Rj =i„..Nr. Thus, in a step E3, a first transmitting transducer E; of the transmitting probe, connected to the excitation means 60 via the demultiplexer 70, excites the guided ultrasonic wave which propagates through the structure.

[0076] In a step E4, a first receiving transducer Rj is connected to the acquisition means 80 via the multiplexer 90. Then, in a step E5, the detection signal delivered by this receiving transducer Rj is recorded for a duration T.

[0077] Steps E4 and E5 are repeated, incrementing j from one to Nr. Thus, Fonde is re-emitted by the first transmitting transducer Ei=i of the transmitting probe and received by a second receiving transducer Rj and so on until the last receiving transducer RNr.

[0078] When the detection signal has been recorded for all receiving transducers Rj=i„..Nr, the process loops back to step E3, incrementing i. Thus, a second transmitting transducer E; of the transmitting probe is connected to the excitation means 60 via the demultiplexer 70 and transmits the guided ultrasonic wave that propagates through the structure. The acquisition steps E4 and recording steps E5 are again performed sequentially for all receiving transducers, and the process loops back to step E3, incrementing i, and so on until the last transmitting transducer ENr.

[0079] Thus, the scanning of the structure by the ultrasonic guided waves is carried out in the same order for all the emitting transducers.

[0080] Once the measurement cycle is complete, the tomographic reconstruction algorithm is implemented in a post-processing step E6. Thus, at the end of the measurement cycle as just described, a matrix M of detection signals is obtained, as illustrated with reference to [Fig. 3]. These time-domain detection signals are stored in memory and then processed to extract the relevant data. From this dataset, called projections, the tomographic reconstruction algorithm is designed to calculate a map of the parameter of interest, in this case, the thickness loss. Finally, in a step E7, the reconstruction results are displayed via the human-machine interface.

[0081] We will now describe in more detail the operation of the tomographic reconstruction algorithm implemented in step E6.

[0082] The principle of tomography is to perform a series of measurements at different angles (projections) all around the unknown object, in this case the corrosion defect. By summing the spread projections, the physical and / or geometric properties of the defect can be reconstructed.

[0083] In the context of the present invention, a transmission tomography inspection system is used, employing a parallel array of transmitting and receiving transducers, completely or partially surrounding the area of ​​the structure to be inspected, so as to observe variations in the propagation signals of guided waves. As explained previously, wave propagation between each pair of transmitting and receiving transducers is assumed to occur along straight rays, including through the defect. Diffraction effects are neglected here.

[0084] The corrosion defect in the structure to be controlled can advantageously be revealed by a localized loss of thickness. With reference to Figure 7, the distribution of the loss of thickness can be defined as a two-dimensional function / (X, y), which, at each point with coordinates (x, y) of the plane of the structure to be controlled, gives the value of the loss of thickness, and the projections pÿ are the set of measurements of the integrated values ​​of this function along the straight lines rü formed by the wavefront between each pair of transmitting transducers Ei=i,...jNe and receiving transducers Rj=i„„Nr.

[0085] p" - J jjf Çx,y)ds

[0086] To evaluate this integral, the chosen guided mode must be sensitive to the thickness variation in the operating frequency range. Furthermore, propagation velocities are among the parameters sensitive to this type of defect, in particular the C-group velocity. Therefore, according to the invention, the time of flight Tv will be used to estimate the average thickness variation between each pair of transmit / receive transducers.

[0087] Knowing the guided wave dispersion curve diagram in the plate illustrated in Figure 8, the SH1 mode will be excited, according to the principles explained above in relation to the probe's selectivity. The sensitivity is directly related to the amount of dispersion at the chosen operating point. Consequently, the dispersion at this point must be significant. In this example, the excitation frequency is set around 100 kHz. The signals are recorded up to 3000 ps, ​​and a sampling frequency of 5 MHz is used. From a series of guided wave transmission measurements in the plate between the transmitting and receiving transducers, performed for the healthy plate and for the plate with defects, the variation in time of flight is deduced, established in this example as: ATV = 3.54 jus.

[0088] Figure 9 shows the group velocity dispersion curves as a function of the frequency-thickness product in MHz x mm. The transverse guided mode SH1 exhibits sufficient dispersion around the 2.5 MHz x mm point and propagates at a given average velocity. Passing through the thickness loss resulting from the defect in the plate will lead to a decrease in the average velocity.

[0089] Knowing the distance between the transmitting and receiving transducers E and R, from the variation in time of flight determined for the SH1 mode, we can deduce the variation in the average group speed: AC = '10-67 m / s, according to the example. Then, by projecting this average group velocity variation onto the dispersion curve of the transverse guided mode SH1 in Figure 9 and knowing the excitation frequency used, i.e. 100 kHz, we can then determine the corresponding average thickness variation: AH = 0.1513 mm, according to this example.

[0090] We will deploy this method of determining the variation in thickness in the plate for all pairs of transmit / receive transducers (E;, Rj) of the multi-element transmit / receive probes implemented in the control system of the invention.

[0091] Thus, the propagation distances Dij between each pair of transducers Given the transmission / reception values ​​(E, Rj), the average group velocities 'J can be easily quantified by the wave packet times of flight: Y v

[0092] JJ = ^0 g “ JJ y1

[0093] Considering that the transverse guided mode SH1 propagates at a speed (70 to the given excitation frequency, and that traversing a thickness loss results in a decrease in average speed, which is expressed as follows: aV o di, C = C - AC ggg then by projecting this onto the dispersion curve in Figure 9 and knowing the For the excitation frequency, the corresponding average thickness can be determined. For SHn modes, a simplified relation of this thickness function is written:

[0095] Where n GN is the mode index. CT is the propagation speed of transverse waves in the material of the structure, f„ is the center excitation frequency.

[0096] The accuracy of the reconstruction depends essentially on the accuracy of the time-of-flight estimation ^7. To extract the latter with good accuracy, known signal processing techniques are applied.

[0097] Figure 10 shows a detailed flowchart of the reconstruction algorithm, implemented in post-processing step E6 of Figure 6.

[0098] It should be noted that the method described here assumes that reference detection signals acquired for a healthy state of the structure are available. However, it could still be used in the opposite case by considering a variation of the detection signals relative to nominal values.

[0099] A first step E60 implements windowing of the detection and reference signals for the structure to be monitored, accompanied by an optional dispersion correction step E61. This consists of multiplying the time signals by a tapered cosine VF(i) window (also known as a "Tukey window") to isolate the wave packet of the chosen mode. The width and bounds of the window are automatically adjusted according to the characteristics of the excitation signal and the propagation distance. This results in the following windowed signals: 101001 (tk).W(rk),k=l,...,Ns S, d

[0101] where tk G[0, T] is the discrete time, Ns is the number of samples in the signal. The indices s and d correspond respectively to the healthy and defective state of the structure. The signal amplitudes are then normalized with respect to the maximum.

[0102] Dispersion distorts the excitation waveform as it propagates. The wave packet spreads out in time and space with a decrease in amplitude. This makes processing experimental data difficult. To overcome this problem and improve the accuracy of the estimates, algorithms

[0103]

[0104]

[0105]

[0106]

[0107]

[0108]

[0109]

[0110] [YES] dispersion correction can be used in an E61 step. The next step, E62, involves interpolating the obtained signals to increase the number of points N and thus the resolution. After conditioning, the time-of-flight estimation is performed in step E63. Several techniques can be used, including: envelope calculation using the Hilbert transform, cross-correlation, and time-frequency analysis using the Wigner-Ville distribution. These techniques can be used depending on the signals and the approach being considered. The envelope method is defined by default. It offers high accuracy and is applicable to each signal in the matrix. Cross-correlation is very effective and recommended when reference signals are available. However, it only works for well-isolated wave packets. In contrast, time-frequency analysis is more advantageous in the case of multiple modes. Then, the projection calculation is implemented in step E64. From the estimated flight times, knowing the distances between the transmitting and receiving transducers, the average group speeds are calculated. Through the £ From the dispersion curve (see the example in Figure 9), the average thicknesses are deduced. The function f(x, y) is then calculated by subtracting the functions d thicknesses, as follows: f(x,y)=H,(x,y)-Hd(x,y) Therefore, the projections are obtained through the following calculation: Working in relative mode, that is, with respect to reference signals acquired in the structure's healthy state, minimizes errors induced by the experimental measurement system. If reference detection signals are unavailable, the nominal value can be subtracted. The results of the projection calculations are stored in a projection matrix p = [y>]. Consider the example of [Fig. 11], which schematically illustrates the plate of [Fig. 8], instrumented with two parallel arrays of transmitting transducers Ei=i,...jNe and receiving transducers Rj=i„..Nl, respectively, arranged on either side of the plate 50”, opposite each other with respect to the defect 51”. The transmitting and receiving transducer arrays extend here along the y-direction of the plate. Ne = Nr = 20 in this example. The transmitting / receiving transducers are arranged with a constant interval of 100 mm between them. As an alternative to the aforementioned arrangement of transmitting / receiving transducers, one could provide a network of transmitting transducers and a network of receiving transducers. parallel reception arrays extending along the x direction of the plate. These two arrangement variants can also be considered in combination, namely a first parallel array of transmit / receive transducers extending along the y direction of the plate and a second parallel array of transmit / receive transducers extending along the x direction of the plate.

[0112] The diagram on the right of Figure 11 illustrates a row of the projection matrix. More precisely, it represents the result of determining the average thickness variation along the guided wave paths from loop measurements taken for each pair of transmitting / receiving transducers Ei=i / Rj=i.,Nr. Thus, for each sequentially excited transmitting transducer of the plurality of transmitting transducers, the average thickness variation AH is calculated for each receiving transducer, in other words, for each signal acquired sequentially from the plurality of receiving transducers, according to the principles explained above. This yields a projection. By combining all the transmitting transducers, the projection matrix is ​​obtained, the rows of which are composed of the projections corresponding to the thickness variation values ​​obtained for each transmitting transducer of the system.

[0113] By applying an inverse transformation to the projection matrix, the properties of the monitored structure can be reconstructed. However, a final step E65 before reconstruction is the filtering of the projections.

[0114] Filtering the projections significantly reduces backpropagation background noise and improves reconstruction accuracy. It is performed for each projection p. — ^pl> 1 ph 2 ... p'> Nr} in the Fourier domain via multiplication by a filter function K(v) of type: Ramp, Hanning, Cosine, etc. The inverse Fourier transform recovers the projection in the spatial domain T—along the line of the receiving transducers—: 101151 S,-O)= [01161 Qp)=

[0117] The ramp filter is the default choice. Note that interpolation must be used to calculate the transformations.

[0118] Following the filtering step, a reconstruction step E66 is implemented. This reconstruction step is the final step in the post-processing. Instead of the backprojection method used in the classical approach, which requires specific mathematical development depending on the sensor network topology, we use the elliptic ray approach developed for the RAPID algorithm described in patent document CN102928511 cited in the preamble to this description.

[0119] This approach consists of backprojecting each element of the matrix Q = [ç'7] independently in the form of an ellipse between each pair of transmit / receive transducers, as illustrated in [Fig. 12]. Thus, as illustrated in [Fig. 13], the filtered projections are spread in the form of an ellipse across each pair of transmit / receive E / Rj transducers in order to reconstruct the defect. The sum of the radii allows the characteristics of the monitored area to be accentuated and thus the thickness loss to be reconstructed.

[0120] To do this, the structure is first meshed. The geometric function G^(x, y) is then calculated for each pair of transducers. E / Rj transmission / reception at any point d(x, y) of the mesh:

[0121]

[0122] This function corresponds to the quotient of the sum of the distances between the points ten, y) to the transmitting transducer E^x^ y.) and to the receiving transducer R.ix., 4 r yj with respect to the distance jjd between them. It is equal to 1 for all the points in the direct path and also increases with distance. To form the elliptic radius, a thresholding is applied. Thus, the elliptic distribution for a radius (Ej, R) is defined by:

[0123] {fi-G..(x, V) y) = L / L ■ 0, if fi < G.^x, y)

[0124] where fi is the threshold, called the shape factor, used to control the size of the ellipse. Finally, the approximation of the thickness loss at each point of the mesh is obtained by linear summation of the ellipses of all possible pairs of transmit / receive transducers such that: 101251 / (x,n =

[0126] The resolution of the reconstruction depends strongly on the size of the ellipses. If this size is too large, the defect cannot be resolved. Otherwise, the reconstruction converges towards linear backprojection and highlights the inadequacy of the measurement data. To overcome this difficulty, the form factor fi is optimized as a function of the number of transducers.

[0127] Figures 14a and 14b show the results of the tomographic reconstruction, based on the detection signals acquired according to the example of the plate in Figure 11. Thus, with reference to Figure 14a, a MAP map of the thickness loss of the 50" plate is obtained, providing, at any point (x, y) of the plate, a value The quantified thickness loss is represented primarily using an SCA color scale (in meters in this example), with a color assigned to each of several ranges of thickness loss values ​​derived from the measurements. This color scale facilitates easy interpretation of the resulting map, both in terms of the location of the thickness losses and their quantitative assessment. The rectangular corrosion defect 51” is thus clearly visible on the map. Figure 14b illustrates, in solid lines, the variation of the estimated thickness profile of the plate along the y-axis. This estimated profile corresponds to a cross-sectional view of the thickness loss map along y, following a cutting plane passing through a median axis of the rectangular defect 51”, thus providing a third dimension that allows for a more refined analysis of the defect.The theoretical variation of the plate thickness profile along this same cross-sectional plane is also shown (in dashed lines). The reconstruction results show relatively good agreement with the theoretical values ​​of the thickness loss.

[0128] Alternatively, the reconstruction of the defect on the plate could be carried out using a circular array of transmit / receive transducers, for example 40 in number, arranged on the plate so as to surround the defect. The reconstruction of the defect could also be carried out using a second parallel array of transmit / receive transducers extending along the x-direction of the plate.

[0129] Thus, by prior selection of an appropriate operating mode and frequency, the time of flight is used to estimate the thickness loss between each pair of transmit / receive transducers. The estimated thickness loss is backprojected into an ellipse connecting the transducers of each transmit / receive transducer pair. The sum of all the contributions from the transducer pairs provides a map of the spatial distribution of the thickness loss. Backprojecting into an elliptical pattern instead of a linear pattern advantageously overcomes the problem of insufficient measurement data. Therefore, a tomographic image can be obtained with fewer transducers. Furthermore, the filtering process implemented prior to reconstruction reduces artifacts and increases reconstruction accuracy.Furthermore, the simplicity of the summation gives the method great flexibility with respect to the geometry of the transducer networks.

Claims

1. Demands A system for controlling the corrosion of a metallic structure using guided ultrasonic waves produced by magnetostrictive effect, comprising: - at least one emission probe (E) comprising at least one membrane formed of a strip (2E) of a magnetostrictive material, this strip having two opposite faces, one of which is directed inwards and adapted to be made integral with the structure to be monitored, and the other is directed outwards and on which an excitation coil is placed to create in the vicinity of the membrane a variable magnetic field in relation to an electrical excitation signal applied to the terminals of the coil, and - at least one receiving probe (R) comprising at least one membrane formed of a strip (2R) of a magnetostrictive material, this strip having two opposite faces, one of which is directed inwards and is adapted to be made integral with the structure to be monitored, and the other is directed outwards and on which is placed a detection coil to deliver an electrical detection signal related to the variation of the magnetic field in the membrane as a function of a received ultrasonic wave, - Said transmitting (E) and receiving (R) probes being intended to be installed on either side of an area to be inspected of said structure; said excitation / detection coil comprises a plurality of solenoids (3E / 3R) arranged on the strip, forming an array of independent transmit / receive transducers (Ei=i / Rj=i,...Nr), said system comprising: - excitation means (60) by applying an alternating excitation voltage to an emitting transducer at a predetermined excitation frequency, - switching means (70) for sequentially controlling, in time, the excitation of each transmitting transducer in the transmitting transducer network (E,-| XJpar said excitation means, - acquisition means (80) for sequentially controlling in time the acquisition of the detection signal delivered by each receiving transducer of the receiving transducer network (Rj=i„..Nr), in response to the signal generated by each transmitting transducer, - means (100) for recording a detection matrix corresponding to the set of acquired detection signals, - means for processing (100) the detection matrix by means of a tomographic reconstruction algorithm intended to reconstruct a two- or three-dimensional map of the area to be inspected, providing a distribution of the variation in thickness of the structure at any point in the area to be inspected.said system being characterized in that said solenoids are rectangular in shape, the width of which is defined so as to generate and detect, for a selected excitation frequency, a specific guided mode adapted for corrosion control, including at least the SH mode.

2. System according to claim 1, characterized in that the transmitting transducer array and the receiving transducer array are arranged so as to at least partially surround the area to be inspected.

3. System according to any one of the preceding claims, characterized in that the receiving transducer array is arranged parallel to the transmitting transducer array.

4. System according to any one of the preceding claims, characterized in that said structure being a tubular structure, the transmitting and receiving transducer arrays extend along at least a portion of the circumference of the tubular conduit, in a plane transverse to the axis of the tubular structure.

5. System according to claim 4, characterized in that the transmitting and receiving transducer arrays extend parallel to the axis of the tubular structure, diametrically opposite to each other, over the entire length of the area to be inspected.

6. A method for controlling corrosion in a metallic structure by guided ultrasonic waves produced by magnetostrictive effect, comprising the following steps: - a control system is provided according to any one of claims 1 to 5, - the excitation of each transducer of the transmitting transducer network is sequentially controlled in time by said excitation means, - the acquisition of the detection signal delivered by each transducer of the receiving transducer network is sequentially controlled in time, in response to the signal generated by each transmitting transducer, - the detection matrix corresponding to the set of acquired detection signals is recorded, - said detection matrix is ​​exploited by means of a suitable tomographic reconstruction algorithm to reconstruct a two- or three-dimensional map of the area to be inspected, providing a distribution of the thickness loss of the structure at any point of the area to be inspected.

7. A method according to claim 6, characterized in that a prior selection of specific guided mode and excitation frequency is made, adapted to corrosion control.

8. A method according to claim 7, characterized in that the time-of-flight variation for said specific guided mode related to the thickness variation along the wave path between each pair of transmit / receive transducers is determined, the average velocity variation for said guided mode is deduced therefrom, and the corresponding average thickness variation between each pair of transmit / receive transducers is determined by projecting this average velocity variation onto the dispersion curve of said specific guided mode expressing the group velocity as a function of the excitation frequency-thickness product of said structure, the result of determining the average thickness variation along the guided wave paths for each pair of transmit / receive transducers constituting a projection matrix.

9. A method according to claim 8, characterized in that said mapping is obtained by backprojecting all the projections of the projection matrix in the form of an ellipse connecting the transmitting / receiving transducers of each pair of transducers

10.

11. of transmission / reception and by summing the contributions of each pair of transmission / reception transducers. A method according to claim 9, characterized in that the back projection step is preceded by a projection filtering step. Method according to claim 10, characterized in that the filtering step is carried out for each projection in the Fourier domain via multiplication by a filter function.

Images