RU2757056C1 - Method for detecting and localizing damage in thin-walled structures using lamb waves - Google Patents

Abstract

FIELD: thin-walled structures damage detection.

SUBSTANCE: invention is intended for detecting and localizing damage in thin-walled structures. The substance of the invention lies in the fact that a frame with 8 piezoelectric transducers (PT) installed at the tops of the square and in the middle of its sides is mounted on an intact structure using a fixing device; with the help of 4 reversible PT located in the middle of the sides of the square, the Lamb wave trains are alternately excited in the structure; with the help of the specified reversible PT, as well as the specified PT at the vertices of the square, alternately receive the signals passed through the structure, these signals are recorded and stored using a digital computerized system; then the frame with PT is dismantled; mounting and dismounting of the frame from the PT and recording the signals are repeated several times and, based on the difference signals, the histograms of the discrimination parameter are formed, which are necessary to select the defect detection threshold.

EFFECT: providing the possibility of detecting and localizing a defect, both in metals and in composite materials, with a small number of PT, as well as a fairly simple equipment and processing algorithm.

1 cl, 2 dwg

Description

The invention relates to the direction of technology for non-destructive testing, called "systems for monitoring the state of structures" (in English literature - system health monitoring or SHM). The method for detecting and localizing damage in thin-walled structures is based on the excitation of pulses of ultrasonic Lamb waves with the help of several piezoelectric transducers (PPs), their reception also with the help of several PPs and analysis of the received signals, with the help of which the position and nature of the damage is determined. PCBs are mounted on a removable frame, which is installed on an undamaged structure to record the reference acoustic field. Such measurements are made several times by mounting and dismounting the sensors to accumulate statistical data necessary to select the damage detection threshold. After the completion of the operation cycle of the structure, during which its damage may occur, the frame with the PP is again installed on the structure and a new (disturbed) acoustic field is recorded. To extract information about damage, a comparison of the disturbed wave field with the reference one and subsequent processing using an algorithm based on the physical laws of the interaction of two-dimensional Lamb waves with inhomogeneities (damages) is used. The proposed method also includes assessing the size of the damage.

SHM systems are known from numerous scientific publications and patents, which consist of a network of reversible PCBs. These PCBs can be removable, glued to a thin-walled structure (plate) or embedded in it (if it is made of composite material). Emitting PPs form Lamb wave (VL) pulses in the structure, which are received by other PPs. In this way, a lot of overhead transmission lines are formed. Some of them may overlap. With the help of a digital computerized data acquisition system, a set of signals is recorded that correspond to a known undamaged state of a structure. This distribution of signals is called the "base line". The emerging damage (defect) changes the recorded signals. Comparing the current distribution with the reference one, the position and nature (size) of the damage are determined. The above-mentioned systems differ in the PCB placement schemes, the method of forming the propagation paths of the probing pulses (by switching the emitter-receiver pairs) and, most importantly, information processing algorithms.

For example, in patent CN 110389176 A (Fatigue damage diagnosis method for carbon fiber reinforced matrix composite structure), the data is pre-processed using a continuous time-frequency wavelet transform. Then a certain "damage index" is calculated and with the help of a neural network, which is trained on examples of artificially introduced damages, the position and characteristics of the fatigue damage are determined.

In patents CN 106404911 A (Real time delay single-mode Lamb wave phased array system for plate structure detection) and US 20110058453 A1 (Lamb wave dispersion compensation for EUSR approach to ultrasonic beam steering) it is proposed to use a phased antenna array to determine the position of damage, which works for both radiation and reception. In other words, a well-known approach is used, which is typical for radio and acoustic location. Thanks to electronic scanning of the beam and pulsed radiation, it becomes possible to determine the position of the defect over a sufficiently large area. In this case, it is not required to move the PCB along the structure.

On the contrary, CN 103792287 A (Large-area structural damage detection method based on Lamb wave) uses a single emitting-receiving PCB, which is rearranged along the structure. With the help of frequency scanning, the cutoff frequency of the antisymmetric mode of the A1 overhead line is determined, then signals are emitted at a lower frequency, where only the A0 mode exists, and the entire structure is scanned in order to determine the position and nature of the defect.

The closest in terms of the location of sensors and processing algorithms are the technical solutions described in patent US 7458266 B2 (Method and apparatus for detecting a load change upon a structure and analyzing characteristics of resulting damage). They are taken as a prototype. This patent deals with a regular grid of PCBs that can operate only for reception (passive mode), or for transmission and reception (active mode). In active mode, the position of the damage is determined by the "transmissive" method (per pass) and / or by the method of reflections. The disadvantage of the algorithm used in the transmission method is the need to use many points of intersection of the propagation paths of overhead lines to determine the position of the defect. At the same time, in order to achieve good accuracy, it is necessary to install a large number of sensors and use complex commutation, which increases the volume of equipment and complicates processing.

The reflection method works well in metals, where overhead lines, unlike composite materials, do not experience strong absorption. Therefore, in structures made of composite materials, it is practically inapplicable.

Compared to the prototype, the proposed method has the following advantages:

- can be used in both metals and composite materials;

- efficient with a small number of PCBs;

- simplifies hardware and processing algorithm.

The objective of the invention is to create a method for detecting and localizing a defect, which (method) can be used both in metals and in composite materials, while it is efficient with a small number of PCBs and simplifies the equipment and processing algorithm.

The problem is solved by the fact that in the method of detecting and localizing damage in thin-walled structures, a frame with 8 piezoelectric converters (PP) installed at the tops of the square and in the middle of its sides is mounted on an intact structure using a fixing device;

with the help of 4 reversible PPs located in the middle of the sides of the square, the Lamb wave trains are alternately excited in the structure;

with the help of the specified reversible PP, as well as the specified PP at the vertices of the square, alternately receive the signals passed through the structure, these signals are recorded and stored using a digital computerized system;

then the frame with PP is dismantled;

mounting and dismantling the frame with the PCB and recording the signals are repeated several times and, based on the difference signals, generate histograms of the discrimination parameter necessary to select the defect detection threshold;

additionally, using simulators of different diameters, as necessary, accumulate histograms to determine the size of the defect;

further, in order to detect defects obtained by the specified thin-walled structure during its operation, the frame with the PCB is installed in the same position, the above-mentioned operations of emission-reception of said signals are repeated and the disturbed signals are stored;

perturbed and reference signals are subtracted and mean amplitudes are calculated for the difference signals in time windows, the parameters of which are preliminarily determined from the group velocity of the mode of the zero mode of Lamb waves;

the amplitudes of the difference signals are normalized to the amplitudes of the reference ones, calculated in the same time windows;

the discrimination parameter is compared with a threshold, and if exceeded, a defect detection decision is made;

4 directions per defect are calculated according to the formula

where Angle _k is the calculated value of the angular position of the defect, determined during probing with the k-th probe, Diff_amp _{n, k} is the normalized amplitude of the difference signal at the sensor n when it is irradiated by the k-th probe, AngW _{n, k} is the weight function (angular coordinates receivers n relative to the k-th PP);

coordinates of 6 points of intersection of directions per defect are calculated according to the formulas

where X, Y - coordinates of the point of intersection of directions 1 and 2; X ₁ , Y ₁ - coordinates of the 1st radiator; X ₂ , Y ₂ - coordinates of the 2nd radiator, ϕ ₁ and ϕ ₂ - estimates of the angles measured from the OX axis; γ is the angle between the directions;

the position of the defect is determined by averaging the coordinates of the points of intersection of the directions, while the points lying outside the square, as well as those obtained from the intersection of almost parallel directions, are excluded before averaging.

In a particular case, if it is necessary to control a large surface, a grid of square cells consisting of PPs is placed on it, while the emitting PPs are omnidirectional and their probing signals are used in several neighboring cells.

The proposed method for detecting and localizing a defect is based on the general laws of the interaction of waves with inhomogeneities of the medium in which they propagate. If an obstacle with a diameter D is encountered on the path of a plane two-dimensional wave with a length λ, a partial shadow region is formed ^{behind it at a distance R satisfying the inequality R <D 2 / λ.} With increasing distance due to diffraction, this region expands and in the so-called "wave zone" at R> D ² / λ, a diverging cylindrical wave with a certain angular dependence of the field amplitude appears. In the middle of the corner sector behind the defect (in the direction of the emitting PCB - defect), the attenuation is maximum, and towards the edges the effect decreases. The angular width of this sector can be estimated by the ratio θ≈λ / D, which is applicable if λ / D is significantly less than 1.

In this case, the field behind the defect can be represented as the sum of the reference field (the field in the plate without a defect) and the field scattered by the defect. Therefore, subtracting the reference field from the field distorted by the inhomogeneity, one can obtain a scattering diagram. When the diameter of the inhomogeneity is significantly larger than the wavelength, the scattering pattern is elongated in the direction of wave propagation from the emitting PP to the defect. Therefore, the scattered signal (recall that it is obtained by subtracting the perturbed and reference fields) in this direction is significantly larger than the backscattered signal. Hence, it follows that the location "in the light" is energetically more favorable than the location using the backscatter signal. To implement the "transmissive" sensing scheme, the receivers must be located within a certain sector behind the inhomogeneity. By registering perturbations in the angular distribution of the difference field amplitude with the help of several receivers, one can, first, establish the fact of the appearance of an inhomogeneity, and second, estimate the direction toward it. The direction to the inhomogeneity in this case can be determined by calculating the "center of gravity" of the angular distribution of the difference field by the formula (1)

where Angle _k is the calculated value of the angular position of the defect determined during probing with the k-th probe, Diff_amp _{n, k} is the normalized amplitude of the difference signal at the sensor n when it is irradiated by the probe with number k, AngW _{n, k} is the weight function (angular coordinates receivers n relative to PP k).

If Lamb waves are used for transmissive probing, the picture becomes more complicated due to their mode structure. It is known that the modes of these waves are divided into two groups: symmetric and antisymmetric. The zero (lowest) antisymmetric mode is a flexural wave, in which all layers of the plate move in phase, and in the zero symmetric ("peristaltic") mode, the motion is symmetric about the midline of the plate. The phase and group velocities of the symmetric mode are higher than that of the antisymmetric one. Therefore, the wave pulse excited in the plate breaks down during propagation into several signals that reach the receiver at different times. The authors have experimentally established that a fast symmetric mode almost does not "feel" such inhomogeneities as a hole, delamination, and surface mass loading. But the slow antisymmetric mode is noticeably weakened by them. Therefore, when subtracting the reference and disturbed signals, in most cases, the first arrival of the slow zero mode is confidently distinguished. This allows you to select the time windows in which further analysis of the difference signals should be performed. In these windows, the average signal amplitudes are calculated, which are then used to calculate the direction according to formula (1). To exclude the influence on the estimation of the direction of the difference in the distances between emitters and receivers, these amplitudes are normalized to the amplitudes of the direct reference signals.

FIG. 1 illustrates the described method using the example of the location of the PCB on the structure under study at the nodes of a regular grid of squares. Let us consider the operation of the system within one cell, when odd RIs located in the middle of the sides of the square are used for radiation. To calculate the direction in relation to each of them, 7 receiving PPs, located in the 180 ° sector, should be used. The spacing between the beams connecting the emitter and receivers is on average 30 °. This makes it possible to record in sufficient detail the angular distribution of the difference (scattered) field, which is necessary for a reliable calculation of the direction.

4 emitters must be switched on in turn and emit pulsed signals consisting of several periods of the carrier frequency. PP that do not participate in radiation, while working for reception. To calculate according to the data of 4 sounding cycles using formula (1) of 4 directions, the numbers of the emitting transducers should be k = 0 ... 3, and the receiving ones - n = 0 ... 7.

As a result of these calculations, 4 directions per defect are obtained, which have 6 intersection points. The coordinates (X, Y) of each of these points in the XOY coordinate system in FIG. 1 can be found by the formulas (2)

Here: X, Y - coordinates of the point of intersection of directions 1 and 2; X ₁ , Y ₁ - coordinates of the 1st radiator; X ₂ , Y ₂ - coordinates of the 2nd radiator, ϕ ₁ and ϕ ₂ - estimates of the angles measured from the Ox axis; γ is the angle between the directions.

In general, the coordinates of the intersection points do not match. Obviously, to refine the results, they must be averaged in some way. Before averaging, intersection points that are outside the hexagon and those obtained from the intersection of nearly parallel directions must be removed. For this, an estimate of the angle γ can be used. Experience shows that the smallest measurement error is given by the median averaging. The errors obtained in experiments with composite plates are about 10% of the side of the square, at the vertices of which the PCBs are located.

Receiving-emitting PP should be omnidirectional, this makes it possible to use the radiation of one PP in another neighboring cell.

All of the above refers to an idealized situation when a system of piezoelectric transducers can be fixed to a structure in a stationary manner. If we are talking about diagnostics, for example, of small aircraft, such placement is impossible. Therefore, in this method, it is proposed to install the PCB on a rigid frame, which can be installed and removed from the aircraft using a fixing device. The frame must be reinstalled before starting the service cycle. After that, a survey of the reference (undisturbed) acoustic field should be carried out, which should be recorded and stored in the data collection and processing system for subsequent comparison with the measurement data obtained after the expiration of the aircraft operation time, when its damage could have occurred. As experience shows, the records of the reference field obtained during repeated mounting and dismounting of the frame with the PCB on the structure under study, which were not even subjected to any force effects, are somewhat different. These differences are due to various external factors, the most important of which are the acoustic contact of the PP with the surface and temperature. This circumstance makes it necessary to develop a criterion for assessing the differences between the reference and disturbed fields. For this purpose, by comparing the records of the reference field during multiple mounting - dismantling of the PCB, a histogram of a certain discrimination parameter, calculated from their difference, should be built. Based on this histogram, a detection threshold is selected, when exceeded, a decision on the presence of a defect is made. A similar approach can be used to estimate the size of the damage.

The described procedure is illustrated in FIG. 2, in which the presented histograms were obtained using a surface mass load in the form of metal cylinders of two diameters. If the discrimination parameter (in this case, this is the maximum of all the maxima of the normalized differences in 4 directional dimensions) exceeds the value of 0.4, it is considered that the defect is detected and has a diameter of at least 10 mm. If this parameter is more than 0.65, the diameter is not less than 40 mm. Similar histograms should be built for each specific construction. In this case, the use of a simulator is necessary, since otherwise, the structure would have to be “broken” every time.

Claims (18)

1. A method for detecting and localizing damage in thin-walled structures, characterized in that a frame with 8 piezoelectric transducers (PP) installed at the tops of the square and in the middle of its sides is mounted on an intact structure using a fixing device; with the help of 4 reversible PPs located in the middle of the sides of the square, the Lamb wave trains are alternately excited in the structure; with the help of the specified reversible PP, as well as the specified PP at the vertices of the square, alternately receive the signals passed through the structure, these signals are recorded and stored using a digital computerized system; then the frame with PP is dismantled; mounting and dismantling the frame with the PCB and recording the signals are repeated several times and, based on the difference signals, generate histograms of the discrimination parameter necessary to select the defect detection threshold; additionally, using simulators of different diameters, as necessary, accumulate histograms to determine the size of the defect; further, in order to detect defects obtained by the specified thin-walled structure during its operation, the frame with the PCB is installed in the same position, the above-mentioned operations of emission-reception of said signals are repeated and the disturbed signals are stored; perturbed and reference signals are subtracted and mean amplitudes are calculated for the difference signals in time windows, the parameters of which are preliminarily determined from the group velocity of the mode of the zero mode of Lamb waves; the amplitudes of the difference signals are normalized to the amplitudes of the reference ones, calculated in the same time windows; the discrimination parameter is compared with a threshold, and if exceeded, a defect detection decision is made; 4 directions per defect are calculated according to the formula

where Angle _k is the calculated value of the angular position of the defect, determined during probing with the k-th probe, Diff_amp _{n, k} is the normalized amplitude of the difference signal on the sensor n when it is irradiated by the k-th probe, AngW _{n, k} is the weight function (angular coordinates receivers n relative to the k-th PP); coordinates of 6 points of intersection of directions per defect are calculated according to the formulas

where X, Y - coordinates of the point of intersection of directions 1 and 2; X ₁ , Y ₁ - coordinates of the 1st radiator; X ₂ , Y ₂ - coordinates of the 2nd radiator, ϕ ₁ and ϕ ₂ - estimates of the angles measured from the OX axis; γ is the angle between the directions; the position of the defect is determined by averaging the coordinates of the points of intersection of the directions, while the points lying outside the square, as well as those obtained from the intersection of almost parallel directions, are excluded before averaging. 2. The method according to claim 1, characterized in that, if it is necessary to control a large surface, a grid of square cells consisting of PPs is placed on it, while the emitting PPs are omnidirectional and their probing signals are used in several adjacent cells.

Images