Diameter pipe damage detection method

Abstract

The invention relates to the technical field of damage detection, in particular to a diameter pipe damage detection method, which comprises the following steps: acquiring a condition of a pipeline in a non-damage state as a reference signal, and acquiring ultrasonic guided wave signals of the pipeline under different damage conditions; extracting a first feature damage echo reflection coefficient R, a second feature damage time domain peak amplitude P, a third feature arrival time T and fourth feature damage time domain signal energy E from the ultrasonic guided wave signal; a full-connection neural network is adopted, physical information constraints are embedded, a damage echo reflection coefficient R and the damage size # imgabs0 # meet a monotone increasing relation, a mixed loss function training neural network of physical loss and data loss is introduced, and finally # imgabs1 # and # imgabs2 # are output. Therefore, by learning a plurality of signal features in a complex environment and correcting model deviation by using physical constraints, the robustness in a dynamic environment is remarkably improved, and high-precision damage detection is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of damage detection, and in particular to a method for detecting damage to a diameter pipe. Background Art

[0002] Small-diameter pipes (typically less than 25mm in diameter) are widely used in industries such as petrochemicals, nuclear energy, aerospace, and microelectronics manufacturing, carrying critical tasks such as transporting high-temperature, high-pressure fluids, cooling media, or precision gases. Due to long-term exposure to complex operating conditions such as corrosion, vibration, and fatigue, small-diameter pipes are prone to damage such as cracks, corrosion thinning, and weld defects. Therefore, efficient and accurate damage detection technology is crucial to the safe operation of small-diameter pipes.

[0003] Ultrasonic guided waves (UGWs) are currently considered an ideal technology for small-diameter pipe inspection due to their long-distance propagation capability (single-point excitation can cover several meters) and high sensitivity (capable of detecting even tiny damage or cracks). However, their application in small-diameter pipes still faces the following challenges: The reduced pipe diameter significantly increases the modal complexity of UGWs, making signal propagation susceptible to interference and significantly reducing the signal-to-noise ratio. Furthermore, UGWs have poor adaptability to dynamic environments. Under complex operating conditions such as high pressure and vibration, traditional probes struggle to maintain stable coupling with the pipe wall, resulting in unreliable inspection results.

[0004] It should be noted that the information disclosed in this background technology section is only intended to increase understanding of the overall background of the present invention, and should not be regarded as an admission or any form of implication that the information constitutes prior art already known to those skilled in the art. Summary of the Invention

[0005] The present invention provides a method for detecting damage to a bore pipe, characterized in that the method comprises the following steps: Collect the pipeline in an undamaged state as a reference signal, and collect the corresponding ultrasonic guided wave signals of the pipeline in different damage conditions; extracting a first input feature, a second input feature, a third input feature, and a fourth input feature from the ultrasonic guided wave signal; The first input feature is the damage echo reflection coefficient R, , A1 is the amplitude of the damage reflection echo, and A0 is the amplitude of the incident wave; The second input feature is the damage time domain peak amplitude P. The received ultrasonic guided wave signal with damage information is subtracted from the reference signal collected in the damage-free state to obtain a damage reflection signal. The maximum absolute value of the damage reflection signal in the time domain is extracted to obtain the damage time domain peak amplitude P. The third input feature is the arrival time T, which refers to the time difference between the emission of the excitation signal and the peak value of the first reflected wave. The relationship between T and the distance D of the damaged distance sensor is: , is the wave speed; The fourth input feature is the damage time domain signal energy E, , where x is the damage reflection signal, N is the total length of the damage reflection signal, and n is the sampling point; A fully connected neural network is used. The fully connected neural network mainly consists of an input layer, a hidden layer, and an output layer. The input layer includes 4 nodes, which correspond to the first input feature, the second input feature, the third input feature, and the fourth input feature, respectively. The number of hidden layers and the number of nodes in each layer are adjusted according to the training effect. The activation function adopts the ReLu function. The output layer includes 2 nodes, which correspond to the damage distance prediction value predicted by the neural network. and the predicted value of lesion size ; Embed physical information constraints, the physical constraints are: damage echo reflection coefficient R and damage size Satisfies the monotonically increasing relationship, that is, , where S is ; Introducing loss function; Introducing loss function, total loss function L total for:

[0006]

[0007]

[0008] Among them, L main Main damage, L constraint is the physical constraint loss, λ is a hyperparameter, and the formula N represents the total number of training samples, i For the current sample, D pred is the damage distance prediction value predicted by the neural network, S pred Represents the predicted value of the damage size, D true represents the true value of the damage distance, S true Represents the true value of the damage size; The final output is the true value of the damage distance and the true value of the damage size .

[0009] Furthermore, the collected original waveguide signal is subjected to wavelet threshold denoising.

[0010] Furthermore, Z-score normalization processing is performed on the first input feature, the second input feature, the third input feature, and the fourth input feature.

[0011] Furthermore, a detection probe system is used to realize diameter pipe damage detection, the detection probe system includes a split fixture, the split fixture includes a first arc fixture and a second arc fixture, the first arc fixture is connected to the second arc fixture, the first arc fixture and the second arc fixture are detachable, a cavity is formed between the first arc fixture and the second arc fixture, the cavity is used to accommodate the diameter pipe to be measured, a first excitation probe and a first receiving probe are provided inside the first arc fixture, and a second excitation probe and a second receiving probe are provided inside the second arc fixture to form a one-transmit-one-receive guided wave excitation-receiving mode, the first excitation probe, the first receiving probe, the second excitation probe and the second receiving probe can all be extended into the cavity to contact the diameter pipe to be measured.

[0012] Furthermore, the first arc-shaped fixture and the second arc-shaped fixture are combined together through positioning pins, positioning holes and fixing bolts.

[0013] Furthermore, the first excitation probe, the first receiving probe, the second excitation probe and the second receiving probe are bendable flexible arc-shaped piezoelectric probe sensors.

[0014] The present invention provides a method for detecting damage to a diameter pipe. By learning multiple signal features in complex environments and correcting model deviations using physical constraints, the method significantly improves robustness in dynamic environments such as high pressure and vibration, thereby achieving high-precision damage detection.

[0015] Other features and beneficial effects of the present invention will be described in the following description, and some of the technical features and beneficial effects can be obviously derived from the description or understood by practicing the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, some of the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0017] Figure 1 is a schematic diagram of a training and prediction process of a physical constraint neural network provided by an embodiment of the present invention; Figure 2 It is a schematic diagram of the ultrasonic guided wave echo signal received by the probe; Figure 3 is a schematic diagram of the damage echo signal; Figure 4 This is a schematic diagram of the structure of the ultrasonic guided wave damage detection probe system for diameter pipes; Figure 5 This is a schematic diagram of the ultrasonic guided wave detection system for diameter pipes.

[0018] Reference numerals: 11-first arc-shaped fixture; 12-second arc-shaped fixture; 20-diameter pipe to be measured; 21-first excitation probe; 22-first receiving probe; 31-second excitation probe; 32-second receiving probe; 40-locating pin; 41-locating hole; 42-fixing bolt; 50-preload adjustment screw; 51-compression spring; 52-pressure plate. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments; the technical features designed in different implementation modes of the present invention described below can be combined with each other as long as they do not conflict with each other; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0020] In the description of the present invention, it should be understood that the terms "center", "lateral", "up", "down", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or component referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first" and "second" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more. In addition, the term "including" and any variations thereof all mean "at least including".

[0021] See also Figures 1 to 5 , Figure 1is a schematic diagram of the training and prediction process of a physical constraint neural network provided by an embodiment of the present invention. Figure 2 This is a schematic diagram of the ultrasonic guided wave echo signal received by the probe. Figure 3 is a schematic diagram of the damage echo signal. Figure 4 This is a schematic diagram of the structure of the ultrasonic guided wave damage detection probe system for diameter pipes. Figure 5 Schematic diagram of a bore tube ultrasonic guided wave detection system. As shown in the figure, a bore tube damage detection method provided by an embodiment of the present invention includes the following steps: The baseline signal is collected from the undamaged pipeline, and corresponding ultrasonic guided wave signals are collected for different damage conditions. These ultrasonic guided wave signals carry damage information. Multiple sets of ultrasonic guided wave signals corresponding to different damage locations (distances) and sizes (such as crack lengths or hole diameters) are used for subsequent feature extraction and dataset construction. In some embodiments, wavelet threshold denoising can be performed on the collected raw guided wave signals to eliminate the effects of high-frequency electromagnetic interference and mechanical vibration noise.

[0022] A first input feature, a second input feature, a third input feature and a fourth input feature are extracted from the ultrasonic guided wave signal carrying damage information.

[0023] The first input feature is the damage echo reflection coefficient R, , A1 is the amplitude of the damage reflection echo, and A0 is the amplitude of the incident wave. Figure 2 The figure shows the ultrasonic guided wave echo signal received by the piezoelectric probe. The damage echo reflection coefficient R is obtained by calculating the ratio of the damage reflection echo amplitude A1 to the incident wave amplitude A0.

[0024] The second input feature is the damage time domain peak amplitude P. The received ultrasonic guided wave signal with damage information is subtracted from the reference signal collected in the damage-free state to obtain the damage reflection signal. Figure 3 As shown, the maximum absolute value of the damage reflection signal in the time domain is extracted to obtain the damage time domain peak amplitude P.

[0025] The third input feature is the arrival time T, which is the time difference between the excitation signal and the first reflected wave peak. The relationship between T and the distance D of the damaged distance sensor is: , is the wave speed.

[0026] The fourth input feature is the damage time domain signal energy E, and the integral of the square of the damage reflection time domain signal is calculated. , where x is the damage reflection signal, N is the total length of the damage reflection signal, and n is the sampling point.

[0027] In some embodiments, Z-score normalization is performed on the first input feature, the second input feature, the third input feature, and the fourth input feature to perform data normalization to prevent dimensional differences from affecting training.

[0028] A fully connected neural network is used. The fully connected neural network mainly consists of an input layer, a hidden layer, and an output layer. The input layer includes 4 nodes, which correspond to the first input feature, the second input feature, the third input feature, and the fourth input feature respectively. The number of hidden layers and the number of nodes in each layer are adjusted according to the training effect. The activation function uses the ReLu function. The output layer includes 2 nodes, which correspond to the damage distance prediction value predicted by the neural network. and the predicted value of lesion size .

[0029] Embed physical information constraints, the physical constraints are: damage echo reflection coefficient R and damage size Satisfies the monotonically increasing relationship, that is, , where S is .

[0030] The loss function is introduced. The total loss function uses weighted summation to balance the main loss and physical constraints. A mixed loss function of physical loss + data loss is introduced to train the neural network. The total loss function L total for:

[0031]

[0032]

[0033] Among them, L main The main loss, L constraint is the physical constraint loss, λ is a hyperparameter (such as 0, 1), and the formula N represents the total number of training samples, i For the current sample, D pred is the damage distance prediction value predicted by the neural network, S pred Represents the predicted value of the damage size, D true represents the true value of the damage distance, S true Represents the true value of the damage size.

[0034] The final output is the true value of the damage distance and the true value of the damage size .

[0035] During the data training process, the data sample set can be divided into 70% as the training set, 15% as the validation set, and 15% as the test set. Figure 1 As shown, using the Adam optimizer (learning rate 1e -4 ), with a batch size of 32, and trained for 200 rounds based on a mixed dataset (including corrosion, crack, and weld defect samples), using the early stopping method to monitor damage on the validation set. Subsequently, the trained neural network model was deployed to the host computer software processing center, where the collected ultrasonic guided wave signals were preprocessed and feature extracted. The preprocessed feature vectors [R, P, T, E] were input into the trained neural network, which output the true value of the damage distance. and the true value of the damage size .

[0036] In some embodiments, as Figure 4 As shown, a detection probe system is used to detect diameter pipe damage. The detection probe system includes a split fixture, which includes a first arc fixture 11 and a second arc fixture 12. The first arc fixture 11 is connected to the second arc fixture 12. The first arc fixture 11 and the second arc fixture 12 are detachable. A cavity is formed between the first arc fixture 11 and the second arc fixture 12, and the cavity is used to accommodate the diameter pipe 20 to be measured. A first excitation probe 21 and a first receiving probe 22 are provided inside the first arc fixture 11, and a second excitation probe 31 and a second receiving probe 32 are provided inside the second arc fixture 12 to form a one-transmit-one-receive guided wave excitation-receiving mode. The first excitation probe 21, the first receiving probe 22, the second excitation probe 31 and the second receiving probe 32 can all be extended into the cavity to contact the diameter pipe 20 to be measured.

[0037] Furthermore, the first arc-shaped fixture 11 and the second arc-shaped fixture 12 are combined together by a positioning pin 40 , a positioning hole 41 and a fixing bolt 42 .

[0038] Furthermore, the first excitation probe 21 , the first receiving probe 22 , the second excitation probe 31 and the second receiving probe 32 are bendable flexible arc-shaped piezoelectric probe sensors.

[0039] Furthermore, the detection probe system also includes a preload adjustment screw 50. The movement of the pressure plate 52 is controlled by adjusting the preload adjustment screw 50, thereby adjusting the tightness of the spring. The pressure plate 52 is installed on the probe base, and four compression springs 51 are installed on the pressure plate 52. The pressure plate 52 is fixed to the arc-shaped probe base shell by screws. Specifically, the pressure plate 52 is moved downward or upward by adjusting the knob, and acts on the compression spring 51, thereby adjusting the tightness of the compression spring 51. The pressure plate 52 acts on the probe to achieve compression or separation between the probe and the pipe wall. The probe is compressed with the pipe wall to be tested, thereby forming a mechanical coupling to excite ultrasonic guided waves. When not in use, the preload adjustment screw 50 can be loosened, and the probe is retracted under the action of the compression spring 51 to protect the probe surface chip.

[0040] Two probe holes are reserved on the first arc-shaped fixture 11 and the second arc-shaped fixture 12 , through which the probes are in direct contact with the wall of the small-diameter tube to achieve mechanical coupling.

[0041] The piezoelectric chip of the probe adopts a thickness shear type PZT piezoelectric ceramic chip to excite the T (0,1) mode guided wave signal.

[0042] The working surface of the probe protrudes 0.5mm from the fixture, and is equipped with a silicone rubber buffer layer. It is connected to the external system through elastic contact terminals to form a dynamic impedance stable interface.

[0043] During testing, the first arc-shaped fixture 11 and the second arc-shaped fixture 12 are combined through the positioning pin 40 to place the tested pipe in the cavity. After the fixture is closed, the preload adjustment screw 50 cooperates with the pressure plate 52 to achieve a close fit between the probe and the pipe wall. Figure 5 The figure shows a schematic diagram of the ultrasonic guided wave detection system. The excitation probe is electrically connected to the power amplifier and signal generator in sequence, while the receiving probe is electrically connected to the signal acquisition device and the host computer, enabling the excitation and reception processing of the guided wave signal. An external signal generator is connected to the power port of the excitation piezoelectric probe via a wire, thereby stimulating the ultrasonic guided wave signal. The signal acquisition device collects the reflected signal through the receiving probe and sends it to the host computer software for analysis of damage characteristics, enabling pipeline damage detection and identification of damage location and size.

[0044] In some embodiments, a split fixture can be placed over the surface of the small-diameter pipe to be tested. The preload adjustment knob and spring pressure ensure a tight, conformal coupling between the arc-shaped piezoelectric probe and the pipe wall. The probe module utilizes a two-transmitter, two-receiver configuration, with an excitation probe and a receiving probe in each of the upper and lower fixtures. The probe excitation frequency range is 50kHz-200kHz, stimulating torsional T(0,1) mode guided waves.

[0045] An external signal source generates a high-voltage pulse excitation signal, which, after power amplification, drives the excitation probe to generate ultrasonic guided waves. A receiving probe collects the reflected echo signal propagating through the pipeline and stores the raw waveform data using a high-speed data acquisition card (sampling rate ≥ 10MS / s). Experiments collected a set of ultrasonic guided wave signals from an undamaged pipeline as a baseline signal, as well as multiple sets of ultrasonic guided wave signals corresponding to different damage locations (distances) and sizes (such as crack lengths or hole diameters) for subsequent feature extraction and dataset construction.

[0046] It should be noted that the diameter tube mentioned in the present invention may also refer to tubular objects such as pipes and conduits, and the size of the diameter tube is not limited to less than 25 mm.

[0047] In summary, the present invention provides a method for detecting damage in a diameter pipe. By learning multiple signal features in complex environments and correcting model deviations using physical constraints, the method significantly improves the robustness in dynamic environments such as high pressure and vibration, thereby achieving high-precision damage detection.

[0048] In addition, those skilled in the art should understand that, although there are many problems in the prior art, each embodiment or technical solution of the present invention may be improved in only one or several aspects, without having to simultaneously solve all the technical problems listed in the prior art or background art. Those skilled in the art should understand that any content not mentioned in a claim should not be construed as limiting the claim.

[0049] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting damage to a pipe, characterized in that: The diameter pipe damage detection method comprises the following steps: Collecting the pipeline in an undamaged state as a reference signal, and collecting the corresponding ultrasonic guided wave signals of the pipeline in different damage states, wherein the ultrasonic guided wave signals carry damage information; extracting a first input feature, a second input feature, a third input feature, and a fourth input feature from the ultrasonic guided wave signal carrying damage information; The first input feature is the damage echo reflection coefficient R, , A1 is the amplitude of the damage reflection echo, and A0 is the amplitude of the incident wave; The second input feature is the damage time domain peak amplitude P. The received ultrasonic guided wave signal with damage information is subtracted from the reference signal collected in the damage-free state to obtain a damage reflection signal. The maximum absolute value of the damage reflection signal in the time domain is extracted to obtain the damage time domain peak amplitude P. The third input feature is the arrival time T, which refers to the time difference between the emission of the excitation signal and the peak value of the first reflected wave. The relationship between T and the distance D of the damaged distance sensor is: , is the wave speed; The fourth input feature is the damage time domain signal energy E, , where x is the damage reflection signal, N is the total length of the damage reflection signal, and n is the sampling point; A fully connected neural network is used. The fully connected neural network mainly consists of an input layer, a hidden layer, and an output layer. The input layer includes 4 nodes, which correspond to the first input feature, the second input feature, the third input feature, and the fourth input feature, respectively. The number of hidden layers and the number of nodes in each layer are adjusted according to the training effect. The activation function adopts the ReLu function. The output layer includes 2 nodes, which correspond to the damage distance prediction value predicted by the neural network. and predicted value of lesion size ; Embed physical information constraints, the physical constraints are: damage echo reflection coefficient R and damage size Satisfies the monotonically increasing relationship, that is, , where S is ; Introducing loss function; The final output is the true value of the damage distance and the true value of the damage size .

2. A method for detecting damage to a bore pipe according to claim 1, characterized in that: The collected original waveguide signal is denoised by wavelet threshold.

3. The method for detecting damage to a bore pipe according to claim 1, wherein: Z-score normalization processing is performed on the first input feature, the second input feature, the third input feature, and the fourth input feature.

4. A method for detecting damage to a bore pipe according to claim 1, characterized in that: A detection probe system is used to detect damage to the diameter pipe. The detection probe system includes a split fixture, which includes a first arc fixture and a second arc fixture. The first arc fixture is connected to the second arc fixture, and the first arc fixture and the second arc fixture are detachable. A cavity is formed between the first arc fixture and the second arc fixture, and the cavity is used to accommodate the diameter pipe to be measured. A first excitation probe and a first receiving probe are provided inside the first arc fixture, and a second excitation probe and a second receiving probe are provided inside the second arc fixture to form a one-transmit-one-receive guided wave excitation-receiving mode. The first excitation probe, the first receiving probe, the second excitation probe and the second receiving probe can all be extended into the cavity to contact the diameter pipe to be measured.

5. A method for detecting damage to a bore pipe according to claim 4, characterized in that: The first arc-shaped fixture and the second arc-shaped fixture are combined together through positioning pins, positioning holes and fixing bolts.

6. A method for detecting damage to a bore pipe according to claim 4, characterized in that: The first excitation probe, the first receiving probe, the second excitation probe and the second receiving probe adopt bendable flexible arc-shaped piezoelectric probe sensors.

7. A method for detecting damage to a bore pipe according to claim 4, characterized in that: The detection probe system further comprises a preload adjustment screw, which is used to control the movement of the pressure plate and thereby adjust the tightness of the spring.

8. The method for detecting damage to a bore pipe according to claim 1, wherein: In the step of introducing the loss function, the total loss function L total for: Among them, L main Main damage, L constraint is the physical constraint loss, λ is a hyperparameter, and the formula N represents the total number of training samples, i For the current sample, D pred is the damage distance prediction value predicted by the neural network, S pred Represents the predicted value of the damage size, D true represents the true value of the damage distance, S true Represents the true value of the damage size.

Images