Quality ultrasonic non-destructive testing method and system for water pipes

Abstract

This invention discloses an ultrasonic non-destructive testing method and system suitable for water pipes, belonging to the field of ultrasonic non-destructive testing technology. It includes: acquiring full-matrix capture data arranged along the pipe axis and extracting time-domain ultrasonic echo sequences from each test section; performing adaptive threshold segmentation on each echo sequence based on local peak density to separate suspected defect echo segments; constructing a defect echo morphology membership vector based on the waveform rise steepness, fall tail length, and echo width of each segment; arranging the membership vectors of each section according to spatial position to form a defect morphology evolution sequence, and performing variational mode decomposition on it to obtain global trend components and local fluctuation components; jointly determining the defect type and axial extension endpoint based on the slope sign change points of the global trend component and the zero-crossing point distribution density of the local fluctuation component. This invention can accurately identify pipe defect types and axial ranges, with high detection reliability.

Description

Technical Field

[0001] This invention relates to the field of ultrasonic nondestructive testing technology, specifically to a method and system for ultrasonic nondestructive testing of the quality of water pipes. Background Technology

[0002] During long-term service, water pipes are susceptible to pitting and diffusion corrosion due to corrosion from the transported medium, fluid erosion, and external loads. These defects propagate along the pipe's axial direction. If not detected and identified in a timely manner, they can reduce the pipe wall's pressure-bearing capacity, leading to pipe rupture and leakage risks. Therefore, a non-destructive assessment of the pipe's internal quality condition is necessary without excavation or disruption of operation.

[0003] Current ultrasonic nondestructive testing techniques mostly employ a single probe or conventional phased array to scan point-by-point along the pipe's axial direction, extracting the portion of the echo exceeding a fixed amplitude threshold as suspected defect signals. In practical engineering, the inhomogeneity of pipe materials, variations in coupling conditions, and internal grain noise within the pipe wall can cause significant fluctuations in the background amplitude of ultrasonic echoes. A fixed threshold can easily lead to numerous false detections in areas with high background noise, while in areas with low noise, it may miss small-amplitude true defect echoes. Multi-channel echo sequences extracted directly from full-matrix capture data contain a large amount of redundant information, and existing methods lack quantitative characterization of the essential features of the echo morphology, failing to effectively and stably distinguish defect echoes from noise under strong background interference.

[0004] In determining the axial extent and type of defects, existing technologies often rely on observing the echo characteristics of a single cross-section or manually analyzing sequence data. This makes it difficult to simultaneously extract macroscopic trend changes and microscopic morphological fluctuations from the continuous spatial echo evolution along the pipeline's longitudinal direction. For axially extending diffusion corrosion and discretely distributed localized pitting corrosion, they exhibit different patterns in the longitudinal evolution of the echoes. Traditional methods lack the means to simultaneously separate the trend component and fluctuation component from the echo evolution sequence and make a joint determination. This leads to inaccurate positioning of the defect's axial extension endpoint and an inability to reliably distinguish between diffusion corrosion and localized pitting corrosion. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for ultrasonic nondestructive testing of water pipes, which solves the problems of false detection and missed detection of defect echoes caused by background amplitude fluctuations in ultrasonic testing of water pipes, as well as the problem of not being able to accurately distinguish defect types and determine axial extension endpoints from echo evolution sequences.

[0006] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides an ultrasonic non-destructive testing method suitable for water pipe quality, comprising: Acquire full matrix capture data from multiple ultrasonic phased array probes arranged along the pipeline axis, and extract the time-domain ultrasonic echo sequence of each detection section from it; For each time-domain ultrasonic echo sequence, adaptive threshold segmentation based on local peak density is performed to separate suspected defect echo segments; Based on the waveform rising edge steepness, falling edge tail length, and echo width of each suspected defect echo segment, construct the defect echo morphology membership vector for that segment. The defect echo morphology membership vectors of all detected sections are arranged according to their spatial positions to form a defect morphology evolution sequence in the longitudinal direction of the pipeline. Variational mode decomposition is then performed on this sequence to separate the global trend component and the local fluctuation component. Based on the slope sign change points of the global trend component and the zero-crossing point distribution density of the local fluctuation component, the defect type and axial extension endpoint are jointly determined.

[0007] The above method fully acquires the acoustic field information of the pipe section by capturing full matrix data, and combines multi-level processing of adaptive threshold segmentation, morphological membership vector and variational mode decomposition to effectively distinguish between diffusion corrosion defects and local pitting defects. At the same time, it accurately identifies the extension range of defects along the pipe axis, thereby improving the reliability and quantification accuracy of ultrasonic nondestructive testing.

[0008] As a technical solution of the present invention, the specific process of acquiring full matrix capture data from multiple ultrasonic phased array probes arranged along the pipeline axis is as follows: each ultrasonic phased array probe is controlled to act as a transmitting probe in sequence, and all other probes act as receiving probes simultaneously, acquiring the full-channel time-domain signal under each transmission, and storing the full matrix capture data of each detection section in matrix form; the full matrix capture data of each detection section is mapped onto a two-dimensional spatial grid composed of the pipeline axis and circumference according to the probe geometric coordinates, generating the original sound field spatiotemporal matrix of that section. Preferably, when mapping to the two-dimensional spatial grid, bilinear interpolation is used to fill the signal amplitude on non-grid nodes to obtain a normalized original sound field spatiotemporal matrix, thereby ensuring the spatial continuity of the sound field data.

[0009] The preferred step for extracting the time-domain ultrasonic echo sequence for each detection cross-section is as follows: For the original acoustic field spatiotemporal matrix of each detection cross-section, median fusion is performed on the time-domain signals of multiple receiving channels at the same axial position along the circumferential direction to obtain a single-channel fused echo at that axial position; the single-channel fused echoes at all axial positions within the same detection cross-section are then connected in series according to the probe arrangement order to form the time-domain ultrasonic echo sequence for that detection cross-section. Circumferential median fusion effectively suppresses uncorrelated noise and improves the signal-to-noise ratio of the echo sequence.

[0010] The specific method for performing adaptive threshold segmentation based on local peak density for each time-domain ultrasound echo sequence is as follows: The global amplitude mean and the peak frequency within a local window are calculated for each time-domain ultrasound echo sequence. Using the global amplitude mean as the initial baseline, the segmentation threshold is dynamically adjusted according to the peak frequency within the local window, increasing the threshold in areas of high peak density and decreasing it in areas of low peak density. The dynamically adjusted segmentation threshold is then used to binarize the time-domain ultrasound echo sequence, marking intervals that continuously exceed the threshold as a suspected defect echo segment. The peak frequency within the local window is calculated as follows: A time window of a preset length is taken, centered on the current sampling point. The number of local maxima within this window is counted, and this number is divided by the window length to obtain the peak frequency. This adaptive thresholding mechanism can highlight genuine defect echoes under strong background noise while suppressing spurious peak interference.

[0011] Preferably, after separating the suspected defect echo segments, the ratio of the start and end time width of each suspected defect echo segment to the time interval of adjacent suspected defect echo segments is calculated. If this ratio is lower than a preset interference screening threshold, the suspected defect echo segment is identified as noise interference and removed. All remaining suspected defect echo segments are then renumbered in chronological order and used as valid input for subsequent steps. This further filters out isolated random electrical noise and minute impurity echoes, reducing the false alarm rate.

[0012] The process of constructing the defect echo morphology membership vector is as follows: For each suspected defect echo segment, the median slope of the waveform rising from the amplitude to the peak before the peak is calculated as the rising edge steepness; the time required for the waveform to fall to half the peak after the peak is calculated as the falling edge tail length; and the time span between the two half-amplitude points of the waveform is calculated as the echo width. The rising edge steepness, falling edge tail length, and echo width are input into a preset defect morphology fuzzy membership function to obtain their respective morphology membership scores. The three scores are combined to form the defect echo morphology membership vector for that segment. The preset defect morphology fuzzy membership function is preferably an S-shaped membership function, where the rising edge steepness, falling edge tail length, and echo width correspond to their respective membership mapping curves, and the threshold parameters of the three curves are pre-calibrated based on the pipe material and wall thickness. Quantifying the echo waveform characteristics through the morphology membership vector allows for a more refined characterization of acoustic impedance changes caused by defects, improving the ability to identify different types of defects such as corrosion and cracks.

[0013] The preferred steps for forming and decomposing the longitudinal defect morphology evolution sequence of the pipeline are as follows: Following the spatial coordinate order along the pipeline axis for each detection section, extract the maximum vector magnitude from the defect echo morphology membership vector of all suspected defect echo segments in each detection section, using this as the cross-sectional defect characterization value for that section. Arrange the cross-sectional defect characterization values ​​of all detection sections in ascending order of axial coordinates to form the longitudinal defect morphology evolution sequence of the pipeline. Input this sequence into a variational mode decomposition model, setting the number of decomposition modes to two, and initializing the center frequency and bandwidth constraints of each mode. Iteratively update each mode and its center frequency until the convergence condition is met, obtaining the first mode as the global trend component and the second mode as the local fluctuation component. The obtained global trend component reflects the overall trend of defect severity along the pipeline, while the local fluctuation component captures the detailed fluctuations within the defect, achieving decoupling of multi-scale information on defect morphology.

[0014] The process of jointly determining the defect type and axial extension endpoint is as follows: First-order difference calculation is performed on the global trend component, and the sign change points where the difference value changes from positive to negative or from negative to positive are recorded. The axial position corresponding to these change points is used as candidate boundary points for the defect type. The number of zero-crossing points of the local fluctuation component within the neighborhood of the candidate boundary points is counted, and the ratio of this number to the neighborhood length is calculated as the zero-crossing point distribution density. If the distribution density exceeds a preset threshold, the defect type is determined to be a diffusion-type corrosion defect; otherwise, it is determined to be a local pitting corrosion defect. This joint determination method comprehensively utilizes trend reversal and local detail features, which can significantly improve the accuracy of defect classification.

[0015] Furthermore, when the defect type is determined to be diffusion-type corrosion, the point where the slope sign changes in the global trend component is taken as the starting endpoint of the defect, and the lowest point of the last monotonically decreasing interval of the global trend component sequence is taken as the ending endpoint. The axial range between the starting and ending endpoints is output as the axial extension range of the defect. When the defect type is determined to be local pitting corrosion, the axial positions corresponding to the three largest positive peaks in the local fluctuation component are extracted, and their average coordinates are output as the defect center position. This provides accurate axial location and extension range estimation for different types of defects, offering a reliable basis for pipeline safety assessment and repair decisions.

[0016] This invention also provides an ultrasonic non-destructive testing system suitable for water pipes, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of the aforementioned detection method. This system achieves high-precision detection, classification, and axial range determination of defects in water pipes through multi-level analysis of the full matrix captured data, including adaptive threshold segmentation, morphological membership vector construction, and variational mode decomposition. It has the advantages of reliable detection and strong anti-interference capability.

[0017] The technical effects and advantages provided by the present invention in the above technical solution are as follows: An adaptive threshold segmentation method based on local peak density is employed. This method calculates the global amplitude mean and the peak frequency within a local window of the time-domain ultrasonic echo sequence. Using the global amplitude mean as the initial baseline, the segmentation threshold is dynamically increased when the peak frequency within the local window rises, and correspondingly decreased in areas with sparse peaks. The dynamically adjusted threshold is then used to binarize and segment the echo sequence, marking continuous out-of-limit intervals as suspected defect echo segments. This approach adaptively changes the segmentation threshold with the local peak density of the echo, automatically suppressing false alarms caused by noise in areas with large background noise fluctuations, and improving detection sensitivity in low-density areas where weak defect echoes appear. This maintains a high capture rate for true defects while reducing the probability of false detections, effectively overcoming the problem of missed and false detections caused by the incompatibility of the pipe's acoustic field when using a fixed threshold.

[0018] By analyzing the defect evolution sequence based on variational mode decomposition, the longitudinal defect morphology evolution sequence of the pipeline is decomposed into a global trend component and a local fluctuation component. The slope sign change points of the first-order difference of the global trend component are calculated to obtain candidate locations for defect type boundaries. Simultaneously, the distribution density of zero-crossing points of the local fluctuation component in the neighborhood of the candidate points is statistically analyzed to jointly determine the defect type. When the distribution density exceeds a threshold, it is identified as a diffusion corrosion defect, and the starting and ending points of the axial extension are located using the slope sign change point and the lowest point of the monotonically decreasing interval at the end of the trend sequence. When it is below the threshold, it is identified as a local pitting corrosion defect, and the average coordinate of the positive peak value with the largest amplitude in the fluctuation component is used as the defect center location. This method simultaneously separates macroscopic trend and microscopic fluctuation modes from the continuous longitudinal echo of the pipeline. It captures the axial extension boundary of the defect using the directional change of the trend component and distinguishes the typical morphological differences between diffusion corrosion and pitting corrosion using the zero-crossing distribution density of the fluctuation component. It can automatically identify defect types and accurately extract the axial extension interval or center location without relying on manually setting the axial determination length. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0020] Figure 1 This is a flowchart of the ultrasonic non-destructive testing method for water pipe quality; Figure 2 This is a flowchart of the full matrix capture data space mapping and median fusion echo sequence generation process; Figure 3 This is a flowchart of adaptive threshold segmentation and interference removal based on local peak density; Figure 4 This is a flowchart for constructing the membership vector of defect echo morphology. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] See Figure 1 This invention provides an ultrasonic non-destructive testing method for water pipes, comprising the following steps: acquiring full-matrix capture data from multiple ultrasonic phased array probes arranged along the pipe axis, and extracting the time-domain ultrasonic echo sequence for each test section; performing adaptive threshold segmentation based on local peak density on each time-domain ultrasonic echo sequence to separate suspected defect echo segments; constructing a defect echo morphology membership vector for each suspected defect echo segment based on the waveform rising edge steepness, falling edge tail length, and echo width; arranging the defect echo morphology membership vectors of all test sections according to spatial position to form a defect morphology evolution sequence along the pipe longitudinal direction, and separating the global trend component and local fluctuation component by performing variational mode decomposition on the sequence; and jointly determining the defect type and axial extension endpoint based on the slope sign change points of the global trend component and the zero-crossing point distribution density of the local fluctuation component.

[0023] Example 1: In specific implementation, please refer to Figure 2The process of acquiring full-matrix capture data from multiple ultrasonic phased array probes arranged along the axial direction of a pipeline is implemented as follows: An array of ultrasonic phased array probes arranged circumferentially along the pipeline is controlled, with each ultrasonic phased array probe sequentially acting as a transmitting probe. When one ultrasonic phased array probe is selected as a transmitting probe, all other ultrasonic phased array probes in the array simultaneously act as receiving probes. The transmitting probe emits ultrasonic pulse signals towards the pipeline wall. These ultrasonic pulse signals propagate within the pipeline wall and generate reflections or scattering. All receiving probes synchronously acquire the full-channel time-domain signals of their respective channels. For a single transmission event, the full-channel time-domain signals acquired by all receiving probes are combined into a two-dimensional data matrix according to the correspondence between the receiving probe number and the transmitting probe number, denoted as the submatrix of a single transmission event. The process iterates through all ultrasonic phased array probes, using each as a transmitting probe and performing both transmitting and receiving processes. This yields a set of sub-matrices equal in number to the number of ultrasonic phased array probes. All sub-matrices are arranged and combined according to the transmitting probe numbers to form full matrix capture data corresponding to a single detection cross-section. This full matrix capture data is stored in matrix form, with the two dimensions corresponding to the transmitting and receiving probe numbers, respectively. Each element within the matrix represents a time-domain signal sequence. In some embodiments, the ultrasonic phased array probe array is arranged circumferentially around the pipe to form a detection cross-section. Multiple detection cross-sections are set along the pipe axis, each corresponding to an independent ultrasonic phased array probe array. The above process is performed independently on each detection cross-section to obtain full matrix capture data for multiple detection cross-sections.

[0024] The process of mapping the full matrix capture data of each detection section onto a two-dimensional spatial grid composed of the pipe's axial and circumferential directions according to the probe's geometric coordinates is performed on a per-detection-section basis. A two-dimensional spatial grid is pre-established, covering the pipe wall region of the detection section. The two coordinate directions of the two-dimensional spatial grid are the pipe's axial direction and the pipe's circumferential direction. The grid nodes of the two-dimensional spatial grid are uniformly distributed according to preset axial and circumferential angular spacing. The physical geometric coordinates of each ultrasonic phased array probe in actual space, including axial and circumferential coordinates, are read from the full matrix capture data. Based on the geometric coordinates of the transmitting and receiving probes, the projection intersection point of the acoustic beam path corresponding to each transmit-receive combination on the two-dimensional spatial grid is determined. For a time-domain signal sequence acquired by a transmit-receive combination, the signal amplitude at each sampling time in the time-domain signal sequence is allocated to the corresponding spatial grid nodes according to the correspondence between acoustic time and spatial location. For spatial location points that do not directly fall on grid nodes, bilinear interpolation is used to fill in the signal amplitude at non-grid nodes. The calculation formula for bilinear interpolation is: in, Represents the axial coordinates of the spatial point to be interpolated. Represents the circumferential coordinates of the spatial point to be interpolated; This represents the axial coordinate of the first adjacent grid node surrounding the point to be interpolated in space. This represents the axial coordinate of the second adjacent grid node surrounding the point to be interpolated in space. and satisfy Furthermore, the two coordinates are those of two adjacent nodes on the axial grid; This represents the circumferential coordinates of the first adjacent grid node surrounding the point to be interpolated in space. This represents the circumferential coordinates of the second adjacent grid node surrounding the point to be interpolated. and satisfy Furthermore, the two coordinates are those of two adjacent nodes on the circumferential grid; Indicates coordinates as The signal amplitude on the adjacent grid nodes, Indicates coordinates as The signal amplitude on the adjacent grid nodes, Indicates coordinates as The signal amplitude on the adjacent grid nodes, Indicates coordinates as The signal amplitudes on the adjacent grid nodes, all four signal amplitudes mentioned above have been allocated during the mapping process from the full matrix capture data to the grid nodes; The coordinates obtained by bilinear interpolation are: The signal amplitude on the non-mesh nodes is obtained by filling all non-mesh nodes on the two-dimensional spatial grid that are not directly assigned signal amplitudes using the bilinear interpolation method described above. This yields a normalized original sound field spatiotemporal matrix, where each row of the original sound field spatiotemporal matrix corresponds to an axial position, each column corresponds to a circumferential position, and the matrix elements are the signal amplitudes.

[0025] The process of extracting the time-domain ultrasonic echo sequence for each detection section is performed on a normalized original acoustic field spatiotemporal matrix. Time-domain signals from multiple receiving channels at the same axial position are extracted along the circumferential direction of the original acoustic field spatiotemporal matrix. Here, the time-domain signals of the receiving channels correspond to the signal sequences at the circumferential column indices of the row containing that axial position in the original acoustic field spatiotemporal matrix. The multiple circumferential time-domain signal sequences extracted at the same axial position are sorted by their median amplitude at each identical sampling time point. The median of the sorted values ​​is taken as the fused amplitude, thus obtaining the single-channel fused echo for that axial position. The single-channel fused echo is a time-domain sequence with the same number of sampling points as the original time-domain signal. The above-described median fusion operation along the circumferential direction is performed on all axial positions covered by the original acoustic field spatiotemporal matrix to obtain single-channel fused echoes for multiple axial positions. The single-channel fused echoes from all axial positions within the same detection cross-section are strung together in the axial order of the ultrasonic phased array probe, with the first axial position's single-channel fused echo placed first, followed by subsequent axial positions' single-channel fused echoes, forming the time-domain ultrasonic echo sequence for that detection cross-section. The time-domain ultrasonic echo sequence is a long sequence, its length equal to the number of axial positions multiplied by the length of a single single-channel fused echo.

[0026] Example 2: In specific implementation, please refer to Figure 3 The process of performing adaptive threshold segmentation based on local peak density to separate suspected defective echo segments for each time-domain ultrasound echo sequence is implemented as follows: The global amplitude mean of the time-domain ultrasound echo sequence is calculated. The global amplitude mean is the arithmetic mean of the absolute values ​​of the amplitudes of all sampling points in the time-domain ultrasound echo sequence. The time-domain ultrasound echo sequence is denoted as... ,in Indicates the sampling time index. , Let be the total number of sampling points in the time-domain ultrasound echo sequence, then the global amplitude mean is... The following formula is used to derive: Simultaneously, the peak frequency of the time-domain ultrasound echo sequence within a local window is calculated. The peak frequency within the local window is calculated as follows: based on the current sampling point... Centered on a time window of a preset length, the window length is denoted as . The sampling point range covered by the time window is ,in This indicates a floor function. Within this time window, the number of all local maxima is counted. The criterion for determining a local maximum is: for a given sampling point within the time window... If its amplitude Simultaneously satisfy and Then it is determined Let be a local maximum point. The number of local maxima within the time window is counted and denoted as . .Will Divide by window length Get the current sampling point Frequency of peak occurrence within the corresponding local window Its expression is: Among them, window length The value of is determined based on the nominal width of the ultrasonic pulse and the sampling period of the time-domain ultrasonic echo sequence, so that the window length The corresponding physical time length covers the duration of a single complete ultrasound pulse. When the sampling period is 10 nanoseconds and the ultrasound pulse width is 500 nanoseconds, the window length is... Set to 50 sampling points.

[0027] global amplitude mean As the initial benchmark for segmentation threshold, it is based on the frequency of peak occurrence within the local window. Dynamically adjust the current sampling point The segmentation threshold is set at a specific point. The dynamically adjusted rule increases the segmentation threshold in regions with high peak frequency and decreases it in regions with low peak frequency. One implementation method is to set an adjustment coefficient. , The value is a real number greater than 0, determined by statistical analysis of the difference in echo peak distribution between intact and typical defective areas of the pipeline. The value is a constant between 0.5 and 2.0. The dynamically adjusted segmentation threshold is then... Determined by the following formula: in, Represents all sampling points in the entire time-domain ultrasound echo sequence The mean. When Greater than hour, It has increased from the baseline initial value; when Less than hour, The value was reduced from the baseline initial value.

[0028] Using dynamically adjusted segmentation thresholds The time-domain ultrasound echo sequence is segmented into points using binarization. For each sampling point in the time-domain ultrasound echo sequence... its amplitude and Comparison: If Then mark the sampling point as a valid signal point; if If a sampling point is marked as an invalid signal point, then the interval of sampling points that are continuously marked as valid signal points is determined as a suspected defect echo segment, and the start and end times of the suspected defect echo segment are recorded.

[0029] After separating all suspected defect echo segments, the following process is performed: Calculate the start and end time widths of each suspected defect echo segment, defined as the difference between the end time and start time of the suspected defect echo segment. Simultaneously, calculate the time interval between the suspected defect echo segment and its next adjacent suspected defect echo segment, defined as the difference between the start time of the next suspected defect echo segment and the end time of the current suspected defect echo segment. Calculate the ratio of the start and end time widths to the time interval, and compare this ratio to a preset interference screening threshold. The interference screening threshold is set based on the characteristics of typical noise signals and actual defect signals collected statistically, and its value ranges from 0.1 to 0.5. If the ratio of the start and end time widths to the time interval is lower than the interference screening threshold, the corresponding suspected defect echo segment is identified as noise interference and removed; otherwise, the corresponding suspected defect echo segment is retained. The above elimination process is applied to all suspected defect echo segments contained in the current time-domain ultrasound echo sequence. Then, all the suspected defect echo segments that remain after elimination are renumbered in chronological order, and the renumbered suspected defect echo segments are used as valid inputs for subsequent steps.

[0030] Example 3: In specific implementation, please refer to Figure 4 The process of constructing a membership vector of defect echo morphology based on the steepness of the rising edge, the length of the falling edge tail, and the echo width of each suspected defect echo segment involves performing the following operations independently for each suspected defect echo segment.

[0031] For a suspected defective echo segment, first locate the waveform peak point of the suspected defective echo segment. Search for the sampling point with the largest absolute amplitude within the time domain sampling point interval covered by the suspected defective echo segment, and record the time index of this sampling point as the peak time. The amplitude corresponding to this sampling point is recorded as the peak amplitude. .

[0032] The rise edge steepness is calculated as follows: on the time-domain waveform of the suspected defect echo segment, the time from the start time of the suspected defect echo segment to the peak time is extracted. The waveform interval between the two points is designated as the rising edge interval. Within the rising edge interval, the waveform slope at each sampling point is calculated. The slope is calculated by dividing the amplitude difference between the sampling point and the previous sampling point by the sampling time interval. After obtaining the slope value sequence of all sampling points within the rising edge interval, the median of this slope value sequence is taken. This median slope is used as the rising edge steepness of the suspected defect echo segment, denoted as . .

[0033] The falling edge tail length is calculated as follows: on the time-domain waveform of the suspected defective echo segment, the peak time is taken as the starting point. The waveform interval between the end time of the suspected defective echo segment and the end time of the next waveform segment is defined as the falling edge interval. Within the falling edge interval, the waveform amplitude is first observed to drop to its peak amplitude. The sampling point corresponding to half of the value is found, i.e., the point that satisfies the condition. The first sampling point, whose time index is denoted as . .calculate and The time difference between them is the length of the falling edge tail, denoted as . , If the waveform amplitude does not decrease to half of the peak amplitude within the falling edge interval, then the length of the falling edge interval is taken as the falling edge tail length.

[0034] The echo width is calculated as follows: On the entire time-domain waveform of the suspected defective echo segment, find two sampling points where the waveform amplitude drops to half its peak value on both sides. At the peak time... On the left, find the condition that is met. The last sampling point, i.e., the left half amplitude value point, is denoted by time . At peak time On the right, find the condition that is met. The last sampling point, i.e., the right half-amplitude value point, is denoted by time . .calculate and The time span between these two points is called the echo width, denoted as . , .

[0035] Obtain the steepness of the rising edge Length of the trailing edge and echo width Next, these three feature quantities are input into the preset defect morphology fuzzy membership function. The defect morphology fuzzy membership function adopts the S-shaped membership function. The mathematical expression is: in, The waveform features representing the input are used to calculate the membership score for the steepness of the rising edge. for When calculating the membership score of the trailing edge length, for When calculating the membership score of echo width for ; To control the steepness factor of the S-shaped membership function curve, The range of values ​​for is positive real numbers. The specific values ​​are pre-calibrated based on the pipe material and wall thickness; To control the threshold center parameter of the S-shaped membership function curve, The range of values ​​for is real numbers. The specific values ​​are also pre-calibrated based on the pipe material and wall thickness. This is the base of the natural logarithm. For rising edge steepness, falling edge tail length, and echo width, separate sets of steepness factors and threshold center parameters are used, i.e., the membership mapping curve parameters corresponding to rising edge steepness are... and The membership mapping curve parameters corresponding to the length of the falling edge tail are: and The membership mapping curve parameters corresponding to the echo width are: and The calibration process for the three sets of parameters involves collecting ultrasonic echo data from calibration specimens with standard artificial defects, extracting corresponding feature quantities, and adjusting the steepness factor and threshold center parameter to maximize the distinguishability of the membership score output by the S-shaped membership function between defective and non-defective samples, given the pipe material and wall thickness. Finally, the optimized parameter values ​​are stored as preset values.

[0036] The rise along the steepness Substitute and The sigmoid membership function with respect to the parameter yields the morphological membership score of the rising edge steepness, denoted as . ; the length of the trailing edge Substitute and The sigmoid membership function for the parameter yields the morphological membership score for the length of the falling edge tail, denoted as . ; Echo width Substitute and The S-shaped membership function with respect to parameters yields the morphological membership score of the echo width, denoted as . The three morphological membership scores are combined into a three-dimensional vector, i.e. This vector is the defect echo shape membership vector of the suspected defect echo segment. The above process is performed on all suspected defect echo segments to obtain the defect echo shape membership vector of each suspected defect echo segment.

[0037] Example 4: In practice, the process of arranging the membership vectors of the defect echo morphology of all detection sections according to their spatial positions to form a longitudinal defect morphology evolution sequence of the pipeline is achieved in the following way.

[0038] An ultrasonic phased array probe array arranged along the pipeline axis corresponds to multiple detection sections, each with a unique axial spatial coordinate. For each detection section, it contains all suspected defect echo segments processed and retained in the aforementioned steps. Each suspected defect echo segment has a constructed defect echo morphology membership vector. For a suspected defect echo segment within a given detection section, its defect echo morphology membership vector is represented as follows: Calculate the vector magnitude of the membership vector of the defect echo shape. The vector magnitude is calculated as follows: in, The vector magnitude representing the membership vector of the defect echo morphology. The morphological membership score represents the steepness of the rising edge. The morphological membership score represents the length of the falling edge tail. The morphological membership score represents the echo width.

[0039] The algorithm iterates through all suspected defect echo segments within a given detection cross-section, calculating the vector magnitude of the defect echo morphology membership vector for each segment, resulting in a set of vector magnitudes. The maximum value from this set is selected as the cross-sectional defect characterization value for that detection cross-section. If a detection cross-section contains no suspected defect echo segments, its cross-sectional defect characterization value is set to zero.

[0040] The test sections are arranged in ascending order of their spatial coordinates along the pipeline axis. The defect characterization values ​​for each section are extracted, and these values ​​are then arranged in ascending order of their axial coordinates to form a one-dimensional sequence. This sequence represents the longitudinal defect morphology evolution sequence of the pipeline, denoted as […]. ,in Indicates the axial coordinate index of the test section. , This represents the total number of cross-sections detected.

[0041] In some embodiments, axial coordinate index There is a one-to-one correspondence between the physical axial coordinates of the test sections and the physical axial coordinates of the test sections. This applies when the test sections are uniformly distributed along the axial direction and the axial spacing between adjacent test sections is a fixed value. At that time, axial coordinate index Corresponding physical axial coordinates ,in This represents the physical axial coordinate of the first detection section.

[0042] In the formation of the defect morphology evolution sequence in the longitudinal direction of the pipeline Subsequently, variational mode decomposition was performed on the defect morphology evolution sequence to separate the global trend component and the local fluctuation component. The variational mode decomposition process was implemented as follows.

[0043] Defect morphology evolution sequence The input signal is fed into the variational mode decomposition model. The number of decomposition modes in the variational mode decomposition is set to two, i.e. The number of decomposition modes will be... The basis for setting it to 2 is that the spatial variation of the defect morphology evolution sequence along the axis macroscopically manifests as two modes: a slowly changing trend component and a rapidly fluctuating detail component. These two modes can be separated into two independent modal components: the first mode corresponds to the global trend component with a larger scale of change, and the second mode corresponds to the local fluctuation component with a smaller scale of change.

[0044] In the initialization stage of the variational mode decomposition model, the time-domain signals of the two modes are initialized, and the initial values ​​are both set as zero vectors. The length of the zero vector is equal to the defect morphology evolution sequence. length Initialize the center frequencies of both modes, with initial values ​​ranging from 0 to... The initial values ​​of the center frequencies of the two modes are randomly selected within the range and denoted as follows: and The number in parentheses indicates the iteration count, with an initial iteration count of 0. Set bandwidth constraint parameters. Bandwidth constraint parameters Controlling the bandwidth of each modal component, The value of is a real number, determined by examining the spectral distribution range of the defect morphological evolution sequence. The value ranges from 500 to 5000. Set the convergence tolerance. Convergence tolerance Used to determine whether the iterative process meets the convergence condition. Values to Real numbers between.

[0045] Variational mode decomposition enters the iterative update phase, and the following operations are performed in each iteration. For For each of the *n* modes, using the estimates of other modes at the current iteration number, update the frequency domain expression of the current mode through Wiener filtering. Update the *n*th mode. When expressing the frequency domain of a mode, The value is either 1 or 2, and the calculation is performed using the following frequency domain update formula. After the frequency domain update, update the [number]th [unit]. The center frequency of each mode is updated by calculating the centroid of the power spectrum of that mode.

[0046] After each iteration updates all modes and their center frequencies, the mean square error of the time-domain signal of each mode between two adjacent iterations is calculated. The convergence tolerance is considered when the mean square error is less than the convergence tolerance. When the iterative process meets the convergence condition, the iteration stops. The time-domain signal of the first mode obtained when the convergence condition is met is taken as the global trend component, denoted as... The time-domain signal of the second mode obtained when the convergence condition is met is taken as the local fluctuation component, denoted as... Global trend components Reflecting the sequence of defect morphological evolution Macroscopic variation trend along the pipeline axis, local fluctuation components Reflecting the sequence of defect morphological evolution Detailed fluctuation characteristics along the axial direction of the pipe.

[0047] Example 5: In practice, the process of jointly determining the defect type and the axial extension endpoint based on the slope sign change points of the global trend component and the zero-crossing point distribution density of the local fluctuation component is implemented in the following manner.

[0048] Obtain the global trend component obtained from variational mode decomposition. and local fluctuation components ,in Indicates the axial coordinate index of the test section. The value ranges from 1 to integers, The total number of cross-sections detected. Global trend component. It is a one-dimensional sequence with a sequence length equal to Local fluctuation components It is a one-dimensional sequence with a sequence length equal to .

[0049] global trend components Perform first-order difference calculations. The first-order difference sequence is denoted as... The first-order difference is calculated as follows: for the axial coordinate index From 2 to Calculate at each position For axial coordinate index , The value is set to 0. This yields the first-order difference sequence. Then, traverse middle From 2 to For each element, determine whether the signs of two adjacent first-order differences have changed. The condition for determining the sign change is: when... and At that time, the determination is made at the axial coordinate index. The sign of the difference value changes from positive to negative at the point where it occurs; and At that time, the determination is made at the axial coordinate index. The sign of the difference value changes from negative to positive at a certain point. The axial coordinate index that satisfies any of the above sign change conditions... The points where the slope sign changes are recorded, and each point corresponds to a specific axial position. These axial positions are used as candidate points for defect type boundaries. All candidate points for defect type boundaries form a set, denoted as . ,in This refers to the index of the candidate point for the defect type boundary. , This represents the total number of candidate points for the defect type boundary.

[0050] For each defect type boundary candidate point Then, take its neighborhood interval. The neighborhood interval is defined as follows: based on the candidate points of the defect type boundary. Centered on a fixed number of axial coordinate indices extend to the left and right sides, with the extension length denoted as . , This indicates the number of sampling points for a single-sided extension. The value is determined based on the axial inspection resolution of the pipeline and the characteristic scale of the defect. When the axial spacing of the inspection sections is 10 mm and the typical axial dimension of the defect to be inspected is 50 mm, Setting it to 5 ensures that the physical axial length covered by the neighborhood interval matches the typical axial dimension of the defect. The axial coordinate index range covered by the neighborhood interval is... If the left boundary of the neighborhood interval is less than 1, then the left boundary is truncated to 1; if the right boundary of the neighborhood interval is greater than 1, then the left boundary is truncated to 1. Then the right boundary will be truncated as .

[0051] In the neighborhood interval Within, statistical local fluctuation components The number of zero-crossing points. The criterion for determining zero-crossing points is: for two adjacent axial coordinate indices within a neighborhood interval... and If the local fluctuation component is in The value at the location With The value at the location satisfy ,and and If they are not both zero, then it is determined that... and There exists a zero-crossing point. Examine each pair of adjacent axial coordinate indices within the neighborhood interval, count the number of pairs that satisfy the zero-crossing point criterion, and use this count as a candidate boundary point for the defect type. The number of zero-crossing points in the neighborhood is denoted as .

[0052] The formula for calculating the zero-crossing point distribution density is as follows: in, Indicates the first Distribution density of zero-crossing points in the neighborhood of candidate points at the boundary of each defect type Indicates the first The number of zero-crossing points obtained by statistically analyzing the neighborhood of each candidate point at the boundary of a defect type. The denominator represents the number of sampling points extending unilaterally from the neighborhood. and Multiplication represents the total number of axial coordinate index intervals within the neighborhood interval. Zero-crossing point distribution density. The range of values ​​for is a real number greater than or equal to 0 and less than or equal to 1.

[0053] The calculated zero-crossing point distribution density Compare with a preset distribution density threshold. The preset distribution density threshold is denoted as... , The value is obtained through statistical analysis of calibration samples with known defect types. The calibration samples include diffuse corrosion defect samples and localized pitting corrosion defect samples confirmed by destructive testing. The distribution density of zero-crossing points in the neighborhood of candidate boundary points for each defect type is extracted for both types of samples. The value that minimizes the classification error between the two types of samples is selected as the threshold. , The value of is a real number between 0.2 and 0.5. Greater than At that time, determine the candidate boundary points of the current defect type. The corresponding defect type is diffusion-type corrosion defect; when Less than or equal to At that time, determine the candidate boundary points of the current defect type. The corresponding defect type is local pitting defect.

[0054] For cases identified as diffusion-type corrosion defects, determine the axial extension endpoints of the defect. Then, assign the global trend component... All points where the slope sign changes are arranged in ascending order of axial coordinate index. The first point where the slope sign changes is selected as the starting endpoint of the diffusion-type corrosion defect. The axial position of the starting endpoint is denoted as... In the global trend component sequence Up, from the starting endpoint Begin searching in the direction of increasing axial coordinate index to identify... The monotonically decreasing interval. The method for determining a monotonically decreasing interval is: for a continuous interval of axial coordinate indices, if each axial coordinate index... place All less than If the interval is zero, then the interval is a monotonically decreasing interval. The lowest point within the last identified monotonically decreasing interval is taken as the termination point of the diffusion-type corrosion defect. The lowest point refers to the global trend component within that monotonically decreasing interval. The axial coordinate index with the smallest value is denoted as the axial position of the termination endpoint. . Start endpoint and termination endpoint The axial range between these values ​​is output as the axial extension interval of the defect, and the axial coordinate index contained within the axial extension interval is from... arrive All integers.

[0055] For cases identified as localized pitting defects, the local fluctuation component is extracted. The axial positions corresponding to the three largest positive peaks in the intermediate amplitude. (In the local fluctuation component) The search is performed to find positive peak values, which are sampling points where the local fluctuation component value is greater than zero and also greater than the values ​​of its left and right adjacent points. All found positive peak values ​​are sorted in descending order of amplitude, and the top three positive peak values ​​with the largest amplitudes are selected. The axial coordinate indices corresponding to these three positive peak values ​​are recorded, denoted as [index]. , and Calculate the arithmetic mean of these three axial coordinate indices, denoted as . , .Will Output the defect center location as a local pitting defect.

[0056] In some embodiments, the ultrasonic nondestructive testing system for water pipes includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it performs all the steps described above: acquiring full matrix capture data and extracting time-domain ultrasonic echo sequences, performing adaptive thresholding based on local peak density, constructing defect echo morphology membership vectors, forming defect morphology evolution sequences and performing variational mode decomposition, and jointly determining the defect type and axial extension endpoint.

[0057] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. An ultrasonic non-destructive testing method applicable to the quality of water pipes, characterized in that, Includes the following steps: Acquire full matrix capture data from multiple ultrasonic phased array probes arranged along the pipeline axis, and extract the time-domain ultrasonic echo sequence of each detection section from it; For each time-domain ultrasonic echo sequence, adaptive threshold segmentation based on local peak density is performed to separate suspected defect echo segments; Based on the waveform rising edge steepness, falling edge tail length, and echo width of each suspected defect echo segment, construct the defect echo morphology membership vector for that segment. The defect echo morphology membership vectors of all detected sections are arranged according to their spatial positions to form a defect morphology evolution sequence in the longitudinal direction of the pipeline. Variational mode decomposition is then performed on this sequence to separate the global trend component and the local fluctuation component. Based on the slope sign change points of the global trend component and the zero-crossing point distribution density of the local fluctuation component, the defect type and axial extension endpoint are jointly determined.

2. The ultrasonic non-destructive testing method for water pipe quality according to claim 1, characterized in that, The specific steps for acquiring full matrix capture data from multiple ultrasonic phased array probes arranged along the pipeline axis include: Each ultrasonic phased array probe is controlled to act as a transmitting probe in sequence, while all other probes act as receiving probes at the same time. The full-channel time-domain signal under each transmission is collected, and the full matrix capture data of each detection section is stored in matrix form. The full matrix capture data of each detection section is mapped onto a two-dimensional spatial grid composed of the axial and circumferential directions of the pipe according to the probe's geometric coordinates, generating the original sound field spatiotemporal matrix of that section.

3. The ultrasonic non-destructive testing method for water pipe quality according to claim 2, characterized in that, When mapping the full matrix capture data of each detection section onto a two-dimensional spatial grid formed by the axial and circumferential directions of the pipe according to the probe's geometric coordinates, bilinear interpolation is used to fill the signal amplitude on non-grid nodes to obtain a normalized original sound field spatiotemporal matrix.

4. The ultrasonic non-destructive testing method for water pipe quality according to claim 2, characterized in that, The step of extracting the time-domain ultrasonic echo sequence for each detection section specifically includes: For the original sound field spatiotemporal matrix of each detection section, the time-domain signals of multiple receiving channels at the same axial position are fused along the circumferential direction to obtain the single-channel fused echo at that axial position; The single-channel fused echoes from all axial positions within the same detection cross section are connected in series according to the probe arrangement order to form the time-domain ultrasonic echo sequence of that detection cross section.

5. The ultrasonic non-destructive testing method for water pipe quality according to claim 1, characterized in that, The step of performing adaptive threshold segmentation based on local peak density for each time-domain ultrasound echo sequence to separate suspected defect echo segments specifically includes: Calculate the global amplitude mean and the peak frequency within a local window for each time-domain ultrasound echo sequence. Use the global amplitude mean as the initial value and dynamically adjust the segmentation threshold according to the peak frequency within the local window, so that the threshold increases in the peak-dense region and decreases in the peak-sparse region. The time-domain ultrasonic echo sequence is binarized and segmented using a dynamically adjusted segmentation threshold, and intervals that continuously exceed the threshold are marked as a suspected defect echo segment.

6. The ultrasonic non-destructive testing method for water pipe quality according to claim 5, characterized in that, The peak frequency within the local window is calculated as follows: taking the current sampling point as the center, a time window of a preset length is taken, the number of local maxima within the window is counted, and the number is divided by the window length to obtain the peak frequency.

7. The ultrasonic non-destructive testing method for water pipe quality according to claim 5, characterized in that, After separating the suspected defective echo segments, the following steps are also performed: Calculate the ratio of the start and end time width of each suspected defect echo segment to the time interval of adjacent suspected defect echo segments. If the ratio is lower than the preset interference screening threshold, the suspected defect echo segment is identified as noise interference and is removed. All suspected defective echo segments retained after removal are renumbered in chronological order and used as valid input for subsequent steps.

8. The ultrasonic non-destructive testing method for water pipe quality according to claim 1, characterized in that, The step of constructing the defect echo morphology membership vector of each suspected defect echo segment based on the waveform rising edge steepness, falling edge tail length, and echo width specifically includes: For each suspected defective echo segment, the median slope of the waveform rising from the amplitude to the peak before the peak is calculated as the rising edge steepness, the time required for the waveform to fall to half the peak after the peak is calculated as the falling edge tail length, and the time span between the two half amplitude points of the waveform is calculated as the echo width. The rising edge steepness, falling edge trailing length, and echo width are input into the preset defect morphology fuzzy membership function to obtain their respective morphology membership scores. The three scores are combined to form the defect echo morphology membership vector of the sub-segment.

9. The ultrasonic non-destructive testing method for water pipe quality according to claim 8, characterized in that, The preset defect morphology fuzzy membership function is an S-shaped membership function, in which the rising edge steepness, falling edge tail length and echo width correspond to their respective membership degree mapping curves, and the threshold parameters of the three curves are pre-calibrated according to the pipe material and wall thickness.

10. An ultrasonic non-destructive testing system for water pipes, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the ultrasonic non-destructive testing method for water pipes as described in any one of claims 1 to 9.

Images