Determination method for ultrasonic flaw clutter separation and sensitivity threshold of flaw detection vehicle

Abstract

The invention relates to a flaw detection vehicle ultrasonic flaw clutter separation and sensitivity threshold determination method, which comprises the following steps: collecting ultrasonic echo signals of each detection channel in the operation process of a flaw detection vehicle, and carrying out reverse analysis based on the consistency relationship between the echo phase and the propagation path; continuously sampling multiple frames of echo signals, and calculating the change rate of the amplitude, the frequency and the main lobe width of a main frequency band along with time; when the change rate exceeds a preset stability threshold value, judging that the signal is an unstable clutter signal by combining interface disturbance characteristics and removing the signal; when stepped descending, slow attenuation or tail reflection signals occur, the damage confirmation logic is executed, and the corresponding area is marked as a potential damage area; signal wave energy is evaluated based on the echo peak amplitude and the duration, and if the wave energy does not reach a system response threshold, a sensitivity threshold is improved; and if the wave energy exceeds the preset upper limit and does not conform to the structural response characteristics, reducing a sensitivity threshold value or implementing signal suppression so as to realize insensitive switching of the sensitivity of the flaw detection system.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of ultrasonic flaw detection technology, and in particular to a method for separating ultrasonic wave damage noise and determining sensitivity threshold of a flaw detection vehicle. BACKGROUND

[0002] The prior art, such as the ultrasonic signal processing scheme disclosed in a steel rail defect intelligent detection method and system based on ultrasonic signals (CN113720910A), has provided a certain degree of intelligent support in steel rail defect detection, but from the perspective of the required damage noise separation accuracy and sensitivity threshold dynamic adjustment of the flaw detection vehicle in actual operation, the scheme still has several technical bottlenecks and application limitations in key links, mainly in the following aspects: first, the patent method emphasizes the removal of noise signals and the extraction of effective ultrasonic signals in the signal extraction stage, but does not model the multi-dimensional characteristics such as the interference of the echo signal phase starting point, the frequency mutation, and the main lobe energy distribution in the actual operation of the flaw detection vehicle, so that when facing complex rail surface environments such as welds, rust, oil stains, and other interface disturbance scenes, it is easy to cause the misjudgment problem of mixing noise signals into structural reflection signals, and cannot effectively support signal consistency analysis and propagation path closed-loop verification functions, resulting in that the noise and real damage cannot be distinguished with high robustness; in addition, the patent only selects echo groups based on the signal density ratio index, and does not implement system modeling on the phase continuity, propagation symmetry, and energy attenuation mode in the signal evolution process, so it lacks a precise elimination mechanism for interference type noise or tail shadow type reflection defects.

[0003] Secondly, the damage classification in the scheme relies on an expert system for post-processing, but it does not explicitly state whether high-order quantitative indicators such as time sequence dynamic characteristics, signal main lobe width evolution trend, and waveform envelope change trajectory are introduced into the classification basis, so that the system lacks the ability to recognize nonlinear characteristics in the damage evolution process and lacks real-time response capability, especially in terms of dynamic switching of sensitivity threshold, the patent does not provide a wave energy control mechanism and feedback closed-loop structure, and the system cannot adaptively adjust the sensitivity when the signal wave energy does not meet the response standard or exceeds the upper limit of the structure, which easily causes the failure of the flaw detection sensitivity dull switching or the increase of the damage omission rate.

[0004] Again, at the waveform analysis level, the patent does not introduce high-order derivative or multi-segment line fitting strategy to identify abnormal structural signal characteristics such as step-down, main lobe shift, adjacent segment angle mutation, etc. present in the echo envelope, nor does it explicitly use any mathematical modeling method to perform symmetry reverse verification analysis on the ultrasonic echo path, so it lacks a propagation rationality evaluation function or a consistency loss function mechanism. In addition, the scheme does not introduce track maintenance history interface or material state parameters to participate in signal stability threshold modeling, resulting in fixed and rigid detection threshold setting, which is difficult to adapt to detection needs under different track types and different working conditions, especially in high-speed running or frequent switching environment, there is a significant response lag problem.

[0005] Further, the patent does not use a multi-channel time difference positioning model to perform reverse mapping of the echo source, nor does it explain whether it has the ability to identify features such as waveform main frequency segment mutation rate, average frequency shift rate, or main lobe symmetry, so there is a risk of error accumulation for boundary ambiguous or overlapping echo signals in actual application. At the same time, the continuity judgment of the signal slope also does not introduce a second-order derivative positive and negative sign statistical density judgment mechanism, so it cannot accurately identify the unstable segment of the waveform, and cannot participate in the subsequent envelope weight adjustment and boundary barycenter dynamic correction, making it difficult to release the risk of rigid jump for edge signal processing. SUMMARY

[0006] The purpose of the present application is to provide a method for separating ultrasonic damage noise and determining the sensitivity threshold of a flaw detection vehicle, thereby solving some of the problems and deficiencies pointed out in the background art.

[0007] The technical scheme adopted by the present application to solve the above technical problems is as follows: a method for separating ultrasonic damage noise and determining the sensitivity threshold of a flaw detection vehicle, comprising: collecting ultrasonic echo signals of each detection channel during the operation of the flaw detection vehicle, recording the phase starting point, energy center position and propagation path time of each frame signal; based on the consistency relationship between the echo phase and the propagation path, perform reverse analysis, when the detection result shows that the phase shift is irreversible or the propagation path does not form a closed loop, the signal is determined as a non-structural reflection signal or a noise signal; A plurality of frames of echo signals are continuously sampled, the amplitude, frequency and main lobe width of the main frequency band are calculated, and the rate of change with time is calculated; when the rate of change exceeds the preset stability threshold, the interface disturbance characteristics are combined to determine that the signal is an unstable noise signal and is rejected; for the signals after screening, the amplitude change of the waveform envelope in the time distance direction is detected, the energy attenuation characteristics of the envelope are determined by calculating the shrinkage trend; when there is a step-down, slow attenuation or tail reflection signal, perform damage confirmation logic, and mark the corresponding area as a potential damage area; The wave energy of a signal is evaluated based on the peak amplitude and duration of an echo, and if the wave energy does not reach a system response threshold, the sensitivity threshold is increased, and if the wave energy exceeds a preset upper limit and does not meet a structural response characteristic, the sensitivity threshold is reduced or signal suppression is implemented, so that the sensitivity of the flaw detection system is switched between dull and sensitive.

[0008] Further, the phase starting point of the ultrasonic echo signal collected by each detection channel during the operation of the flaw detection vehicle is recorded, and the recording includes measuring the phase value and phase continuity of the initial wave peak of the signal, for identifying potential interference type clutter signals. The propagation path time of each echo is calculated using the time difference method between multiple channels to realize reverse positioning of the echo source position. Based on the consistency relationship between the phase information, propagation path, and energy distribution characteristics, a consistency evaluation function is constructed The propagation rationality of the echo signal is reversely analyzed to determine whether the echo is a real structural reflection signal, and the calculation formula is as follows: ; Wherein: is a wave source consistency loss function, which is used to quantify the deviation between the phase behavior, path symmetry, and energy distribution of the current signal; represents the total number of flaw detection channels; is the phase change amount of the th channel, defined as the phase difference between the current frame and the starting frame; is the corresponding theoretical propagation path model function, based on the detection position and the propagation angle ; is the echo energy density of the th channel at position and time ; is the spatial second derivative of the energy density, used to evaluate the symmetry or abnormal bending of the energy distribution; to represents the time window range of the current analysis frame; When the output value of the function exceeds the preset deviation threshold, the system determines that the signal does not meet the propagation consistency condition and is classified as an interference type clutter signal or a non-structural reflection; when the function output value converges to a local minimum value, it indicates that the signal has phase continuity, path reducibility, and energy symmetry, confirming that it is a real structural reflection signal.

[0009] Further, the change rate of the main frequency band includes two dimensions of instantaneous main frequency mutation rate and average frequency offset rate; while calculating the change rate of the main lobe width, the main lobe symmetry index is extracted as an auxiliary feature for judging signal stability; wherein the preset stability threshold is dynamically adjusted according to the material category and surface state of the detected track section within the running period.

[0010] Further, the interface disturbance feature includes signal consistency interference correction on the recognition results of the weld, joint and rail surface corrosion area; wherein the amplitude change calculation of the waveform envelope contains the second derivative judgment on the envelope slope continuity; the detection of the stepped decline is based on the multi-segment line fitting strategy of the envelope curve.

[0011] Further, the interface disturbance feature recognition includes calling the interface through the historical track section maintenance record, and forcibly correcting the signals of the existing known weld and corrosion section for consistency; when identifying the weld interference, the system performs a symmetry difference calculation, and if the left and right adjacent signals appear reverse deformation, the signal is marked as a disturbance area; the identification of the rail surface corrosion area uses the combination of high-frequency component fluctuation intensity exceeding the interference critical threshold as the corrosion interference judgment standard.

[0012] Further, the continuity judgment of the envelope slope uses the second derivative positive and negative sign statistics of the slope change in each window section; when calculating the second derivative, if there are three or more consecutive sampling points with derivative sign change, the signal segment is classified as a waveform unstable segment; the fluctuation interval of the envelope slope second derivative is used to dynamically adjust the envelope boundary; the multi-segment line fitting strategy is to divide the envelope curve into multiple linear segments, and calculate the angle change amount of adjacent segments for identifying the stepped mutation.

[0013] Further, the second derivative positive and negative sign statistics include calculating the sign change density, when the density exceeds the set proportion threshold, the system identifies the segment as a high fluctuation area to trigger the envelope smoothing operation; the slope change of each window segment is updated adaptively with a sliding step, and the step is adjusted according to the signal sampling rate and amplitude gradient.

[0014] Further, the calculation of the second derivative is performed by bidirectional difference accumulation on the first derivative sequence; the recognition result of the waveform unstable segment is cached and used for subsequent envelope correction weight calculation, and the dynamic adjustment of the envelope boundary is completed by calculating the waveform energy distribution center, so that the boundary moving direction is consistent with the waveform energy offset direction.

[0015] Further, when calculating the angle change amount of adjacent segments, the cosine similarity method is used, and the angle direction mutation is used as the stepped feature criterion.

[0016] Further, after the angle direction mutation is identified as a step feature, the feature is analyzed in combination with the waveform envelope energy slope to construct a multi-feature damage confirmation strategy; the judgment result of the angle change direction is cached in a time sequence mutation map, which is used for monitoring the track continuity and evaluating the damage behavior evolution trend.

[0017] The method for ultrasonic damage clutter separation and sensitivity threshold determination of the inspection vehicle of the present application can realize adaptive identification and accurate separation of ultrasonic signals in a complex track environment by introducing phase consistency analysis, energy symmetry evaluation and dynamic judgment mechanism of the second derivative of envelope slope. Through comprehensive analysis of the phase starting point, energy center and propagation path time of the echo signal, this method can complete the elimination of non-structural reflection signals in the time domain, space domain and frequency domain, significantly improve the clutter recognition accuracy and the structure restoration ability of damage echo. The use of multi-channel time difference reverse positioning and path symmetry verification enables the inspection system to have self-learning and spatial tracing capabilities, effectively reduces the manual misjudgment rate, and realizes automatic consistency correction of typical interference areas such as welds and corrosion sections.

[0018] In addition, the present application realizes dynamic self-adjustment of envelope boundary and multi-index fusion judgment of damage mutation characteristics by using the second derivative positive and negative sign statistics of envelope slope, waveform angle cosine similarity analysis and energy distribution gravity calculation. This method not only can adaptively adjust the sensitivity threshold, but also can improve the detection gain when the wave energy is insufficient, and reduce the response threshold when there is a high risk of misjudgment. It can also track the evolution trend of damage behavior through the time sequence mutation map, so that the inspection vehicle can realize higher signal-to-noise ratio, more stable real-time damage detection and sensitivity control in high-speed running environment. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 The present application is an ultrasonic damage clutter separation and sensitivity determination main process.

[0020] Figure 2 The present application is an ultrasonic signal consistency discrimination and clutter interference elimination function relationship diagram.

[0021] Figure 3 The present application is an envelope slope and angle feature dynamic identification flowchart.

[0022] Figure 4 The present application is an ultrasonic signal abnormality processing and damage identification flowchart of the high-speed rail inspection vehicle of embodiment 1.

[0023] Figure 5 The present application is an ultrasonic signal envelope segmentation and structural response identification flowchart of the inspection vehicle of embodiment 2. DETAILED DESCRIPTION

[0024] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0025] Combined with appendix Figure 1 This invention relates to a method for separating ultrasonic damage clutter and determining sensitivity thresholds in a flaw detection vehicle. This method involves synchronously acquiring ultrasonic echo signals generated during operation via multiple detection channels deployed on the vehicle. Feature extraction is performed on each frame of the acquired echo signal, including recording its phase start point, energy center position, and signal propagation path time. The phase start point is obtained by analyzing the initial rising edge characteristics of the echo signal; the energy center position is calculated based on the power distribution of the main lobe region of the signal waveform; and the propagation path time is calculated based on the time interval between transmission and reception combined with the positional relationship between the channels. After acquiring the above data... Based on this, the system constructs a consistency judgment model that uses phase change behavior and path propagation trajectory as joint criteria. By comparing the fitting degree between the actual propagation path of the current frame signal and the theoretical structural propagation path, and monitoring the evolution trend of the signal phase in the spatial distribution, when it is identified that the phase change of the echo exhibits nonlinear drift and cannot form a closed-loop or invertible path trajectory in the spatial structure, it is determined that the signal does not conform to the structural echo law. The signal is then marked as an unstructured reflection signal or clutter signal and removed from the subsequent damage identification process, thereby improving the system's echo source identification capability in complex orbital environments.

[0026] The system continuously samples and processes multiple frames of echo signals acquired during the continuous operation of the flaw detection vehicle. It sequentially extracts the amplitude characteristics, center frequency, and main lobe width of each frame within its main frequency band, and constructs a time-axis trajectory based on the continuous frame sequence. Then, it calculates the rate of change of these features over time to obtain a stability index of the frequency domain characteristics. When the rate of change of any feature exceeds the system's set stability threshold, it indicates that the signal has a severe temporal disturbance or transient change. Based on this, the system analyzes the interface disturbance characteristics in conjunction with track structure surface information, including the matching of known interference areas such as weld transitions, surface corrosion, or joint wear. If the judgment result points to a human or structural interference source, the signal is identified as an unstable clutter signal and discarded. For the remaining frames after discarding... The system further extracts the envelope curve of the selected signal in the time-distance direction and performs continuity detection. By monitoring the evolution trend of the envelope amplitude with propagation time, it calculates the contraction rate and energy attenuation law. If the signal envelope shows a significant step-like decrease, that is, a segmented sudden drop occurs in the continuous propagation interval, or a slow attenuation process, that is, the signal amplitude continuously weakens within a certain time distance range rather than a sudden decrease, or a secondary wave group echo with structural characteristics appears in the rear of the main reflected wave, i.e., tail reflection characteristics, the system triggers the damage confirmation logic, identifies the spatial segment where the corresponding signal is located as a suspected damage area, and includes it in the subsequent defect location and level assessment process, thereby achieving accurate restoration of potential damage signals under complex waveforms and effective elimination of clutter interference.

[0027] A joint evaluation method based on echo peak amplitude and duration is used to quantitatively calculate signal wave energy. The system extracts the maximum amplitude and calculates the effective duration of each frame of ultrasonic echo signal, using these two parameters as the main parameters to form a wave energy evaluation index to determine whether the signal has the ability to trigger a system structural response. If the evaluation result indicates that the current signal wave energy is lower than the system's preset minimum response threshold, it indicates that the signal may be close to noise level due to weak penetration or severe attenuation. The system accordingly increases the sensitivity threshold to enhance the receiving gain and improve the ability to capture weak defect signals, thereby preventing the omission of effective damage information. Conversely, when the echo signal wave energy exceeds the system's defined energy limit, but its signal morphology, spectral characteristics, or propagation path behavior is inconsistent with typical structural responses, exhibiting characteristics such as asymmetric waveforms, frequency band drift, or abnormal paths, the system classifies this type of signal as non-structural interference or strong clutter. Based on the judgment level, the sensitivity threshold is reduced to avoid excessive system response. Simultaneously, a signal suppression mechanism can be activated to limit its impact in subsequent processing, thereby achieving dynamic switching between sensitivity and stability in the flaw detection system.

[0028] Combined with appendix Figure 2To accurately identify and classify ultrasonic signal sources from different detection channels, the system continuously records the phase start point information of the ultrasonic echo signals acquired by each detection channel during the operation of the flaw detection vehicle. The recording process includes measuring the phase value of the initial peak of the signal and analyzing the phase continuity. Potential interference clutter signals are identified by judging the phase change trend. When phase continuity is interrupted or nonlinear drift occurs, the system marks that segment of the signal as a suspected interference source and proceeds to the subsequent consistency verification stage.

[0029] Subsequently, the system employs a multi-channel time difference method to calculate the propagation path time of the echoes from each channel. By statistically analyzing the arrival time differences of the received signals from different channels, the location of the echo source can be reversed under the condition of known probe spacing and sound velocity, thereby establishing a mapping relationship of the echo propagation path in the spatial domain. This calculation process enables the flaw detection system to have the ability to direct the reflection source, providing a data basis for distinguishing between structural reflections and random clutter.

[0030] After obtaining phase information, propagation path, and energy distribution characteristics, the system constructs a consistency evaluation function to assess the rationality of signal propagation. This function comprehensively analyzes the correlation between signal phase changes, path symmetry, and energy spatial distribution. Its calculation expression is as follows: ; in: This represents the source consistency loss function, used to quantify the deviation between the current signal's phase behavior, path symmetry, and energy distribution. The total number of channels participating in the testing; For the first The phase change of a channel is defined as the phase difference between the current frame and the starting frame; This is a theoretical propagation path model function, based on the detection location. From the perspective of communication The geometric relationships are determined; For the first The passage is in the location and time The echo energy density distribution below; This is the spatial second derivative of the energy density, used to characterize the symmetry of the signal in its spatial distribution and the degree of local anomalous curvature; to Indicates the time window range of the current analysis frame.

[0031] During operation, the system processes the function. The system calculates and monitors the output value in real time. When the output value of the function exceeds the preset deviation threshold, it indicates a significant inconsistency between the phase evolution and the theoretical path, and that the energy distribution exhibits asymmetric characteristics. The system determines that such signals do not meet the propagation consistency condition, classifies them as interference clutter or unstructured reflection signals, and discards them. When the function result gradually converges to a local minimum, it indicates that the signal can be closed-loop reconstructed on the spatial path, with continuous phase changes and symmetrical energy distribution. Based on this, the system confirms that the signal originates from a real structural reflection and uses it as a valid damage characteristic signal in the subsequent quantitative assessment and defect judgment process, thereby achieving quantitative judgment of the authenticity of the ultrasonic signal and verification of propagation consistency.

[0032] function The derivation process includes: Set the total number of channels to For any one For channels, the phase evolution features are extracted first. That is, the channel in the current frame The phase offset relative to the signal's starting frame directly reflects the likelihood of structural disturbances or interference along the signal path; to account for spatial propagation characteristics, the orbital coordinate position is... From the perspective of communication As parameters, construct the theoretical propagation path function. This function can approximately predict the actual reflection path based on parameters such as structural geometry and material sound velocity. Since clutter interference often causes spatial asymmetry in energy reflection, an energy density function is introduced to address this issue. Indicates channel In spatial location ,time The signal power distribution under the given conditions, and take its pair The second derivative To characterize the spatial curvature of energy changes, the greater the symmetry change, the stronger its spatial non-reducibility, ultimately within a continuous time window. Integrate the derivative to obtain the cumulative energy deviation of the channel within a given spatial segment.

[0033] To achieve consistent quantification, residual terms are constructed. and to The average value of all channels is calculated to form the overall consistency loss function. When the function value is at a local minimum, it indicates that the current frame signal has reversible propagation path, phase stability, and energy symmetry, which means it belongs to a real structural reflection signal. If the function output value is consistently high or fluctuates drastically, it is considered a manifestation of phase disorder, path breakage, or structural asymmetry, corresponding to interference clutter or unstructured reflection sources, which can be eliminated accordingly.

[0034] The system analyzes the dominant frequency characteristics of continuously acquired ultrasonic echo signals during flaw detection and constructs a signal change trend over time. To more accurately assess frequency change behavior, the system calculates the instantaneous dominant frequency abrupt change rate and the average frequency offset rate of the dominant frequency band. The instantaneous dominant frequency abrupt change rate is used to identify rapid frequency drift caused by sudden interference, while the average frequency offset rate is used to measure the long-term stability trend of the signal's dominant frequency. Together, they constitute a two-dimensional index of the dominant frequency band change rate. Based on the above frequency behavior, the system simultaneously extracts the signal's main lobe width and calculates its change rate to determine the spatial broadening or compression effect of the waveform structure. Further, it extracts the main lobe symmetry index. This symmetry index, by analyzing the balance of energy distribution on both sides of the main lobe waveform, determines whether the signal exhibits deformation or disturbance characteristics. This parameter serves as an auxiliary stability feature input to the system's stability determination module, and together with the frequency change index, forms the basis for multi-feature clutter identification. Considering that different track structures, material types and surface conditions can significantly affect the propagation characteristics of ultrasonic signals, the system introduces a preset stability threshold dynamic adjustment mechanism. The threshold is corrected and updated based on external factors such as the material type, rail type, surface roughness level and temperature and humidity environment of the detected track segment during operation, so as to realize adaptive adjustment of clutter identification criteria in actual environment.

[0035] The system introduces an interface disturbance feature-assisted correction mechanism. When the flaw detection vehicle passes through welds, track joints, or areas with surface corrosion, the ultrasonic echo signal often generates unstructured clutter due to local geometric distortion, discontinuities in the reflecting surface, or material defects. To address this phenomenon, the system identifies and marks the weld location, joint distribution, and corroded sections by calling up existing historical track maintenance data and real-time acquired track surface geometric information, and marks the echo signals collected in these areas as "disturbance zone signals." In subsequent processing, the system implements consistency interference correction for disturbance zone signals. This involves identifying the typical signal morphology of such areas and the amplitude, frequency, and phase variation trends between the signals and those in adjacent non-disturbance sections, constructing correction parameters to achieve signal restoration or weight reduction, thereby reducing its interference with the damage identification process.

[0036] For the perturbation-corrected echo signal, the system further extracts its envelope features and calculates the change process of the envelope amplitude in the time-distance direction, paying particular attention to the continuity of the slope change of the envelope curve. During this process, the system calculates the second derivative of the envelope slope and judges the signal stability by detecting the continuity of the sign of the slope change. If multiple derivative sign reversals occur in consecutive sampling points, it is considered that the signal has a non-stationary perturbation. The system then marks this segment as a waveform anomaly region and performs subsequent feature recognition processing. In identifying signals that may exhibit energy attenuation, to extract typical step-like attenuation characteristics caused by damage, the system uses a multi-segment line fitting strategy to segment and model the envelope curve. By finding multiple approximately linear fitting segments and calculating the angle change and slope abrupt change positions between adjacent segments, it judges whether there is a sudden decrease in amplitude. If the fitting result contains multiple line segment connection points with significantly different slopes, and the overall trend of the envelope curve shows a segmented decrease, the system confirms that the signal exhibits typical step-like descent characteristics, thereby triggering the damage confirmation process and marking the spatial location of the signal as a high-risk segment.

[0037] The system introduces an identification and intervention mechanism based on interface disturbance characteristics to proactively identify and suppress structural noise sources. The system pre-establishes a data interface with the track maintenance system to access historical track section structural maintenance records, including weld construction time, rail surface grinding batches, and areas requiring localized corrosion reinforcement. When the system detects that the current vehicle's track section matches a registered disturbance section, it activates a forced signal consistency correction strategy, marking the echo data of that area as a disturbed section. In subsequent signal analysis, waveform correction or weight reduction is implemented to prevent such non-destructive interference from misleading the overall judgment results.

[0038] When handling weld interference, the system calculates based on the spatial symmetry characteristics of the echo signal structure, extracts the echo signal profiles of adjacent channels on the left and right sides of the weld center, and calculates the symmetry difference in waveform shape, energy distribution, and phase change. When the system identifies that the signal in this area exhibits a left-right reversal deformation trend, i.e., the signal is strengthened on one side while it is significantly weakened or reversed on the other side, the system determines that the weld has caused a disturbance to the waveform structure and marks this signal segment as a disturbance area for further processing. For the identification of rail surface corrosion areas, the system extracts the high-frequency components of the signal in the frequency domain and analyzes their fluctuation intensity characteristics. If the high-frequency amplitude exhibits abnormal amplitude jitter or continuous jumps within a short period of time, and its peak value exceeds the system's preset interference threshold, the system determines that the signal is affected by echo distortion caused by corrosion and further marks it as a corrosion interference signal segment.

[0039] The system divides the ultrasonic echo signal into multiple equally wide analysis windows and extracts the trend of the envelope curve slope within each window segment. Then, it performs second derivative operations on these slope sequences and statistically analyzes the distribution of their positive and negative signs to identify the continuity and stability of the signal structure. During this process, the system processes the envelope slope sequence within each analysis window segment. If the signs of the second derivatives of three or more consecutive sampling points alternate (i.e., alternating positive and negative signs), the system determines that the slope fluctuation of that segment is significant, indicating a clear structural discontinuity or disturbance in the signal, and identifies it as an unstable waveform segment. Based on this, the system further extracts the fluctuation range of the second derivative of the envelope slope and dynamically adjusts the envelope boundary accordingly. The expansion or contraction direction of the envelope boundary depends on the local energy gradient shift trend, ensuring that the concentrated signal energy area is completely covered while suppressing abnormal fluctuations from affecting the accuracy of the envelope definition.

[0040] To further identify abrupt changes in the waveform corresponding to damage edges or unstable echo characteristics, the system introduces a multi-segment line fitting strategy to model the envelope curve. The envelope curve is divided into multiple continuous linear fitting segments. Each segment is fitted using the least squares method to obtain an approximate straight line representation. The system sequentially calculates the change in the angle between adjacent fitting segments. When the abrupt change in angle exceeds a preset threshold angle and the directionality changes significantly, the point is identified as an abrupt change point with step-like characteristics, indicating the presence of structural defects or abnormal energy reflection behavior at that location.

[0041] Combined with appendix Figure 3 The system extracts the first-order slope change sequence of the signal envelope within each analysis window and further calculates its second-order derivative to reflect the acceleration characteristics of the envelope change. To accurately capture high-frequency fluctuations in signal disturbance regions, the system not only extracts the direction of slope change but also performs statistical analysis on the positive and negative signs of the second-order derivative. During this statistical process, the system counts the number of switching between positive and negative signs within each window and calculates the sign change density based on the total number of sampling points. When this density value exceeds a preset proportional threshold, such as being higher than 30% of the total number of sampling points, the system determines that the waveform slope in that region fluctuates drastically and identifies that window segment as a high-fluctuation area. After identification, the system triggers an envelope smoothing operation, optimizing the envelope curve segment through low-pass filtering or weighted moving average to reduce the impact of clutter interference on damage feature identification and improve the stability of signal representation.

[0042] Meanwhile, to avoid feature extraction distortion caused by mismatched window lengths, the system adaptively updates the sliding step size for each analysis window segment. The step size determination strategy is based on the sampling rate of the current signal and the gradient magnitude of the envelope amplitude change. When the signal sampling rate is high and the amplitude change is relatively gentle, the system appropriately increases the sliding step size to reduce redundant computational overhead; conversely, when the amplitude gradient changes drastically, the system decreases the step size to obtain more refined waveform resolution. Through this adaptive adjustment mechanism of the sliding step size, the system can balance processing efficiency and recognition accuracy under different operating conditions, ensuring that slope change features and disturbance behaviors are effectively captured in multi-scale windows. This enables accurate calibration and dynamic envelope control of the high-fluctuation segment of the flaw detection signal, providing a stable and reliable basic input for subsequent damage identification.

[0043] The system performs differential operations on the slope sequence both forward and backward in each analysis window, and then superimposes and fuses the differential results to obtain a smooth and physically meaningful second-derivative change curve. Based on this curve, the system determines whether there are significant fluctuations in the waveform and identifies the segment as an unstable segment if certain change conditions are met. The identification results of the unstable segments are cached and stored in the signal processing queue. Subsequently, their identification results are introduced as one of the weighting factors in the envelope correction calculation, thereby assigning a lower weight to the envelope calculation of the unstable segment signal to reduce the impact of interference.

[0044] To achieve intelligent adjustment of the waveform envelope boundary, the system further introduces the waveform energy distribution centroid as a judgment reference. During execution, the system calculates the energy density distribution corresponding to the envelope within the current window and performs mass center analysis on the main energy region of the waveform based on this distribution to determine its centroid position in the spatial dimension. When a continuous offset trend is detected within the boundary region, the system adjusts the value of the envelope boundary line according to the centroid displacement direction, causing the boundary to move compensatorily in the direction of the energy offset trend. This dynamic adjustment method significantly improves the flaw detection system's ability to reconstruct the true boundary of structural reflections, avoiding misjudgments of damage caused by envelope truncation or energy omission. It also effectively suppresses envelope deformation caused by interference clutter, thereby achieving more stable and reliable flaw detection signal envelope extraction.

[0045] The system performs linear segmentation on the waveform envelope curve. For each pair of adjacent linear segments, the system first extracts their slope vectors and constructs the angle parameter. The system then quantifies the trend of change by calculating the cosine of the angle between the two vectors. Specifically, if the direction vector of the previous linear segment is... The direction vector of the next linear segment is The change in the included angle is then evaluated using the cosine similarity function, expressed as a vector. and The cosine value obtained by dividing the inner product by its modulus reflects the degree of similarity between two trends. When this value changes abruptly between consecutive segments, especially from high positive correlation to low or even negative correlation, the system determines that the direction of the angle has changed significantly.

[0046] Based on the aforementioned changes in the included angle direction, the system introduces a direction abrupt change criterion as a trigger signal for the step feature. When a jump in the cosine similarity value is detected, and the corresponding angle change is greater than the preset step abrupt change threshold, the system marks the node as a potential step feature location.

[0047] After extracting the envelope of the ultrasonic echo signal and performing linear piecewise fitting, the system calculates the change in the directional angle between adjacent line segments and identifies the locations of abrupt changes, marking such abrupt events as potential step response signals. The abrupt change in directional angle not only serves as an independent structural feature indicator but is also jointly matched with the energy slope change trend of the waveform envelope corresponding to that location. The system enhances the reliability of the judgment by determining whether the two abrupt changes occur in the same direction and simultaneously exhibit significant shifts.

[0048] Furthermore, to achieve time-series tracking of the evolution of damage behavior, this invention caches the judgment results of the aforementioned angle change direction at the frame level and constructs a time-series mutation map. This map uses time as the horizontal axis and the change in angle between envelope segments as the vertical axis, marking the time points and magnitudes of all triggering directional mutation events. By analyzing the clustering, frequency, and directional shift trends of mutation events in this map, the system can identify interruptions in trajectory continuity and assist in assessing the development direction and rate of change of potential damage. For example, when a region experiences consecutive frame-wide angle mutations accompanied by a decrease in energy slope, the system will trigger enhanced damage confirmation logic and simultaneously adjust the sensitivity configuration of subsequent frames to achieve dynamic response and closed-loop monitoring of damage evolution trends.

[0049] Example 1: Combined with appendix Figure 4 In this embodiment, a high-speed railway line arranges a track flaw detection vehicle for regular inspections during operation. This vehicle is equipped with eight ultrasonic detection channels, each with a sampling frequency of 5MHz and a pulse repetition frequency of 1kHz. During one inspection mission, the vehicle speed was maintained at 60km / h, and multiple abnormal echo points were detected on a certain track section. The analysis will now focus on one inspection time window of the third channel, with a time range from... arrive The corresponding spatial location range is From the perspective of dissemination .

[0050] The system records the initial peak phase value of the echo signal of the current frame in this channel. The phase values ​​for the four consecutive frames are as follows: ; The phase difference between adjacent frames suddenly increases, where Exceeding the system's set upper limit for continuity deviation It was determined that there was a tendency for interference-type clutter.

[0051] The response times of channels 3 and 5 to the same echo are as follows: , ; The lateral offset of the echo source on the sensor array was calculated using the time difference method: ; Among them, the speed of sound Based on the array geometry, the echo source is located 0.236 meters to the left of the front of the 4th channel.

[0052] To verify whether this signal represents a true structural reflection, a consistency evaluation function is constructed: ; Substitute the parameter settings: Number of channels ; No. Channel phase difference estimation Within the range of 0.05 to 0.50π; Theoretical propagation path function ,by ,Pick ,have to ; The echo energy density model is as follows: The second derivative is; ; make , , Integration range The time is 0.002 seconds. The integral value is approximated as 0.32, and finally substituted into the formula for calculation. ; The system's set consistency deviation threshold is The current calculation result significantly exceeds this value, indicating that the echo does not conform to the laws of real structural reflection in terms of phase change, propagation path and energy symmetry. The system ultimately classifies it as an interference clutter signal and removes it from the damage confirmation process.

[0053] After completing the aforementioned consistency assessment and eliminating a batch of initially confirmed interferometric clutter signals, the system continues to process the remaining echo frame signals to further identify potential structural damage areas. (Based on location segment) Taking a meter as an example, the flaw detection vehicle detected multiple short-term enhanced signals. The system analyzed its main frequency band characteristics, envelope change trend and local structure to determine its stability and possible damage attributes.

[0054] Spectral analysis was performed on this signal segment, and the system extracted the main frequencies of five consecutive frames of data as follows: , , , , ; Among them, the maximum inter-frame frequency change is The instantaneous mutation rate is: ; The average frequency offset is calculated as follows: ; The system's preset stability threshold is determined by the rail section's material and surface condition. This section is located in the transition zone of the aluminothermic welded rail, with a medium level of surface corrosion. Therefore, the system sets the mutation rate threshold to be [value missing]. The average offset threshold is Since both test results exceeded the limits, the system marked this segment as a frequency unstable segment.

[0055] The main lobe width and symmetry of the above frame signals were analyzed, and the following parameters were extracted: Changes in main lobe width: From Shrink to ; Main lobe symmetry index: ; in and These represent the integrated energies at half-widths of the left and right main lobes, respectively. Due to symmetry deviation exceeding the system's allowable range... The system further classifies it as an unstable signal segment.

[0056] The system, after accessing the historical track database, identified this section as a concentrated area of ​​welds and detected a continuous 0.4-meter layer of corrosion on the rail surface. To mitigate the waveform shift caused by the welds and corrosion, the system implemented a consistency interference correction algorithm, which includes: Weld interference correction: Calculation of symmetry difference If the threshold of 0.22 is exceeded, the interference intervention process will begin. Corrosion disturbance correction: The energy proportion of high-frequency components in the frequency domain of the detected signal exceeds 45%, and the corrosion effect is confirmed by combining the surface state spectrum; After interference correction, the system renormalizes the spectral energy and adjusts the signal envelope to improve the ability to identify true reflections.

[0057] A polyline fitting analysis was performed on the corrected waveform envelope. The envelope can be divided into three segments in the time-distance direction, and the included angle between adjacent segments was calculated. and The consistent direction of change indicates a clear step-like descent characteristic.

[0058] Simultaneously, the system calculates the second derivative positive and negative sign sequence for each envelope slope segment as follows: First paragraph: +++; Second paragraph: --+; Third paragraph: +--; If the sign changes twice in the second segment, the system determines that the slope is discontinuous, constructs a step-change index based on the angle change trend, and triggers the damage confirmation process.

[0059] The flaw detection vehicle continued its inspection along the track from east to west until it reached section TQ42. The track maintenance system's historical database showed a weld joint record 31.7 meters from the reference point, and a lightly corroded area between 32.1 and 32.5 meters. Upon receiving the echo signal from this section, the system invoked the maintenance record interface for early warning consistency correction, and simultaneously began extracting interference features for comparison and verification.

[0060] At a position of 31.7 meters, the system observed that the main waveform of the signal exhibited a deformation of tilting forward on the left and shifting backward on the right. The system then compared the symmetrical waveforms on the left and right with the symmetrical waveforms on the left and right to calculate the symmetry difference. Set the maximum slope on the left as The maximum slope on the right is The system uses a symmetric difference formula: ; Because the system's interference threshold for the weld zone is set to... The current value exceeds the limit, so it is marked as "weld disturbance zone". The system enters the signal correction process, incorporating this echo segment into the consistency forced normalization algorithm, so that its slope and energy distribution transition to a stable mode, thereby avoiding misidentification.

[0061] When the flaw detection vehicle reached a distance of 32.2 meters, the system detected that the high-frequency component of the echo signal, extracted from the spectrum, accounted for more than 51% of the energy. The specific frequency domain integral is as follows: Integral energy in the high-frequency region (>4.5MHz): ; Total energy integral: ; Calculate the proportion of high-frequency energy: ; This value has exceeded the corrosion interference threshold set by the system. The system identifies this area as a "corrosion high-frequency interference zone." Using the corrosion segment identification tag, the signal in this segment is incorporated into a weighted compensation sequence, and its phase slip, amplitude offset, and boundary constraints are dynamically adjusted.

[0062] The echo envelope after interference compensation is processed using a one-dimensional sliding window, with each window containing 50 sampling points. The slope derivative sequence is calculated using the central difference. Let the first derivative sequence of the slope within a certain window be: ; Calculate the second derivative sign transformation sequence: : Positive → Positive (unchanged); : Positive → Negative (change sign 1); : Negative → Negative (unchanged); : Negative → Positive (change sign 2); : Positive → Negative (change sign 3); If the sign changes three times, the system determines that the window is a "waveform instability segment". This marker will serve as the core basis for subsequent envelope boundary adjustments.

[0063] The system extracts three adjacent slope segments from the waveform envelope of the 32.4-meter segment: Section 1: Fitting Slope ; Section 2: Fitting Slope ; Section 3: Fitting Slope ; Calculate the cosine similarity of the angle between segment one and segment two: ; The angle direction changes from a steep drop to a gradual drop, then drops sharply again. The cosines of segments two to three are: ; Although the cosine values ​​are all 1, the actual variation between the fitted lines is large, indicating that although the directions are the same, the slope values ​​change abruptly. The system combines the change in the angle between the fitted segments and the discontinuity of the slope to identify step-by-step abrupt changes and marks them as "suspected damage response segments".

[0064] Example 2: Combined with appendix Figure 5Based on Example 1, in this example, during the continued depth inspection of the TQ42 track section by the flaw detection vehicle, the system uses the aforementioned signal stability identification strategy to perform more refined structural identification and envelope boundary correction on the dynamic changes of waveform slope: When the detection range is between 32.6 meters and 32.9 meters, the system extracts the envelope curve of a certain channel signal waveform and divides it into several window segments. Each window is set to have 60 sampling points, with an initial sampling rate of 2500 points per second. The system continuously analyzes the slope change of the envelope using a sliding window. Within a certain window, the first derivative sequence is: ; The system calculates the second derivative of the sequence, that is, the difference between adjacent first derivatives: ; Further extraction of its positive and negative sign sequence is as follows: ; The sign changes 6 times, the window length is 60 points, and the sign density is: ; If the system sets the sign density threshold to 0.6, then the current window is clearly a high-fluctuation segment, and the system triggers envelope smoothing operation to include this region in the waveform instability processing strategy.

[0065] The system uses the maximum magnitude gradient within the current window as... And combined with signal sampling rate Calculate the sliding step size ; ; At this point, the window step size is updated from the original setting of 5 points to 8 points, achieving a smoother and more locally adaptive scan and improving recognition accuracy.

[0066] The system has calculated first derivative sequences. Simultaneously perform forward and backward differencing to form a bidirectional difference sequence. and The cumulative results are combined to form a unified stability weight index: ; Taking the fourth point in the above sequence as an example: ; ; ; This value exceeds the system's set fluctuation critical weight of 0.5. The system marks this point and its neighborhood as unstable node cache units and assigns them high correction priority in subsequent envelope boundary adjustments.

[0067] For highly volatile windows, the system does not directly truncate them, but instead translates the envelope boundary based on the centroid of the waveform energy distribution. The waveform energy density within a certain window is: ; The sampling point index is The system calculates the position of the center of gravity: ; The original boundary position was set at the midpoint 4. The system shifted it to the right to the new boundary position 6 to follow the energy offset, thereby ensuring that the envelope processing matches the actual signal energy focusing area.

[0068] After completing the main inspection tasks in the middle section of the TQ42 track, the flaw detection vehicle continued along the predetermined path to conduct detailed analysis of sections suspected of structural fatigue. Following the earlier waveform stability assessment based on slope fluctuations, envelope centroid shift, and symmetry indexing, a cosine similarity calculation method based on abrupt angle changes was further introduced to identify stepped signal features. Combined with the changing trend of envelope energy, multi-feature fusion was used to confirm damage.

[0069] Between 41.2 meters and 41.6 meters above the track position, the waveform envelope curve of a certain detection channel has multiple inflection points. The system uses a multi-segment line fitting algorithm to divide this signal segment into five linear segments, each represented by a vector, denoted as: ; Each vector represents a linear segment. Direction change and amplitude direction change.

[0070] The system calculates the angle between adjacent line segments using cosine similarity. The formula for cosine similarity is: ; Take paragraphs 2 and 3 as examples: ; The dot product is: ; The module length is: , ; ; The system identifies that the cosine value corresponding to the included angle is lower than the set similarity threshold of 0.7, and the vector direction changes from positive to negative, that is, a significant change in direction occurs. It is determined to be a step-like change inflection point, and the corresponding signal slope trend reverses sharply, which has potential structural response characteristics.

[0071] The system further combines the trend of waveform envelope energy slope changes before and after the inflection point for joint judgment. Let the waveform envelope energy slope before the inflection point be: ; The slope of the waveform envelope energy after the inflection point is: ; The system detected that the energy slope dropped sharply from positive 1.4 to negative 3.8, and the trend was consistent with the abrupt change in the angle direction, further confirming that this segment of the signal has the typical characteristics of a sudden structural response.

[0072] The system marks the above identification results as feature nodes and caches them in the time series mutation map. The map records the following fields: Timestamp: T20xx-xx-xx-xx:xx:xx; Location: 41.38 meters; Step abrupt change cosine similarity: 0.614; Energy slope variation: +1.4 to -3.8; Signal morphological stability label: Unstable segment; Triggering cause: abrupt change in the included angle direction + reverse slope of wave energy; This map was referenced again in subsequent channel verifications. Similar reverse slope and angle abrupt behavior were also detected in adjacent channels at the corresponding positions. The system performed damage confirmation logic on it and finally marked this segment as the characteristic area of ​​the initial stage of fatigue crack in the analysis report.

[0073] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A method for separating ultrasonic damage clutter and determining sensitivity threshold using a flaw detection vehicle, characterized in that, include: The ultrasonic echo signals from each detection channel of the flaw detection vehicle are collected during operation, and the phase start point, energy center position and propagation path time of each frame of signal are recorded. Inverse analysis is performed based on the consistency relationship between the echo phase and the propagation path. When the detection results show that the phase shift is irreversible or the propagation path does not form a closed loop, the signal is determined to be an unstructured reflection signal or a clutter signal. The system continuously samples multiple frames of echo signals and calculates the rate of change of amplitude, frequency, and main lobe width of the main frequency band over time. When the rate of change exceeds a preset stability threshold, the signal is determined to be an unstable clutter signal and is removed based on the interface disturbance characteristics. For the filtered signal, the amplitude change of its waveform envelope in the time-distance direction is detected, and its energy attenuation characteristics are determined by calculating the contraction trend of the envelope. When a step-like descent, slow decay, or tail reflection signal occurs, the damage confirmation logic is executed, and the corresponding area is marked as a potential damage area. The signal energy is evaluated based on the peak amplitude and duration of the echo. If the energy does not reach the system response threshold, the sensitivity threshold is increased. If the energy exceeds the preset upper limit and does not meet the structural response characteristics, the sensitivity threshold is decreased or signal suppression is implemented to enable the switching between sensitivity and insensitivity of the flaw detection system.

2. The method for separating ultrasonic damage clutter and determining sensitivity threshold using a flaw detection vehicle according to claim 1, characterized in that, The recording of the echo phase start point includes measuring the phase value and phase continuity of the initial peak of the signal to identify potential interference clutter; the calculation of the propagation path time uses the time difference method between multiple channels to reverse the location of the echo source. The reverse analysis of the consistency relationship includes constructing a symmetry verification corresponding to the echo path to confirm whether the propagation direction of the echo can be closed-loop restored.

3. The method for separating ultrasonic damage clutter and determining sensitivity threshold using a flaw detection vehicle according to claim 1, characterized in that, The rate of change of the main frequency band includes two dimensions: instantaneous main frequency change rate and average frequency offset rate; while calculating the rate of change of the main lobe width, the main lobe symmetry index is extracted as an auxiliary feature for judging signal stability; wherein the preset stability threshold is dynamically adjusted according to the material type and surface condition of the detection track segment during the operation period.

4. The method for separating ultrasonic damage clutter and determining sensitivity threshold using a flaw detection vehicle according to claim 1, characterized in that, The combination of interface disturbance features includes signal consistency interference correction for the identification results of welds, joints and rail surface corrosion areas; the calculation of waveform envelope amplitude change includes second derivative judgment of envelope slope continuity; the step-down detection is based on a multi-segment line fitting strategy of envelope curve.

5. The method for separating ultrasonic damage clutter and determining sensitivity threshold using a flaw detection vehicle according to claim 4, characterized in that, The identification of interface disturbance features includes calling the interface through historical track section maintenance records to perform forced consistency correction on signals with known welds and corrosion sections; when identifying weld interference, the system performs symmetry difference calculation, and if the left and right adjacent signals show reverse deformation, the signal is marked as a disturbance area; the identification of the rail surface corrosion area adopts the criterion of judging corrosion interference by combining the high frequency component fluctuation intensity exceeding the interference critical threshold.

6. The method for separating ultrasonic damage clutter and determining sensitivity threshold using a flaw detection vehicle according to claim 4, characterized in that, The continuity determination of the envelope slope is achieved by statistically analyzing the sign of the second derivative of the slope change within each window segment. When calculating the second derivative, if there are three or more consecutive sampling points where the derivative sign changes, the signal segment is classified as an unstable waveform segment. The fluctuation range of the second derivative of the envelope slope is used to dynamically adjust the envelope boundary. The multi-segment line fitting strategy divides the envelope curve into multiple linear segments and calculates the change in the angle between adjacent segments to identify step abrupt changes.

7. The method for separating ultrasonic damage clutter and determining sensitivity threshold using a flaw detection vehicle according to claim 6, characterized in that, The second-order derivative positive and negative sign statistics include calculating the sign change density. When the density exceeds a set proportional threshold, the system marks the segment as a high fluctuation region to trigger envelope smoothing operation. The slope change of each window segment is adaptively updated using a sliding step size, which is adjusted according to the signal sampling rate and amplitude gradient.

8. The method for separating ultrasonic damage clutter and determining sensitivity threshold using a flaw detection vehicle according to claim 6, characterized in that, The second derivative is calculated by performing bidirectional differential accumulation on the first derivative sequence; the identification result of the unstable segment of the waveform is cached and participates in the subsequent envelope correction weight calculation; the dynamic adjustment of the envelope boundary is completed by calculating the centroid of the waveform energy distribution, so that the boundary movement direction is consistent with the waveform energy offset direction.

9. The method for separating ultrasonic damage clutter and determining sensitivity threshold using a flaw detection vehicle according to claim 6, characterized in that, The calculation of the change in the angle between adjacent segments adopts the cosine similarity method, and the abrupt change in the direction of the angle is used as the criterion for the step feature.

10. The method for separating ultrasonic damage clutter and determining sensitivity threshold using a flaw detection vehicle according to claim 9, characterized in that, After the abrupt change in the included angle direction is identified as a step feature, the feature is analyzed together with the waveform envelope energy slope to construct a multi-feature damage confirmation strategy; the judgment result of the included angle change direction is cached in the time series mutation map to monitor the trajectory continuity and assess the evolution trend of damage behavior.

Images