A method of measuring thickness based on electromagnetic ultrasonic signals

Abstract

The application relates to the technical field of ultrasonic thickness measurement, in particular to a thickness measurement method based on electromagnetic ultrasonic signals. First, an echo signal of a measured object is acquired through an electromagnetic ultrasonic transducer; then, the echo signal is sequentially subjected to fixed gain amplification, primary band pass filtering, adjustable gain amplification and secondary band pass filtering to obtain a filtered signal; subsequently, the signal-to-noise ratio of the filtered signal is calculated, and a corresponding processing strategy is selected according to the signal-to-noise ratio; finally, the initial thickness is calculated according to the peak value interval and the ultrasonic wave speed, the average thickness is obtained after averaging, and the final thickness is obtained through relative standard deviation verification, so that the thickness measurement precision is effectively improved.

Description

Technical Field

[0001] This invention relates to the field of ultrasonic thickness measurement technology, specifically a thickness measurement method based on electromagnetic ultrasonic signals. Background Technology

[0002] In the field of industrial equipment operation and maintenance, especially in industries such as petrochemicals and energy, monitoring the wall thickness of critical pressure-bearing equipment such as pipelines and storage tanks is a core aspect of ensuring the safe operation of equipment.

[0003] In traditional wall thickness measurement techniques, piezoelectric ultrasonic thickness measurement is widely used. It uses a piezoelectric transducer to excite ultrasonic waves and calculates the thickness by utilizing the time difference of the ultrasonic waves' transit time. However, this technique has inherent drawbacks: a couplant needs to be applied between the transducer and the surface of the workpiece being measured to ensure sound energy transmission. The use of the couplant is limited by factors such as temperature and environmental corrosion. Specifically, the couplant is prone to failure under high-temperature conditions, and in highly corrosive environments, it can accelerate surface damage to the equipment. Furthermore, for rough surfaces or irregularly shaped structures (such as pipe bends and tees), it is difficult to apply the couplant evenly, leading to decreased measurement accuracy or even failure to measure.

[0004] To address the aforementioned issues, a non-contact measurement device (electromagnetic ultrasonic transducer, EMAT) has been developed. EMAT establishes a static bias magnetic field on the surface of the workpiece using a permanent magnet. An alternating current is passed through a high-frequency coil to generate a dynamic magnetic field, which, based on the Lorentz force or magnetostrictive effect, excites ultrasonic waves inside the workpiece. The ultrasonic waves propagate to the bottom surface and are reflected, then the EMAT receives the echo signal. This non-contact testing eliminates the need for coupling agents and requires no pretreatment of the workpiece surface, making it suitable for measuring irregularly shaped structures under harsh conditions.

[0005] However, EMAT thickness measurement also suffers from a key performance bottleneck: its energy conversion efficiency is far lower than that of piezoelectric transducers, requiring significantly greater excitation energy to generate effective ultrasonic waves. This results in weak echo signal amplitudes (often in the microvolt range) with high background noise and a low signal-to-noise ratio (SNR). In sub-micron high-precision thickness measurement scenarios, low SNR signals can lead to echo peak positioning errors, consequently causing thickness calculation errors and failing to meet the early monitoring requirements for minute wall thickness reduction. To address these issues, current methods often employ fixed signal amplification, filtering, or averaging algorithms without considering the dynamic differences in signal quality during actual testing. For signals with high SNR, overly complex processing increases computational load and reduces detection efficiency; for signals with low SNR, simple processing cannot effectively suppress noise, making it difficult to guarantee measurement accuracy. Furthermore, existing methods lack a mechanism to verify the dispersion of measurement data, making it impossible to determine the reliability of thickness calculation results, further limiting the application of EMAT in high-precision thickness measurement. Therefore, these issues urgently need to be addressed. Summary of the Invention

[0006] In order to avoid and overcome the technical problems existing in the prior art, the present invention provides a thickness measurement method based on electromagnetic ultrasonic signals, which can effectively improve the thickness measurement accuracy.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A thickness measurement method based on electromagnetic ultrasonic signals includes the following thickness measurement steps:

[0009] S1. Obtain echo signals through an electromagnetic ultrasonic transducer;

[0010] S2. Filter the echo signal to obtain the filtered signal;

[0011] S3. Calculate the signal-to-noise ratio of the filtered signal, select the corresponding signal processing strategy based on the signal-to-noise ratio, convert the filtered signal into a detection signal, and obtain the signal peak value of the detection signal.

[0012] S4. Calculate the thickness based on the signal peak value.

[0013] As a further aspect of the present invention, the process of obtaining the filtered signal is as follows:

[0014] S21. Fixed gain amplification: A fixed gain amplification circuit is used to amplify the echo signal. Amplification is performed to obtain the amplified signal. ,and , For fixed gain, Indicates time;

[0015] S22, First-order bandpass filtering: A first-order bandpass filter circuit consisting of a low-pass filter and a high-pass filter connected in series is used for... The frequency response function of this first-order bandpass filter circuit is as follows: ;

[0016] A bandpass filtering process is as follows: First, the bandpass filter is applied to the filter... Perform a Fourier transform to obtain its frequency domain signal, and then combine the frequency domain signal with... Multiply the products, and finally perform an inverse Fourier transform on the product to obtain the signal. , , Indicates Fourier transform, Indicates the inverse Fourier transform;

[0017] S23. Adjustable gain amplification: An adjustable gain amplification circuit is used for... The signal is amplified twice to obtain the amplified signal. , , Indicates adjustable gain;

[0018] S24. Secondary Bandpass Filter: A secondary bandpass filter circuit with the same structure as the primary bandpass filter circuit is used for... The frequency response function of this secondary bandpass filter circuit is as follows: The filtering process is the same as the first bandpass filtering process, and thus the result is obtained. Finally, the filtered signal after filtering is obtained. .

[0019] As a further aspect of the present invention, the signal processing strategy selection process is as follows:

[0020] S31. Calculate the signal-to-noise ratio of the filtered signal using the following formula:

[0021] ;

[0022] ;

[0023] ;

[0024] In the formula, Indicates the signal-to-noise ratio; This represents the signal power of the effective signal region in the filtered signal; This represents the signal power in the noisy region of the filtered signal; Represents a logarithmic function; Indicates time The echo signal changes; express The square of the absolute value; Indicates the effective signal area The sum of signal power in; Indicates the noise area The sum of signal power in; This indicates an average value operation; Indicates the signal location number;

[0025] S32. Calculate the signal-to-noise ratio (SNR) value and select the corresponding signal processing strategy based on the selection principle. The selection principle is expressed as follows:

[0026] When SNR≥25dB, a simple threshold judgment is used to obtain the peak value of the filtered signal.

[0027] When 15dB≤SNR<25dB, the peak value of the filtered signal is obtained by averaging the signal 32 times and cross-correlation.

[0028] When 5dB≤SNR<15dB, the peak value of the filtered signal is obtained by averaging 64 times + bandpass filtering + cross-correlation calculation.

[0029] If SNR < 5dB, abandon this measurement.

[0030] As a further aspect of the present invention, the division of the effective signal region and the noise region is as follows:

[0031] First, locate the global peak of the filtered signal and record its position on the time axis as follows: The effective signal length of the filtered signal is defined as... , This indicates the total length of the filtered signal;

[0032] Next, the effective signal region is defined as... , This represents the absolute value operation; Indicates the filtered signal number 1 Time axis coordinates at each location; express The time axis coordinates at that location;

[0033] Finally, the noise region is defined as , This represents the total area of ​​the filtered signal.

[0034] As a further aspect of the present invention: when SNR≥25dB, the filtered signal at this time That is, the detection signal The process of obtaining the peak value of the detected signal using a simple threshold is as follows:

[0035] S3A1, Set the peak threshold as Peak spacing is The candidate peak set is , , Indicates the detection signal The Middle Time axis coordinates of each amplitude, Indicates the detection signal middle The amplitude at that point;

[0036] S3A2, For candidate peak set Each amplitude in the range, when it is in its position interval When the amplitude is the maximum value, then this amplitude is stored as the signal peak value in the target peak value set. If the signal peak value is in the target peak set, it is stored elsewhere; otherwise, it is not stored. The storage format is as follows:

[0037] ;

[0038] In the formula, Indicates the detection signal middle The signal peak at that location is in the target peak set. The storage format within.

[0039] As a further aspect of the present invention: when 15dB≤SNR<25dB, the process of obtaining the peak signal of the filtered signal by averaging 32 times and cross-correlation is as follows:

[0040] S3B1. Control the electromagnetic ultrasonic transducer to repeatedly acquire echo signals 32 times under the same detection parameters, and obtain 32 sets of filtered signals after filtering, denoted as... , Indicates the first The filtered signal is obtained after the second acquisition and filtering process. Indicates the collection sequence number. ;

[0041] S3B2. The 32 groups of filtered signals from step S3B1 are superimposed at different time points and then averaged to obtain the averaged signal. ;

[0042] S3B3, with The detection signal is obtained by cross-correlation calculation based on discrete sampling points, using the input as input. The cross-correlation calculation formula is as follows:

[0043] ;

[0044] In the formula, for The total number of discrete sampling points; This is a delay parameter, and its value range is... ; for In the The conjugate signal at each sampling point Indicates the discrete sampling point number;

[0045] S3B4, will Timeline Divide into W equal segments, each segment having a length of... , of which Each segment is represented as ;

[0046] S3B5, in each segment Inside, extract The maximum amplitude is taken as the candidate peak value for this segment, and the time axis coordinate of this candidate peak value is... The corresponding amplitude is , express In segments The time axis coordinate of the largest amplitude, express In segments Central amplitude at ;

[0047] S3B6, Collect all segments Candidate peaks were identified, and a preliminary set of peak locations was formed. With the initial peak set ; express The first in Each segment express In segments The maximum amplitude value in;

[0048] S3B7, using the critical points between adjacent segments Create a dynamic window at the midpoint of the window length, and each dynamic window in Set sequentially along the timeline; the window length of the dynamic window is... Adjust according to adjacent segments The adjustments are as follows:

[0049] ;

[0050] In the formula, α is the height control parameter, and β is the basic length;

[0051] S3B8, a dynamic window covers the regions on both sides of the critical point to form sub-segments. ;when hour, ;when hour, ;

[0052] S3B9, Extract each sub-segment Inside The maximum amplitude is used as the supplementary peak value, and the time axis coordinate of the supplementary peak value is... The corresponding peak size is ; express In sub-segments The time axis coordinate of the maximum amplitude in the data. express In sub-segments Central amplitude at ;

[0053] S3B10. Collect the supplementary peaks of all sub-segments to form a set of supplementary peak locations. With peak size set ;

[0054] S3B11. Set a sliding window on the timeline and set the initial window length of the sliding window to... , The value is for all segments. The average length, and the initial sliding step size are set to , The value is 1 / 2 of the dynamic window length;

[0055] S3B12, The sliding window starts to slide sequentially along the time axis starting point according to the initial sliding step size, and obtains the current sliding window content during each sliding process. The maximum amplitude, if this maximum amplitude is within the range of the previous sliding window position. If the time axis coordinate difference between the maximum amplitudes is less than the set spacing threshold, then the next sliding step size will be increased. , For the first This refers to the sliding step size in the current sliding process. For the first "Next" refers to the sliding step size in the next sliding process. This represents the average time axis coordinate difference between adjacent maximum amplitude values ​​acquired at each position during the sliding window's movement; conversely, the window length for the next sliding process is reduced to... , For the sliding window This refers to the window length during the current sliding process. For the sliding window The next slide refers to the length of the window during the next sliding process;

[0056] S3B13. Globally cover the time axis using a sliding window, extract the maximum amplitude within the sliding window at each sliding position, record each maximum amplitude as the signal peak value, and store it in the target peak value set. middle.

[0057] As a further aspect of the present invention: when 5dB≤SNR<15dB, the steps for obtaining the peak signal of the filtered signal by averaging 64 times + bandpass filtering + cross-correlation calculation are as follows:

[0058] S3C1 controls the electromagnetic ultrasonic transducer to repeatedly acquire 64 echo signals under the same detection parameters, and obtains 64 sets of filtered signals after filtering, denoted as... , Indicates the first The filtered signal obtained after the first acquisition and subsequent filtering process. Indicates the collection sequence number. ;

[0059] S3C2: The 64 groups of filtered signals are superimposed at different time points and then averaged to obtain the averaged signal. ;

[0060] S3C3, to Perform a Fourier transform to convert the time-domain signal into a frequency-domain signal; multiply the frequency-domain signal by the frequency response function of the bandpass filter module; and perform an inverse Fourier transform on the multiplied frequency-domain signal to restore it to the time-domain filtered signal. ;

[0061] S3C4, with For input, through The detection signal is obtained by cross-correlation calculation. The cross-correlation calculation formula is as follows:

[0062] ;

[0063] In the formula, for The number of discrete sampling points; This is a delay parameter, and its value range is... ; for In the The conjugate signal at each sampling point, where b represents the discrete sampling point number;

[0064] S3C5, three core parameters are preset: initial peak spacing Peak spacing increase factor λ and tolerance ;but The theoretical number of segments is The actual number of segments is , Indicates to Round down;

[0065] S3C6, will Divided into segments according to the actual number of segments Spacing range Within each interval range, find local peak points and record the time axis coordinates corresponding to those local peak points. Collect all time axis coordinates to form an initial set of peak locations t D ={ }, , Indicates the first Time axis coordinates corresponding to each local peak point; Indicates the first Time axis coordinates corresponding to each local peak point;

[0066] S3C7, based on t D Calculate the distance between adjacent local peak points, and calculate the average value of the distance between adjacent local peak points. , express and The spacing between them;

[0067] Then based on the average value Calculate the standard deviation of the interval ;

[0068] S3C8, Calculate the standard deviation With tolerance In comparison, when the standard deviation If the distance is less than the given tolerance ε, the current spacing is considered appropriate and no adjustment is needed; proceed to step S3C9. When the standard deviation of the spacing... When the value exceeds the tolerance ε, the initial peak spacing is increased to... Then, obtain the corresponding actual number of segments, and then re-execute steps S3C5 to S3C7 until the standard deviation of the interval between adjacent peaks is reached. If the distance is less than the given tolerance ε, then the distance is the optimal distance.

[0069] S3C9, When the peak spacing meets the stability condition When, the corresponding t at this time D The set of final peak positions corresponds to the time axis coordinates of the detected signal. The effective peak time point is identified, and the amplitude corresponding to the effective peak time point is recorded as the signal peak value and stored in the target peak set P.

[0070] As a further aspect of the present invention, the thickness calculation process is as follows:

[0071] S41. Calculate the peak spacing between two adjacent signal peaks along the time axis in the target peak set P:

[0072] S42. Multiply each peak spacing obtained in step S41 by the sound velocity of the ultrasonic wave in the workpiece being tested. Half of each product is the initial thickness corresponding to each peak spacing.

[0073] S43. Calculate the average value of the initial thickness, which is the final thickness of the workpiece being measured.

[0074] As a further aspect of the present invention: calculating the relative standard deviation of the initial thickness. :

[0075] ;

[0076] In the formula, This represents the total initial thickness in the target peak set P; Indicates the first An initial thickness;

[0077] When the relative standard deviation (RSD) is less than or equal to 0.5%, it indicates that the initial thickness can be used for the final thickness calculation; when the relative standard deviation (RSD) is greater than 0.5%, it indicates that the initial thickness calculation error is large. Adjust the gain signal amplification circuit and repeat steps S3 to S4 until the relative standard deviation (RSD) is less than or equal to 0.5%.

[0078] Compared with the prior art, the beneficial effects of the present invention are:

[0079] 1. This invention uses an electromagnetic ultrasonic transducer to acquire echo signals. Leveraging the non-contact, coupling-free characteristics of electromagnetic ultrasound, it avoids the reliance on workpiece surface pretreatment required by traditional piezoelectric ultrasound. It is adaptable to complex testing environments such as high temperatures and strong corrosion, while minimizing physical damage to the workpiece surface, ensuring testing safety and equipment integrity. Initial filtering removes noise interference from the echo signal, providing a cleaner signal source for subsequent signal analysis. Crucially, it introduces the signal-to-noise ratio (SNR) as the basis for selecting signal processing strategies, breaking away from the traditional "one-size-fits-all" approach. The processing scheme can be dynamically adjusted based on signal quality. When the SNR is high, a simple algorithm ensures testing efficiency; when the SNR is low, multiple rounds of optimization algorithms improve signal quality. This avoids resource waste and ensures testing reliability under different signal quality conditions. Finally, thickness calculation is achieved by combining the signal peak value. The overall process is logically clear, highly practical, and balances testing efficiency and accuracy requirements.

[0080] 2. The different peak-finding methods in this invention are designed for different signal-to-noise ratio scenarios, each with its own advantages and strong adaptability. Firstly, for high signal-to-noise ratio scenarios with SNR≥25dB, a simple threshold judgment peak-finding method is used, which does not require complex calculations. The signal peaks can be filtered out simply by setting the peak threshold and peak spacing, which greatly simplifies the operation process, can quickly obtain effective peak information, significantly improves detection efficiency, and can accurately eliminate secondary peak interference in high signal-to-noise ratio environments, ensuring the accuracy of peak identification. Secondly, for medium signal-to-noise ratio scenarios of 15dB≤SNR<25dB, a method of "32 signal averaging + cross-correlation calculation + three-level peak finding (equal segmentation + dynamic sub-segmentation + sliding window)" is adopted. The 32 signal averaging and cross-correlation calculation first enhance the signal periodicity and reduce noise. The subsequent three-level peak finding forms a progressive logic of "coarse positioning - local correction - global fine-tuning". Equal segmentation quickly establishes the peak framework, dynamic sub-segmentation makes up for the omission of peaks at the segment boundary, and the sliding window eliminates redundant detection and adapts to the local features of the signal by adaptively adjusting parameters, thereby comprehensively improving the completeness and accuracy of peak identification. Thirdly, for low signal-to-noise ratio scenarios of 5dB≤SNR<15dB, the method of "64 signal averaging + bandpass filtering + cross-correlation calculation + adaptive spacing segmented peak finding" is adopted. The 64 signal averaging and additional bandpass filtering further enhance the noise suppression effect. The adaptive spacing segmented peak finding dynamically adjusts the segment spacing by setting the initial spacing, increasing the factor and tolerance. The standard deviation of the spacing between adjacent peaks is used as the judgment criterion to ensure that periodic peaks can still be found stably when the signal quality is poor. This effectively avoids peak misjudgment or omission under low signal-to-noise ratio and ensures the reliability of subsequent thickness calculation. Attached Figure Description

[0081] Figure 1 This is a flowchart of the thickness measurement process of the present invention.

[0082] Figure 2 This is a waveform diagram of the initial echo signal in an embodiment of the present invention.

[0083] Figure 3 This is a waveform diagram of the initial echo signal after 64 averaging operations in an embodiment of the present invention.

[0084] Figure 4 for Figure 3 Waveform after bandpass filtering.

[0085] Figure 5 for Figure 4 Waveform after filtering by cross-correlation algorithm.

[0086] Figure 6 for Figure 5 Waveform diagram after envelope detection calculation. Detailed Implementation

[0087] 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, and 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.

[0088] Please see Figure 1 The embodiments of the present invention mainly include the following contents.

[0089] I. Acquiring Echo Signals

[0090] In this embodiment, an electromagnetic ultrasonic transducer (EMAT) is used to obtain echo signals. The EMAT consists of a permanent magnet, a high-frequency coil, an excitation module, a receiving module, and a control and data transmission module. A static bias magnetic field is established on the surface of the workpiece being tested. After an alternating current is applied, ultrasonic waves are excited based on the Lorentz force / magnetostriction effect, and the echo signal reflected from the bottom surface is received.

[0091] The excitation module includes a full-bridge inverter circuit, a power drive circuit, and an impedance matching circuit. It outputs an excitation voltage of 100-800V, an excitation frequency of 1-10MHz (center frequency of 5MHz), and 1-3 pulses, providing a stable excitation signal for EMAT.

[0092] The receiving module includes a signal amplification circuit and a filtering circuit. The signal amplification circuit includes a fixed-gain amplifier (20dB gain) and an adjustable-gain amplifier (0-80dB), amplifying the microvolt-level echo signal received by EMAT to a processable range. The filtering circuit adopts a bandpass filter structure of "low-pass filter + high-pass filter" in series. The cutoff frequencies of the first and second bandpass filters are both 3.5MHz (lower) - 6.5MHz (upper), accurately filtering out signals in the noise frequency band.

[0093] The control and data transmission module uses an FPGA as the main control chip, coordinates the timing of the excitation and receiving modules, and transmits the processed signals to the computer for software analysis.

[0094] The electromagnetic ultrasonic transducer was controlled under set detection parameters (excitation frequency 5MHz, number of pulses 2, excitation voltage 500V). In this embodiment, the workpiece used for testing was a 20# steel pipe with a sound velocity of 3247m / s. The echo signal of the workpiece was collected and recorded as the initial echo signal. The signal time range is t∈[0,T], where T is the total signal length, determined by the sampling frequency. At a sampling frequency of 5MHz, T=2×10⁻⁶. -5 s.

[0095] II. Echo Signal Filtering

[0096] Through the hardware circuitry and software preprocessing of the receiving module, the... The filtered signal is obtained by performing two-stage amplification and two-stage bandpass filtering. The specific process is as follows:

[0097] 1. Fixed gain amplification

[0098] Using a fixed gain amplifier circuit Amplification, amplified signal Fixed gain The value is 20dB, which boosts the microvolt level signal to the millivolt level, making it easier for subsequent filtering.

[0099] 2. First bandpass filter

[0100] Will The input is a first-order bandpass filter circuit, and the filtering process is as follows:

[0101] First of all, Perform Fourier transform It is converted into a frequency domain signal.

[0102] Next, the frequency domain signal is compared with... Multiplication . ,in, , , , .

[0103] This represents the frequency response function of a low-pass filter, where j is the imaginary unit and f is the frequency variable. It is the cutoff frequency of the low-pass filter. This represents the frequency response function of a low-pass filter; It is the cutoff frequency of the high-pass filter.

[0104] Finally, perform an inverse Fourier transform on the product to obtain the signal. , .

[0105] 3. Adjustable gain amplification

[0106] according to The amplitude is set to an adjustable gain. ,right Secondary amplification, amplified signal This ensures that the signal amplitude is within the optimal range for analysis.

[0107] 4. Secondary bandpass filter

[0108] A secondary bandpass filter circuit with the same structure as the primary bandpass filter circuit is used for... The frequency response function of this secondary bandpass filter circuit is as follows: The filtering process is the same as that of a first-pass bandpass filter, that is... Finally, the filtered signal after filtering is obtained. .

[0109] III. Calculating SNR and Selecting Signal Processing Strategies

[0110] Through software calculation The signal-to-noise ratio (SNR) is determined, and the corresponding signal processing strategy is selected based on the SNR range. Converted into detection signal The signal peak value is extracted, and the specific process is as follows:

[0111] 1. Divide the effective signal region into the noise region.

[0112] Search The global peak value is determined and its position on the time axis is recorded as . The effective signal length of the filtered signal is defined as... (3% of the total signal length) is the effective signal area. , Indicates the signal location number. This represents the absolute value operation; Indicates the filtered signal number 1 Time axis coordinates at each location; express The time axis coordinates at that location. The noise region is defined as... , This represents the total area of ​​the filtered signal.

[0113] 2. SNR Calculation Formula

[0114] Effective signal power noise power ,but .

[0115] 3. Select a signal processing strategy

[0116] Based on the calculated signal-to-noise ratio (SNR) value and the following selection principles, a corresponding signal processing strategy is chosen to achieve a balance between high precision and low computational cost. Each strategy corresponds to its own target peak set. Signal peaks obtained under the same strategy are stored in the same target peak set. middle.

[0117] Strategy 1: When SNR≥25dB, use a simple threshold to determine the peak value of the filtered signal;

[0118] Strategy 2: When 15dB≤SNR<25dB, the peak value of the filtered signal is obtained by averaging the signal 32 times and cross-correlation calculation.

[0119] Strategy 3: When 5dB≤SNR<15dB, use 64 signal averaging + bandpass filtering + cross-correlation calculation to obtain the peak value of the filtered signal;

[0120] Strategy 4: If SNR < 5dB, abandon this measurement.

[0121] 4. Strategy 1

[0122] Taking the surface flatness inspection of 20# steel as an example, The SNR is 28dB, so directly... As a detection signal The signal peak value is extracted through the following steps:

[0123] (1) Parameter setting: Set the peak threshold Peak spacing Candidate peak set .

[0124] (2) Target peak selection: For each Determine if it is in the interval Is the amplitude the maximum value? If it is the maximum value, then this amplitude is taken as the signal peak value and stored in the target peak value set P. Conversely, do not store.

[0125] 5. Strategy Two

[0126] Taking the inspection of bends in petrochemical pipelines as an example, The SNR is 18dB. The calculation method is "32 signal averaging + cross-correlation". The specific steps are as follows:

[0127] (1) Multiple signal acquisition: The EMAT is controlled to repeatedly acquire 32 echo signals under the same parameters, and 32 filtered signals are obtained after filtering. , Indicates the first The filtered signal obtained after the first acquisition and subsequent filtering process. Indicates the collection sequence number. .

[0128] (2) Signal averaging: The average signal is obtained by superimposing the 32 groups of signals at different time points. To reduce the impact of random noise.

[0129] (3) Cross-correlation filtering: Using the input signal, the periodicity of the signal is enhanced through cross-correlation calculation to obtain the detection signal. ,in for The total number of discrete sampling points; This is a delay parameter, and its value range is... ; for In the The conjugate signal at each sampling point.

[0130] (4) Level 3 peak detection

[0131] Peak finding by segmentation: Timeline Divide into W=10 equal segments, each segment having a length of The maximum amplitude value within each segment is extracted as a candidate peak value to form a preliminary peak set.

[0132] Dynamic sub-segment peak finding: using adjacent segment boundaries Centered on, according to Adjust the window length to cover the areas on both sides of the boundary and extract supplementary peak values ​​within the sub-segments.

[0133] Sliding window peak finding: Initial window length Initial step size Slide the window along the time axis; if the peak interval between adjacent windows is less than a threshold... Then increase the step size. If the amplitude changes abruptly, the window length should be reduced. Finally, the global peak value is extracted and stored in the target peak value set. .

[0134] 6. Strategy Three

[0135] Taking high-temperature corrosion pipeline inspection as an example, The SNR is 9.1dB. The calculation method is "64 signal averaging + bandpass filtering + cross-correlation". The specific steps are as follows:

[0136] (1) 64 signal acquisitions and averaging: Under the same detection parameters, the electromagnetic ultrasonic transducer was controlled to acquire 64 sets of filtered signals and average them. Similarly, the corresponding signals were obtained. At this point, the SNR increased to 19.65dB.

[0137] (2) Software bandpass filtering: for Perform a Fourier transform, multiply it by the software-implemented bandpass filter function (cutoff frequency 2MHz-8MHz), and then perform an inverse transform to obtain the time-domain filtered signal. The SNR was further improved to 23.31dB.

[0138] (3) Cross-correlation filtering: according to Calculation, obtained The SNR improved to 26.19 dB.

[0139] (4) Adaptive spacing peak detection:

[0140] First, set the initial peak spacing. Increase factor λ = 1.2, tolerance ε = 0.1 × 10 -6 s, calculate the theoretical number of segments Actual number of segments .

[0141] Will Divide the data into D interval ranges, extract the local peak values ​​in each range, and form an initial peak position set t. D .

[0142] Calculate the average value of the distance between adjacent peaks. with standard deviation ,like If the initial peak spacing is greater than ε, then update the initial peak spacing as follows: Repeat the segmentation and calculation until... ≤ε, at this time t D The final peak position set is given, and the corresponding amplitude is stored in the target peak set P.

[0143] 7. Comprehensive Analysis

[0144] The above signal-to-noise ratio (SNR) classification method is determined based on the actual effect of the noise reduction mechanism in electromagnetic ultrasonic thickness measurement, the balance of system computational complexity, engineering scenario constraints, and sub-micron thickness measurement accuracy requirements. The average SNR improvement after processing is Δ. avg =5log 10 N (where N is the number of samplings), 32 samplings can improve the average by about 7.5dB, and 64 samplings can improve the average by about 9dB; cross-correlation processing (number of independent pulses M=5) can improve the average by about 7dB; bandpass filtering can conservatively improve the average by about 8dB under complex engineering conditions (normal range 6-12dB).

[0145] When SNR≥25dB, the signal quality meets the requirements for high-precision thickness measurement. The simple threshold method (such as setting a peak threshold) has a small computational load and can accurately extract the peak without additional noise reduction processing. It can ensure submicron level accuracy and avoid complex calculations occupying FPGA cache or computer resources, which is in line with the principle of "computation controllable".

[0146] When 15dB ≤ SNR < 25dB, accurate noise reduction is achieved through "32-times signal averaging + cross-correlation calculation". Noise in this range is primarily random, with a low proportion of non-target frequency noise, eliminating the need for bandpass filtering. After combining the 32-times averaging (7.5dB improvement) with cross-correlation processing (7dB improvement), even considering marginal effects, a total improvement of ≥12dB can still be achieved, raising the minimum SNR of 15dB to approximately 27dB, thus meeting the high SNR requirement.

[0147] When 5dB ≤ SNR < 15dB, a combined approach of "64-times signal averaging + bandpass filtering + cross-correlation calculation" is required. This noise range contains a large amount of random noise and non-target frequency noise (such as electromagnetic interference at the engineering site and probe coupling noise). Single or two noise reduction methods cannot meet the required accuracy. 64-times averaging (improving by 9dB) is more effective at reducing strong random noise than 32-times averaging. Bandpass filtering (improving by 8dB) can specifically filter out interference signals deviating from the center frequency (e.g., 5MHz). Cross-correlation processing (improving by 7dB) enhances the periodicity of the echo signal. The combination of these three methods (considering marginal effects, the total improvement is ≥20dB) can increase the minimum SNR of 5dB to approximately 25dB, effectively avoiding peak shifts caused by noise and ensuring the accuracy of thickness calculations.

[0148] When SNR < 5dB, the signal quality is extremely poor, requiring abandonment of the measurement and parameter adjustment. 5dB corresponds to a signal power that is only 3.2 times the noise and an amplitude that is approximately 1.8 times. Even with the aforementioned enhanced noise reduction combination, due to marginal effects and system drift (such as poor probe coupling), the actual improvement may still be less than 20dB, making it difficult to achieve a processed SNR of 25dB, resulting in an extremely high peak extraction error rate. Further improving SNR requires increasing the number of samples (e.g., averaging 128 times) or upgrading the hardware (e.g., increasing the excitation voltage), which doubles the computation time, increases hardware costs, and does not meet the "regular rapid detection" requirements of scenarios such as petrochemical pipelines. Adjusting the detection parameters (e.g., adjusting the excitation voltage to 100-800V and the gain to 0-80dB) can directly improve the original SNR, which is more efficient and accurate than forcibly processing low signal-to-noise ratio signals.

[0149] IV. Calculating Thickness

[0150] The thickness is calculated based on the peak time coordinates in the target peak set P, combined with the sound velocity of the workpiece, and the accuracy is verified by the relative standard deviation (RSD). The specific steps are as follows:

[0151] 1. Calculate the peak spacing: Extract adjacent peak pairs along the time axis in P and calculate the peak spacing between adjacent peak pairs.

[0152] 2. Calculate the initial thickness: According to the principle of ultrasonic transit time, the initial thickness is half of the product of the peak spacing and the ultrasonic velocity in the workpiece being tested.

[0153] 3. Calculate the final thickness and verify accuracy:

[0154] (1) Calculate the average value of the initial thickness.

[0155] (2) Calculate the relative standard deviation , This represents the total initial thickness in the target peak set P; Indicates the first An initial thickness is calculated. When the relative standard deviation (RSD) is less than or equal to 0.5%, the initial thickness can be used for the final thickness calculation. When the RSD is greater than 0.5%, the initial thickness calculation error is large. Adjust the gain signal amplifier circuit and repeat the second to fourth parts until the RSD is less than or equal to 0.5%.

[0156] V. Verification of Implementation Results

[0157] Taking a standard 20# steel workpiece as an example, with a known thickness of 24.000 mm, the following results were obtained using this method:

[0158] In a certain thickness measurement, the excitation center frequency was 5MHz, the excitation signal was a sinusoidal pulse wave, and the sampling frequency was 5MHz, thus obtaining the initial echo signal. like Figure 2 As shown.

[0159] The signal-to-noise ratio of the initial echo signal was calculated to be 9.1 dB, therefore a method of 64-times averaging + bandpass filtering + cross-correlation was adopted.

[0160] The signal after 64 averages is as follows Figure 3 As shown, its signal-to-noise ratio is 19.65dB.

[0161] Using a bandpass filter algorithm, with an upper cutoff frequency of 2MHz and a lower cutoff frequency of 8MHz, the processed signal is as follows: Figure 4 As shown, its signal-to-noise ratio is 23.31 dB. From Figure 4 As can be seen, the fine noise between different peaks has been cleared to some extent, and the main peak of the signal is more obvious.

[0162] Next, a cross-correlation algorithm is used for filtering, and the resulting signal is as follows: Figure 5 As shown, the final signal-to-noise ratio is 26.19 dB.

[0163] Finally, envelope detection calculations are performed on the obtained detection signal, and the processed signal is as follows: Figure 6 As shown, the signal-to-noise ratio is 26.38dB at this time.

[0164] The calculated thickness values ​​are d1 = 23.92 mm, d2 = 23.92 mm, d3 = 23.92 mm, and d4 = 23.92 mm. The corresponding RSD value is 0.167%, indicating relatively consistent data. The final thickness of 23.92 mm is obtained, with a deviation of only 0.33% from the known thickness, demonstrating high calculation accuracy and meeting industrial application requirements.

[0165] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A thickness measurement method based on electromagnetic ultrasonic signals, characterized in that, The thickness measurement steps include the following: S1. Obtain the echo signal of the workpiece under test through an electromagnetic ultrasonic transducer. S2. Filter the echo signal to obtain the filtered signal; S3. Calculate the signal-to-noise ratio of the filtered signal, select the corresponding signal processing strategy based on the signal-to-noise ratio, convert the filtered signal into a detection signal, and obtain the signal peak value of the detection signal. S4. Calculate the thickness of the workpiece being measured based on the signal peak value; The process of selecting a signal processing strategy is as follows: S31. Calculate the signal-to-noise ratio of the filtered signal using the following formula: ； ； ； In the formula, Indicates the signal-to-noise ratio; This represents the signal power of the effective signal region in the filtered signal; This represents the signal power in the noisy region of the filtered signal; Represents a logarithmic function; Indicates time The echo signal changes; express The square of the absolute value; Indicates the effective signal area The sum of signal power in; Indicates the noise area The sum of signal power in; This indicates an average value operation; Indicates the signal location number; S32. Calculate the signal-to-noise ratio (SNR) value and select the corresponding signal processing strategy based on the selection principle. The selection principle is expressed as follows: When SNR≥25dB, a simple threshold judgment is used to obtain the peak value of the filtered signal. When 15dB≤SNR<25dB, the peak value of the filtered signal is obtained by averaging the signal 32 times and cross-correlation. When 5dB≤SNR<15dB, the peak value of the filtered signal is obtained by averaging 64 times + bandpass filtering + cross-correlation calculation. If SNR < 5dB, this measurement should be abandoned. The effective signal region and noise region are divided as follows: First, locate the global peak of the filtered signal and record its position on the time axis as follows: ; Define the effective signal length of the filtered signal as , This indicates the total length of the filtered signal; Next, the effective signal region is defined as... , This represents the absolute value operation; Indicates the filtered signal number 1 Time axis coordinates at each location; express The time axis coordinates at that location; Finally, the noise region is defined as , This represents the total area of ​​the filtered signal.

2. The thickness measurement method based on electromagnetic ultrasonic signals according to claim 1, characterized in that, The process of obtaining the filtered signal is as follows: S21. Fixed gain amplification: A fixed gain amplification circuit is used to amplify the echo signal. Amplification is performed to obtain the amplified signal. ,and , For fixed gain, Indicates time; S22, First-order bandpass filtering: A first-order bandpass filter circuit consisting of a low-pass filter and a high-pass filter connected in series is used for... The frequency response function of this first-order bandpass filter circuit is as follows: ; A bandpass filtering process is as follows: First, the bandpass filter is applied to the filter... Perform a Fourier transform to obtain its frequency domain signal, and then combine the frequency domain signal with... Multiply the products, and finally perform an inverse Fourier transform on the product to obtain the signal. , , Indicates Fourier transform, Indicates the inverse Fourier transform; S23. Adjustable gain amplification: An adjustable gain amplification circuit is used for... The signal is amplified twice to obtain the amplified signal. , , Indicates adjustable gain; S24. Secondary Bandpass Filter: A secondary bandpass filter circuit with the same structure as the primary bandpass filter circuit is used for... The frequency response function of this secondary bandpass filter circuit is as follows: The filtering process is the same as the first bandpass filtering process, and thus the result is obtained. Finally, the filtered signal after filtering is obtained. .

3. The thickness measurement method based on electromagnetic ultrasonic signals according to claim 2, characterized in that, When SNR≥25dB, the filtered signal at this time That is, the detection signal The process of obtaining the peak value of the detected signal using a simple threshold is as follows: S3A1, Set the peak threshold as Peak spacing is The candidate peak set is , , Indicates the detection signal The Middle Time axis coordinates of each amplitude, Indicates the detection signal middle The amplitude at that point; S3A2, For candidate peak set Each amplitude in the range, when it is in its position interval When the amplitude is the maximum value, then this amplitude is stored as the signal peak value in the target peak value set. If the signal peak value is in the target peak set, it is stored elsewhere; otherwise, it is not stored. The storage format is as follows: ； In the formula, Indicates the detection signal middle The signal peak at that location is in the target peak set. The storage format within.

4. The thickness measurement method based on electromagnetic ultrasonic signals according to claim 3, characterized in that, When 15dB≤SNR<25dB, the process of obtaining the peak signal of the filtered signal by averaging 32 signals and cross-correlation is as follows: S3B1. Control the electromagnetic ultrasonic transducer to repeatedly acquire echo signals 32 times under the same detection parameters, and obtain 32 sets of filtered signals after filtering, denoted as... , Indicates the first The filtered signal is obtained after the second acquisition and filtering process. Indicates the collection sequence number. ; S3B2. The 32 groups of filtered signals from step S3B1 are superimposed at different time points and then averaged to obtain the averaged signal. ; S3B3, with The detection signal is obtained by cross-correlation calculation using the input signal. The cross-correlation calculation formula is as follows: ； In the formula, for The total number of discrete sampling points; This is a delay parameter, and its value range is... ; for In the The conjugate signal at each sampling point, where j represents the discrete sampling point number; S3B4, will Timeline Divide into W equal segments, each segment having a length of... , of which Each segment is represented as ; S3B5, in each segment Inside, extract The maximum amplitude is taken as the candidate peak value for this segment, and the time axis coordinate of this candidate peak value is... The corresponding amplitude is , express In segments The time axis coordinate of the largest amplitude, express In segments Central amplitude at ; S3B6, Collect all segments Candidate peaks were identified, and a preliminary set of peak locations was formed. With the initial peak set ; express The first in Each segment express In segments The maximum amplitude value in; S3B7, using the critical points between adjacent segments Create a dynamic window at the midpoint of the window length, and each dynamic window in Set sequentially along the timeline; the window length of the dynamic window is... Adjust according to adjacent segments The adjustments are as follows: ； In the formula, α is the height control parameter, and β is the basic length; S3B8, a dynamic window covers the regions on both sides of the critical point to form sub-segments. ;when hour, ;when hour, ; S3B9, Extract each sub-segment Inside The maximum amplitude is used as the supplementary peak value, and the time axis coordinate of the supplementary peak value is... The corresponding peak size is ; express In sub-segments The time axis coordinate of the maximum amplitude in the data. express In sub-segments Central amplitude at ; S3B10. Collect the supplementary peaks of all sub-segments to form a set of supplementary peak locations. With peak size set ; S3B11. Set a sliding window on the timeline and set the initial window length of the sliding window to... , The value is for all segments. The average length, and the initial sliding step size are set to , The value is 1 / 2 of the dynamic window length; S3B12, The sliding window starts to slide sequentially along the time axis starting point according to the initial sliding step size, and obtains the current sliding window content during each sliding process. The maximum amplitude, if this maximum amplitude is within the range of the previous sliding window position. If the time axis coordinate difference between the maximum amplitudes is less than the set spacing threshold, then the next sliding step size will be increased. , For the first This refers to the sliding step size in the current sliding process. For the first "Next" refers to the sliding step size in the next sliding process. This represents the average time axis coordinate difference between adjacent maximum amplitude values ​​acquired at each position during the sliding window's movement; conversely, the window length for the next sliding process is reduced to... , For the sliding window This refers to the window length during the current sliding process. For the sliding window The next slide refers to the length of the window during the next sliding process; S3B13. Globally cover the time axis using a sliding window, extract the maximum amplitude within the sliding window at each sliding position, record each maximum amplitude as the signal peak value, and store it in the target peak value set. middle.

5. The thickness measurement method based on electromagnetic ultrasonic signals according to claim 1, characterized in that, When 5dB≤SNR<15dB, the steps to obtain the peak value of the filtered signal using 64-times signal averaging + bandpass filtering + cross-correlation calculation are as follows: S3C1 controls the electromagnetic ultrasonic transducer to repeatedly acquire 64 echo signals under the same detection parameters, and obtains 64 sets of filtered signals after filtering, denoted as... , Indicates the first The filtered signal obtained after the first acquisition and subsequent filtering process. Indicates the collection sequence number. ; S3C2: The 64 groups of filtered signals are superimposed at different time points and then averaged to obtain the averaged signal. ; S3C3, to Perform a Fourier transform to convert the time-domain signal into a frequency-domain signal; multiply the frequency-domain signal by the frequency response function of the bandpass filter module; and perform an inverse Fourier transform on the multiplied frequency-domain signal to restore it to the time-domain filtered signal. ; S3C4, with The detection signal is obtained by cross-correlation calculation using the input signal. The cross-correlation calculation formula is as follows: ； In the formula, for The number of discrete sampling points; This is a delay parameter, and its value range is... ; for In the The conjugate signal at each sampling point, where b represents the discrete sampling point number; S3C5, three core parameters are preset: initial peak spacing Peak spacing increase factor λ and tolerance ;but The theoretical number of segments is The actual number of segments is , Indicates to Round down; S3C6, will Divided into segments according to the actual number of segments Spacing range Within each interval range, find local peak points and record the time axis coordinates corresponding to those local peak points. Collect all time axis coordinates to form an initial set of peak locations t D ={ }, , Indicates the first Time axis coordinates corresponding to each local peak point; Indicates the first Time axis coordinates corresponding to each local peak point; S3C7, based on t D Calculate the distance between adjacent local peak points, and calculate the average value of the distance between adjacent local peak points. , express and The spacing between them; Then based on the average value Calculate the standard deviation of the interval ; S3C8, Calculate the standard deviation With tolerance In comparison, when the standard deviation If the distance is less than the given tolerance ε, the current spacing is considered appropriate and no adjustment is needed; proceed to step S3C9. When the standard deviation of the spacing... When the value exceeds the tolerance ε, the initial peak spacing is increased to... Then, obtain the corresponding actual number of segments, and then re-execute steps S3C5 to S3C7 until the standard deviation of the interval between adjacent peaks is reached. If the distance is less than the given tolerance ε, then the distance is the optimal distance. S3C9, When the peak spacing meets the stability condition When, the corresponding t at this time D The set of final peak positions corresponds to the time axis coordinates of the detected signal. The effective peak time point is identified, and the amplitude corresponding to the effective peak time point is recorded as the signal peak value and stored in the target peak set P.

6. A thickness measurement method based on electromagnetic ultrasonic signals according to any one of claims 3-5, characterized in that, The thickness calculation process is as follows: S41. Calculate the peak spacing between two adjacent signal peaks along the time axis in the target peak set P: S42. Multiply each peak spacing obtained in step S41 by the sound velocity of the ultrasonic wave in the workpiece being tested. Half of each product is the initial thickness corresponding to each peak spacing. S43. Calculate the average value of the initial thickness, which is the final thickness of the workpiece being measured.

7. The thickness measurement method based on electromagnetic ultrasonic signals according to claim 6, characterized in that, Calculate the relative standard deviation of the initial thickness ; When the relative standard deviation RSD is less than or equal to 0.5%, it indicates that the initial thickness can be used for the final thickness calculation; when the relative standard deviation RSD is greater than 0.5%, adjust the gain signal amplification circuit and repeat steps S3 to S4 until the relative standard deviation RSD is less than or equal to 0.5%.

8. The thickness measurement method based on electromagnetic ultrasonic signals according to claim 7, characterized in that, The formula for calculating the relative standard deviation is as follows: ； In the formula, This represents the total initial thickness in the target peak set P; Indicates the first An initial thickness.

Images