Method for denoising multi-source signals of PE pipeline based on dual-tree complex wavelet transform

Abstract

The application provides a PE pipeline multi-source signal denoising method based on a dual-tree complex wavelet transform, and belongs to the technical field of signal denoising, and the method comprises the following steps: S1, multi-source signal acquisition and pretreatment; S2, signal collaborative calibration; S3, dual-tree complex wavelet decomposition; S4, noise level estimation; S5, adaptive threshold denoising; S6, signal reconstruction; S7, effect evaluation and optimization. The application adopts a single pendulum wave signal as a benchmark signal, and assists in time synchronization and frequency calibration, thereby improving the accuracy of signal processing, effectively suppressing various noise interferences, and significantly improving the signal-to-noise ratio (SNR) of the denoised signal.

Description

Technical Field

[0001] The present invention relates to the technical field of signal noise reduction, and in particular to a PE pipeline multi-source signal noise reduction method based on dual-tree complex wavelet transform. Background Art

[0002] With the rapid development of industries like petrochemicals and city gas, polyethylene (PE) pipes have become widely used in fluid transportation due to their corrosion resistance and flexibility. However, over long-term operation, PE pipes are susceptible to factors such as internal fluid pressure fluctuations, external environmental vibrations, and temperature changes, leading to problems such as cracks, leaks, and blockages. Timely and accurate monitoring of PE pipeline operating conditions is crucial to ensuring safe and stable operation.

[0003] Currently, multi-source signal monitoring technology is widely used for PE pipeline condition monitoring. It collects signals such as pressure, flow, temperature, and vibration to analyze pipeline operating conditions. However, the collected signals are often subject to various noise interferences, such as vibration noise generated by environmental vibrations and electrical signal noise caused by electromagnetic interference. This noise can seriously affect signal quality, making it difficult to extract pipeline fault characteristics, and thus affecting the accuracy and timeliness of fault diagnosis.

[0004] Existing signal denoising methods, such as traditional Fourier transform filtering, can remove noise to a certain extent. However, because they are based on the assumption of signal stationarity, they are not very effective for non-stationary PE pipeline signals and are prone to losing key signal features. Wavelet transform denoising methods, while capable of processing non-stationary signals, suffer from frequency aliasing and limited directional selectivity, making it difficult to accurately separate noise from useful signals. Furthermore, when processing multi-source signals, existing methods often ignore the correlation between signals and the impact of environmental factors on the signals, making it impossible to fully utilize the information contained in multi-source signals and achieving efficient noise reduction. Therefore, a more effective method for denoising multi-source signals in PE pipelines is urgently needed. Summary of the Invention

[0005] The present invention provides a PE pipeline multi-source signal denoising method based on dual-tree complex wavelet transform to solve one or more of the above problems.

[0006] The PE pipeline multi-source signal denoising method based on dual-tree complex wavelet transform includes:

[0007] S1. Multi-source signal acquisition and preprocessing: Collect pressure signals, flow signals, temperature signals, vibration signals, acoustic emission signals, and strain signals from PE pipelines, perform preliminary filtering and time synchronization processing to obtain pre-processed multi-source signals; collect single pendulum wave signals, perform frequency calibration and time stamping to obtain calibrated single pendulum wave signals;

[0008] S2. Signal collaborative calibration; noise analysis of the calibrated single pendulum wave signal to obtain a single pendulum wave reference signal; time synchronization optimization of the pre-processed multi-source signal based on the single pendulum wave reference signal to obtain a time-synchronized multi-source signal; frequency comparison calibration of the time-synchronized multi-source signal and the single pendulum wave reference signal to obtain a calibrated multi-source signal;

[0009] S3. Dual-tree complex wavelet decomposition: Perform dual-tree complex wavelet transform multi-layer decomposition on the calibrated multi-source signal and the single pendulum wave reference signal, respectively, to obtain low-frequency complex coefficients and high-frequency complex detail coefficients of each layer of the multi-source signal, and low-frequency complex coefficients and high-frequency complex detail coefficients of each layer of the single pendulum wave;

[0010] S4. Noise level estimation: Estimating the noise standard deviation of the high-frequency complex detail coefficients of each layer of the multi-source signal, calculating the difference index between the high-frequency complex detail coefficients of each layer of the multi-source signal and the high-frequency complex detail coefficients of each layer of the single pendulum wave, performing correlation analysis between the signals, and obtaining the noise level estimation result;

[0011] S5. Adaptive threshold denoising: Based on the noise level estimation results, a reference signal is selected, a basic threshold is calculated, and a joint threshold adjustment is performed. A soft threshold function is used to threshold the high-frequency complex detail coefficients of each layer of the multi-source signal to obtain the denoised high-frequency complex detail coefficients of each layer.

[0012] S6. Signal reconstruction: Perform a dual-tree complex wavelet inverse transform on the high-frequency complex detail coefficients of each layer after denoising and the low-frequency complex coefficients of each layer of the multi-source signal, reconstruct the signal layer by layer, and obtain a preliminary reconstructed signal; perform a physical relationship-based consistency check on the preliminary reconstructed signal. If the check fails, return to S5 to adjust the threshold. If the check passes, obtain the final denoised signal.

[0013] S7. Effect evaluation and optimization: Evaluate the final noise reduction signal using the signal-to-noise ratio, mean square error, and peak signal-to-noise ratio indicators, and extract pipeline fault features for feature evaluation. If the evaluation indicators do not meet expectations, adjust the threshold parameters or the number of decomposition layers, and return to S3 for reprocessing until the evaluation indicators meet expectations.

[0014] In this manual, the pressure signal is collected through a piezoresistive pressure sensor, the flow signal is collected through an electromagnetic flowmeter or an ultrasonic flowmeter, the temperature signal is collected through a thermocouple or a thermistor temperature sensor, the vibration signal is collected through an acceleration sensor, the acoustic emission signal is collected through an acoustic emission sensor, and the strain signal is collected through a strain gauge; a bandpass filter is used for preliminary filtering, and an interpolation algorithm is used for time synchronization.

[0015] In this specification, a simple pendulum device is installed next to the pipeline. The swing of the simple pendulum follows the law of simple harmonic motion, and the swing signal of the simple pendulum is collected by a displacement sensor.

[0016] In this specification, the dual-tree complex wavelet transform uses two parallel filter banks. The filter banks of tree A and tree B are designed to meet the analytical signal conditions. The number of decomposition layers is determined to be 3 to 5 layers based on the signal length and noise characteristics.

[0017] In this manual, the noise standard deviation is estimated using the median estimation method, the difference index is calculated as the ratio of the sum of the absolute differences between the high-frequency complex detail coefficients of each layer of each signal and the high-frequency complex detail coefficients of each layer of the simple pendulum wave to the sum of the high-frequency complex detail coefficients of each layer of the simple pendulum wave, and the correlation analysis calculates the correlation coefficient between signals and retains the estimated values ​​with absolute values ​​greater than 0.5.

[0018] In this specification, the reference signal is selected in automatic selection mode or manual specification mode. The automatic selection mode calculates the average correlation coefficient of each signal in each layer and selects the signal with the largest average correlation coefficient as the reference signal; the joint threshold adjustment takes into account the correlation between signals and the difference from the single pendulum wave signal.

[0019] In this manual, the dual-tree complex wavelet inverse transform adopts a dual-tree structure corresponding to the decomposition, and upsamples and filters the denoised high-frequency complex detail coefficients of each layer and the unprocessed low-frequency complex coefficients of each layer layer by layer; the consistency check uses the Bernoulli equation to check the pressure signal and flow signal, and the error is required to be less than 5%.

[0020] In this specification, the signal-to-noise ratio improvement requirement for indicator evaluation is greater than 10dB, the mean square error is less than the set threshold, and the fault feature retention rate for feature evaluation is required to be greater than 70%.

[0021] In this manual, time synchronization optimization uses the period of the single pendulum wave reference signal as a reference to perform interpolation calibration on sensor signals whose time deviation exceeds the set threshold; frequency comparison calibration adjusts the bandpass filter parameters of each signal based on the theoretical frequency of the single pendulum wave reference signal.

[0022] In this specification, the high-frequency complex detail coefficients obtained by dual-tree complex wavelet transform using two sets of filters in parallel are subjected to soft threshold processing on their real and imaginary parts respectively, and then synthesized to obtain the denoised high-frequency complex detail coefficients.

[0023] The embodiments of this specification can achieve at least the following beneficial effects:

[0024] High-Precision Noise Reduction: This solution combines dual-tree complex wavelet transform (DWT) and single pendulum wave technology to achieve high-precision noise reduction for multi-source signals in PE pipelines. The DWT utilizes two parallel filter banks, which offer excellent directional selectivity and anti-aliasing capabilities, accurately separating noise from useful components in the signal. The single pendulum wave signal serves as a reference signal, assisting with time synchronization and frequency calibration. This further improves signal processing accuracy, effectively suppresses various types of noise interference, and significantly enhances the signal-to-noise ratio (SNR) of the denoised signal.

[0025] Collaborative multi-source signal processing: This system fully considers the correlation between multiple signal sources and implements collaborative noise reduction through a dynamic reference signal selection mechanism and a joint threshold adjustment strategy. Based on the correlation coefficients between signals and their differences from a single pendulum signal, it adaptively selects reference signals and adjusts thresholds. This effectively removes noise tailored to the characteristics of different signal types while preserving the physical connections between them. This improves the overall performance of multi-source signal processing and provides more reliable data support for comprehensive pipeline status analysis.

[0026] Effective Fault Signature Preservation: During the noise reduction process, pipeline fault characteristics are effectively preserved through the rational design of threshold processing and the optimization of signal reconstruction algorithms. Key fault characteristics, such as high-frequency acoustic emission signals from pipeline leaks and low-frequency pressure and flow signal changes caused by blockages, are clearly visible after noise reduction. This provides a strong guarantee for accurate pipeline fault diagnosis, helps to promptly identify potential pipeline problems, and reduces the risk of accidents.

[0027] Strong adaptability and robustness: This solution is highly adaptable and robust. Through noise level estimation and feedback optimization, it automatically adjusts noise reduction parameters, such as the number of decomposition layers and threshold coefficients, based on the signal characteristics and noise intensity under different operating conditions. This makes the noise reduction method applicable to a variety of complex environments and operating conditions. Whether it is minor disturbances during the initial operation of the pipeline or strong noise that may occur later, this solution can maintain excellent noise reduction results.

[0028] Improved reliability: High-precision noise reduction and effective retention of fault characteristics significantly enhance reliability and accuracy. This reduces misjudgments and missed diagnosis due to noise, improves the timeliness and accuracy of fault diagnosis, and provides a scientific basis for pipeline maintenance and management, helping to reduce maintenance costs, extend pipeline service life, and ensure safe and stable fluid transportation, thus possessing significant economic and social value. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0030] Figure 1 Schematic diagram of a PE pipeline multi-source signal denoising method based on dual-tree complex wavelet transform involved in some embodiments of the present invention.

[0031] Figure 2The figure is a flow chart of a method for denoising multi-source signals in a PE pipeline based on dual-tree complex wavelet transform in some embodiments of the present invention.

[0032] Figure 3 Schematic diagram of generating a single pendulum wave reference signal involved in some embodiments of the present invention.

[0033] Figure 4 Schematic diagram of multi-source signal synchronization optimization involved in some embodiments of the present invention. DETAILED DESCRIPTION

[0034] Hereinafter, only certain exemplary embodiments are briefly described. As will be appreciated by those skilled in the art, the described embodiments may be modified in various ways without departing from the spirit or scope of the present invention. Therefore, the drawings and description are to be considered as illustrative in nature and not restrictive.

[0035] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential" and the like to indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, or are the orientations or positional relationships in which the products of the present invention are conventionally placed when in use, or are the orientations or positional relationships conventionally understood by those skilled in the art. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore should not be understood as limiting the present invention.

[0036] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature specified as "first" or "second" may explicitly or implicitly include one or more of such features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0037] Furthermore, the terms "mounted," "connected," "connect," and "fixed" should be interpreted broadly. For example, they may refer to fixed connection, detachable connection, or integration; they may refer to direct connection or indirect connection through an intermediate medium; they may refer to internal communication between two components or interaction between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0038] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0039] like Figure 1 and Figure 2 As shown, this embodiment provides a PE pipeline multi-source signal noise reduction method based on dual-tree complex wavelet transform, including:

[0040] S1. Multi-source signal acquisition and preprocessing: Collect pressure signals, flow signals, temperature signals, vibration signals, acoustic emission signals, and strain signals from PE pipelines, perform preliminary filtering and time synchronization processing to obtain pre-processed multi-source signals; collect single pendulum wave signals, perform frequency calibration and time stamping to obtain calibrated single pendulum wave signals;

[0041] S2. Signal collaborative calibration; noise analysis of the calibrated single pendulum wave signal to obtain a single pendulum wave reference signal; time synchronization optimization of the pre-processed multi-source signal based on the single pendulum wave reference signal to obtain a time-synchronized multi-source signal; frequency comparison calibration of the time-synchronized multi-source signal and the single pendulum wave reference signal to obtain a calibrated multi-source signal;

[0042] S3. Dual-tree complex wavelet decomposition: Perform dual-tree complex wavelet transform multi-layer decomposition on the calibrated multi-source signal and the single pendulum wave reference signal, respectively, to obtain low-frequency complex coefficients and high-frequency complex detail coefficients of each layer of the multi-source signal, and low-frequency complex coefficients and high-frequency complex detail coefficients of each layer of the single pendulum wave;

[0043] S4. Noise level estimation: Estimating the noise standard deviation of the high-frequency complex detail coefficients of each layer of the multi-source signal, calculating the difference index between the high-frequency complex detail coefficients of each layer of the multi-source signal and the high-frequency complex detail coefficients of each layer of the single pendulum wave, performing correlation analysis between the signals, and obtaining the noise level estimation result;

[0044] S5. Adaptive threshold denoising: Based on the noise level estimation results, a reference signal is selected, a basic threshold is calculated, and a joint threshold adjustment is performed. A soft threshold function is used to threshold the high-frequency complex detail coefficients of each layer of the multi-source signal to obtain the denoised high-frequency complex detail coefficients of each layer.

[0045] S6. Signal reconstruction: Perform a dual-tree complex wavelet inverse transform on the high-frequency complex detail coefficients of each layer after denoising and the low-frequency complex coefficients of each layer of the multi-source signal, reconstruct the signal layer by layer, and obtain a preliminary reconstructed signal; perform a physical relationship-based consistency check on the preliminary reconstructed signal. If the check fails, return to S5 to adjust the threshold. If the check passes, obtain the final denoised signal.

[0046] S7. Effect evaluation and optimization: Evaluate the final noise reduction signal using the signal-to-noise ratio, mean square error, and peak signal-to-noise ratio indicators, and extract pipeline fault features for feature evaluation. If the evaluation indicators do not meet expectations, adjust the threshold parameters or the number of decomposition layers, and return to S3 for reprocessing until the evaluation indicators meet expectations.

[0047] In some embodiments, the pressure signal is collected through a piezoresistive pressure sensor, the flow signal is collected through an electromagnetic flowmeter or an ultrasonic flowmeter, the temperature signal is collected through a thermocouple or a thermistor temperature sensor, the vibration signal is collected through an acceleration sensor, the acoustic emission signal is collected through an acoustic emission sensor, and the strain signal is collected through a strain gauge; preliminary filtering uses a bandpass filter, and time synchronization uses an interpolation algorithm.

[0048] In some embodiments, a simple pendulum device is installed next to the pipeline, the swing of the simple pendulum follows the law of simple harmonic motion, and the swing signal of the simple pendulum is collected by a displacement sensor.

[0049] In some embodiments, the dual-tree complex wavelet transform uses two parallel filter banks. The filter banks of tree A and tree B are designed to meet the analytical signal conditions. The number of decomposition layers is determined to be 3 to 5 layers based on the signal length and noise characteristics.

[0050] In some embodiments, the noise standard deviation is estimated using the median estimation method, and the difference index is calculated as the ratio of the sum of the absolute differences between the high-frequency complex detail coefficients of each layer of each signal and the high-frequency complex detail coefficients of each layer of the simple pendulum wave to the sum of the high-frequency complex detail coefficients of each layer of the simple pendulum wave. The correlation analysis calculates the correlation coefficient between the signals and retains the estimated values ​​with absolute values ​​greater than 0.5.

[0051] In some embodiments, the reference signal is selected in an automatic selection mode or a manual specification mode. The automatic selection mode calculates the average correlation coefficient of each signal in each layer and selects the signal with the largest average correlation coefficient as the reference signal; the joint threshold adjustment takes into account the correlation between signals and the difference from the single pendulum wave signal.

[0052] In some embodiments, the dual-tree complex wavelet inverse transform adopts a dual-tree structure corresponding to the decomposition, and upsamples and filters the denoised high-frequency complex detail coefficients of each layer and the unprocessed low-frequency complex coefficients of each layer layer by layer; the consistency check uses the Bernoulli equation to check the pressure signal and flow signal, and the error is required to be less than 5%.

[0053] In some embodiments, the signal-to-noise ratio improvement requirement for indicator evaluation is greater than 10dB, the mean square error is less than a set threshold, and the fault feature retention rate requirement for feature evaluation is greater than 70%.

[0054] In some embodiments, time synchronization optimization uses the period of the single pendulum wave reference signal as a reference to perform interpolation calibration on sensor signals whose time deviation exceeds a set threshold; frequency comparison calibration adjusts the bandpass filter parameters of each signal based on the theoretical frequency of the single pendulum wave reference signal.

[0055] In some embodiments, the high-frequency complex detail coefficients obtained by parallel processing of the dual-tree complex wavelet transform using two sets of filters are soft-thresholded on their real and imaginary parts, and then synthesized to obtain the denoised high-frequency complex detail coefficients.

[0056] In some embodiments, multi-source signal acquisition and preprocessing:

[0057] In the actual operation environment of PE pipelines, in order to fully obtain information reflecting the pipeline operation status, it is necessary to collect signals from multiple dimensions and perform preliminary preprocessing operations. The specific process is as follows:

[0058] Signal acquisition:

[0059] Pressure signal: The pressure of the fluid in the pipeline is collected through a piezoresistive pressure sensor. The piezoresistive pressure sensor works based on the piezoresistive effect. When the pressure of the fluid in the pipeline acts on the sensor, the resistance of the piezoresistor inside the sensor changes, thereby converting the physical quantity of pressure into a voltage signal. The conversion formula is: ,in, is the pressure signal at time t; It is the pressure-voltage conversion coefficient determined by the characteristics of the sensor itself; is the voltage value output by the sensor at time t.

[0060] When installing sensors, key nodes of the pipeline are usually selected, such as elbows and near valves, because pressure changes at these locations can more sensitively reflect the operating status of the pipeline.

[0061] Flow signal: electromagnetic flowmeter or ultrasonic flowmeter is used for collection. Electromagnetic flowmeter is based on Faraday's law of electromagnetic induction. When the conductive liquid cuts the magnetic flux lines in the magnetic field, it will generate induced electromotive force on both sides of the pipe. The flow rate is calculated by measuring the induced electromotive force. The conversion formula is: .in, is the flow signal at time t; is the instrument factor of the electromagnetic flowmeter; is the magnetic induction intensity of the magnetic field inside the flowmeter; is the inner diameter of the pipe; is the average flow velocity of the fluid at time t.

[0062] Ultrasonic flowmeters use the time difference between ultrasonic waves propagating downstream and upstream in the fluid to calculate the flow rate. The conversion formula is: ,in, is the calibration coefficient of the ultrasonic flowmeter; L is the length of the sound channel; is the downstream propagation time; is the countercurrent propagation time.

[0063] In actual applications, the appropriate flow meter type will be selected based on factors such as the pipe material, pipe diameter, and fluid properties, and its installation location will be ensured to meet the requirements to ensure the accuracy of flow signal acquisition.

[0064] Temperature signal: Use thermocouples or thermal resistor temperature sensors to collect pipe temperature. Thermocouples are based on the Seebeck effect. Two conductors of different materials form a closed loop. When there is a temperature difference between the two ends, a thermoelectric potential is generated in the loop. The conversion formula is: ,in For the moment Temperature signal; is the thermoelectric potential coefficient of the thermocouple; Thermocouple at time Thermoelectric potential difference between the two ends.

[0065] Thermal resistors reflect temperature changes by measuring resistance values ​​based on the characteristic that resistance changes with temperature. The conversion formula is: ,in, The thermal resistance at temperature The resistance value when The thermal resistor at the reference temperature The resistance value when is the temperature coefficient of the thermal resistor.

[0066] When installing the temperature sensor, corresponding protective measures will be taken to prevent the sensor from being interfered with by external environmental factors and ensure the reliability of the temperature signal.

[0067] Vibration signal: The acceleration sensor is fixed to the pipeline support structure to collect the pipeline vibration acceleration signal. The acceleration sensor can convert the acceleration physical quantity generated by pipeline vibration into an electrical signal. The selection of its installation location is crucial. Generally, it is selected at the pipeline support point, hanger and other locations to accurately capture the vibration of the pipeline during operation. The acceleration physical quantity is converted into an electrical signal. The conversion formula is: ,in, For the moment Vibration signal; is the sensitivity coefficient of the acceleration sensor; For the pipeline at the moment acceleration.

[0068] Acoustic emission signal: An acoustic emission sensor is attached to the outer wall of the pipeline to capture the acoustic emission waves generated by changes in the internal structure of the pipeline. When a fault such as crack expansion or leakage occurs inside the pipeline, elastic waves, namely acoustic emission waves, are generated. The acoustic emission sensor can convert the mechanical energy of the acoustic emission waves into electrical signals. In order to enhance the acoustic coupling between the sensor and the pipeline and improve the signal acquisition effect, a coupling agent is applied to the pipeline surface during installation. The mechanical energy of the acoustic emission wave is converted into an electrical signal. The conversion formula is: ,in, For the moment Acoustic emission signals; is the sensitivity of the acoustic emission sensor; The acoustic emission wave at time sound pressure.

[0069] Strain signal: Strain gauges are attached to the surface of the pipe to collect pipe strain based on the metal resistance strain effect. When the pipe deforms, the strain gauge attached to the pipe surface will also deform, causing its resistance to change. The resistance change is then converted into a voltage signal through a Wheatstone bridge. The conversion formula is: ,in, For the moment The strain signal; is the sensitivity coefficient of the strain gauge; For the strain gauge at time The resistance change; is the initial resistance of the strain gauge.

[0070] When pasting strain gauges, it is necessary to strictly follow the process requirements to ensure that the strain gauges fit tightly against the pipe surface and to ensure the accuracy of strain signal acquisition.

[0071] In some embodiments, the initial pre-processing:

[0072] Filtering: Bandpass filtering is performed on the raw signals collected by each sensor. Bandpass filters allow signals within a specific frequency range to pass through while suppressing signals at other frequencies, effectively removing significant baseline drift and high-frequency noise. When setting the parameters of the bandpass filter, adjustments are made based on the frequency characteristics of different signal types. For example, the normal frequency range of pressure signals is typically low, while the frequency of acoustic emission signals is relatively high. By properly setting the filter's cutoff frequency, it is possible to minimize noise interference while retaining the useful signal.

[0073] Time synchronization: Because different sensors may have different sampling times, an interpolation algorithm is used to time-align sensor signals to ensure consistency across multiple signal sources. This algorithm selects a suitable time reference point and then interpolates the signal values ​​at the missing time points based on the time deviations between each signal and the reference point. This ensures precise alignment of all signals on the time axis, paving the way for subsequent signal processing and analysis.

[0074] In some embodiments, the signals are co-calibrated:

[0075] In order to further improve the quality and accuracy of multi-source signals, it is necessary to use a single pendulum wave signal to perform collaborative calibration on the multi-source signals. The specific steps are as follows:

[0076] Single pendulum wave signal processing:

[0077] Install a high-precision pendulum device: A high-precision pendulum device is installed vertically near the pipeline. The pendulum length is precisely measured, adjusted, and kept fixed. A high-precision displacement sensor is installed at the bottom of the pendulum to collect the pendulum's swing signal. The pendulum's swing follows the law of simple harmonic motion, and its displacement changes with time with a highly stable periodicity, which can serve as a reliable time and frequency reference. The formula for its displacement change with time is: ,in, is a simple pendulum wave at time The displacement signal; A is the amplitude; is the angular frequency of the simple pendulum wave, is the acceleration due to gravity, is the pendulum length; The swing signal of the simple pendulum is collected by a high-precision displacement sensor, and the signal has a highly stable periodicity.

[0078] Signal calibration: Frequency calibration and time stamping are performed on the collected pendulum wave signal. Frequency calibration is achieved by accurately measuring the pendulum's length, calculating the theoretical frequency of the pendulum according to the formula for simple harmonic motion, and comparing and adjusting it with the actual collected signal frequency to ensure the accuracy of the pendulum wave signal frequency. Time stamping assigns a precise timestamp to each sampling point of the pendulum wave signal, making it an accurate time reference signal.

[0079] like Figure 3 As shown, the details are as follows:

[0080] Hardware deployment:

[0081] Adopt high-precision single pendulum device (pendulum length =1m, acceleration due to gravity =9.8m / s²), a laser displacement sensor (sampling frequency 10kHz, resolution 1μm) is installed at the bottom to ensure the swing displacement signal The pendulum is installed on an independent bracket within 5 meters of the pipeline to avoid direct interference from pipeline vibration. The bracket foundation depth is ≥ 2 meters to isolate ground noise.

[0082] Frequency calibration:

[0083] Theoretical frequency calculation:

[0084] According to the simple pendulum period formula , calculate the theoretical period T≈2.007s, corresponding to the frequency .

[0085] Measured frequency correction: Perform FFT spectrum analysis on the collected original signal to extract the main peak frequency , if the deviation exceeds 0.1% (e.g. =0.499Hz), then by adjusting the pendulum length Perform physical calibration to ensure frequency accuracy ≤ 0.01Hz.

[0086] A GPS clock module (1μs accuracy) is used to timestamp the single pendulum wave signal. Each sampling point is labeled with a UTC time stamp to establish an absolute time reference with multiple source signals. Sampling points with a time stamp error exceeding 10μs are corrected using linear interpolation to ensure a time synchronization error of ≤5μs.

[0087] Noise analysis process:

[0088] Noise Source Identification:

[0089] Environmental vibration and noise: Spectrum analysis can be used to identify periodic noise such as 50Hz power frequency interference and 100Hz harmonics of nearby mechanical vibration.

[0090] Sensor noise: Use the white noise test method to calculate the standard deviation of the sensor background noise = 0.5 μm, as the noise analysis threshold.

[0091] Frequency domain noise reduction processing:

[0092] Bandpass filtering: The FIR filter is designed to retain the 0.4-0.6Hz frequency band (single pendulum wave theoretical frequency ±10%), suppress high-frequency environmental noise and low-frequency drift, and the filter stopband attenuation is ≥40dB.

[0093] Wavelet denoising: The filtered signal is decomposed by 3-layer db4 wavelet, and the high-frequency coefficients are processed by soft thresholding (threshold λ=3 ), the noise level is reduced by more than 60% after reconstruction.

[0094] Time domain stability assessment:

[0095] Calculate the periodic fluctuation coefficient of the adjacent 100 periods ,Require ≤0.05%, otherwise recalibrate the simple pendulum mechanical structure. is the standard deviation of the periodic measured values, is the mean of the periodic measured values.

[0096] Generate a stable periodic single pendulum wave reference signal , its signal-to-noise ratio SNR≥40dB, meeting the requirements of multi-source signal calibration.

[0097] like Figure 4 As shown, multi-source signal synchronization optimization:

[0098] Time synchronization optimization: Using the period of the single pendulum wave reference signal as a benchmark, time synchronization optimization is performed on the multi-source signals after preliminary preprocessing. The specific method is to calculate the time deviation between each sensor signal and the single pendulum wave reference signal. For sensor signals whose time deviation exceeds a set threshold (such as 5% of the single pendulum wave period), a high-precision interpolation algorithm is used for calibration. For example, if the time difference between the pressure signal and the single pendulum wave reference signal gradually increases within a certain period of time, by analyzing the periodic variation pattern of the single pendulum wave reference signal, the exact time point that each sampling point in the pressure signal should correspond to is determined. Then, using algorithms such as cubic spline interpolation, the pressure signal is locally resampled to better align the time series of the pressure signal with the single pendulum wave signal, thereby achieving precise time synchronization of the multi-source signals. The same principle applies to other signals.

[0099] The period of the single pendulum reference signal As a reference, if the time deviation from the single pendulum wave reference signal exceeds the set threshold (such as ) sensor signals and use interpolation algorithms for time calibration.

[0100] Frequency comparison calibration: The theoretical frequency of the single pendulum wave reference signal Compare and analyze the frequency of each sensor signal. Obtain the spectrum distribution of each sensor signal through spectrum analysis method. If a component with a frequency close to that of a single pendulum wave and abnormal intensity is found in a sensor signal, it is judged that the component may be interference noise such as environmental vibration. In this case, according to the range of interference frequency, adjust the parameters of the corresponding sensor signal bandpass filter, such as low-frequency cutoff frequency and high-frequency cutoff frequency, to effectively suppress such noise, realize frequency calibration of multi-source signals, and improve signal purity. Calculate the multi-source signal and The cross power spectrum is used to notch filter the frequency components with correlation coefficient < 0.3 (such as 50Hz interference) to ensure that the proportion of components related to the simple pendulum wave frequency in the signal is ≥ 85%.

[0101] In some embodiments, the dual-tree complex wavelet decomposition:

[0102] Dual-tree complex wavelet transform is an advanced signal processing method that can perform multi-layer decomposition of signals and extract the characteristics of signals at different frequency levels. The specific decomposition process is as follows:

[0103] Multi-layer decomposition:

[0104] The calibrated multi-source signal and the single pendulum reference signal are each subjected to a dual-tree complex wavelet transform. The dual-tree complex wavelet transform utilizes two parallel filter banks, namely, tree A and tree B. This unique structural design provides excellent directional selectivity and anti-aliasing capabilities, enabling more accurate capture of signal characteristics. During the transform, the number of decomposition layers is typically set to 3-5, depending on the signal length and noise characteristics. Too few decomposition layers may not adequately separate the noise and useful signal; too many decomposition layers increases computational complexity and may introduce additional errors.

[0105] The filter bank design of tree A and tree B meets the conditions for analyzing the signal. By carefully designing the coefficients and structure of the filters, it is ensured that the phase information and detailed characteristics of the signal can be effectively retained during the decomposition process, avoiding frequency aliasing, thereby improving the quality and accuracy of signal decomposition.

[0106] Decomposition process calculation formula:

[0107] No. Layer tree A filter calculation:

[0108] Low-pass filtering calculates the low-frequency coefficients: .

[0109] High-pass filtering calculates the high-frequency detail coefficient: .

[0110] No. Layer tree B filter calculation:

[0111] Low-pass filtering calculates the low-frequency coefficients: , For time delay.

[0112] High-pass filtering calculates the high-frequency detail coefficient: .

[0113] Coefficient combination calculation:

[0114] Low-frequency complex coefficients: ;

[0115] High-frequency complex detail coefficients: .

[0116] Decomposition process: Determine the number of decomposition layers L, decompose each signal layer by layer according to the above formula, and obtain the low-frequency complex coefficients of each layer and high-frequency complex detail coefficients .

[0117] : Input signal to be decomposed (calibrated multi-source signal and single pendulum wave reference signal), , is the signal length. : low-pass filter coefficients of tree A, , is the filter length. : High-pass filter coefficients of tree A. : low-pass filter coefficients of tree B, , is the filter length. : High-pass filter coefficients of tree B. : No. The low-frequency coefficients generated by tree A after layer decomposition ( Indicates the signal category), , , L is the preset number of decomposition levels. : No. High-frequency detail coefficients generated by tree A after layer decomposition. : No. The low-frequency coefficients generated by tree B after layer decomposition. : No. High-frequency detail coefficients generated by tree B after layer decomposition.

[0118] Coefficient extraction:

[0119] Each decomposition layer generates low-frequency coefficients and high-frequency detail coefficients. The low-frequency coefficients reflect the approximate characteristics of the signal, including the main trends and low-frequency components of the signal. The high-frequency detail coefficients contain detailed characteristics of the signal, such as mutations and edges, and reflect the changes in the signal in the high-frequency band.

[0120] For multi-source signals and single pendulum wave reference signals, the corresponding low-frequency coefficients and high-frequency detail coefficients are retained after each layer of decomposition. These coefficients will serve as the basic data for subsequent processing. Through further analysis and processing, noise reduction and feature extraction of the signal can be achieved.

[0121] In some embodiments, the noise level estimate:

[0122] Accurately estimating the noise level in the signal is a key step in achieving effective noise reduction. This solution uses multiple methods to comprehensively estimate the noise level, as follows:

[0123] Noise standard deviation estimate:

[0124] The median estimation method is used to calculate the noise standard deviation of the high-frequency complex detail coefficients of each layer of the multi-source signal. The median estimation method is insensitive to outliers and can robustly estimate the noise intensity in the presence of burst noise. Its calculation formula is: Noise standard deviation = median of the absolute values ​​of the high-frequency complex detail coefficients / 0.6745. In the specific operation, the absolute values ​​of the high-frequency complex detail coefficients of each layer are first taken, and then the median of these absolute values ​​is calculated. Finally, the median is divided by 0.6745 to obtain the noise standard deviation estimate of the high-frequency complex detail coefficients of this layer. For example, for the high-frequency complex detail coefficients of the second layer of the pressure signal, the above calculation process can accurately estimate the standard deviation of the noise in the signal of this layer, providing an important basis for the subsequent threshold calculation.

[0125] Single signal noise standard deviation estimation: The median estimation method is used to calculate the noise standard deviation of the high-frequency complex detail coefficients of each layer:

[0126] ;

[0127] in, For the Tier The noise standard deviation of the high-frequency complex detail coefficients of the similar signal; For the Tier High-frequency complex detail coefficients of the signal-like object; is a standard normal distribution.

[0128] Difference analysis:

[0129] Calculate the difference index between the high-frequency complex detail coefficients of each layer of each signal and the high-frequency complex detail coefficients of the corresponding layer of the simple pendulum wave. The calculation formula for the difference index is: Difference index = sum of absolute differences / sum of high-frequency complex detail coefficients of the single pendulum wave. By calculating this index, the similarity between the high-frequency complex detail coefficients of each signal and the high-frequency complex detail coefficients of the single pendulum wave can be measured. If the difference index of a signal exceeds the set threshold (such as 0.3), it means that the high-frequency complex detail coefficients of this signal in this layer are significantly affected by noise and require special attention in subsequent noise reduction processing. For example, when there is a mechanical vibration source near the pipeline, the high-frequency complex detail coefficients of the vibration signal may be interfered with. By comparing the difference index with the high-frequency complex detail coefficients of the single pendulum wave, this interference can be detected in a timely manner and appropriate noise reduction measures can be taken.

[0130] Multi-signal joint noise estimation: Calculate the difference index between the high-frequency complex detail coefficients of each signal and the high-frequency complex detail coefficients of the single pendulum wave:

[0131] ;

[0132] in, For the Tier The difference index between the high-frequency complex detail coefficient of the similar signal and the high-frequency complex detail coefficient of the simple pendulum wave; The reference signal of the single pendulum wave is The high-frequency complex detail coefficients of the layer.

[0133] Correlation analysis:

[0134] Calculate the correlation coefficient matrix between signals and retain valid correlation coefficients with an absolute value greater than 0.5. The correlation coefficient reflects the degree of linear correlation between signals. By calculating the correlation coefficient matrix, we can fully understand the interrelationships between multi-source signals. Retaining correlation coefficients with an absolute value greater than 0.5 means that only signal pairs with strong correlations are considered, while signal relationships with weak or no correlation are eliminated, thus screening signal combinations that are practical for noise estimation. For example, in some cases, vibration signals and acoustic emission signals may both be affected by the same mechanical vibration noise source. Through correlation analysis, the noise components of these two signals can be jointly estimated, improving the accuracy and reliability of noise estimation.

[0135] Correlation analysis between signals: Calculate the correlation coefficient between signals ,in . For the Tier Class signal and Correlation coefficient of high-frequency complex detail coefficient of class signal; is the covariance; 、 are the means of the corresponding high-frequency complex detail coefficients; For the The length of the high-frequency complex detail coefficients of the layer. estimated value.

[0136] In some embodiments, adaptive threshold noise reduction:

[0137] Based on the noise level estimation results, the adaptive threshold noise reduction method is used to process the signal to effectively remove noise and retain useful signal features. The specific process is as follows:

[0138] Reference signal selection:

[0139] Automatic selection mode: Calculate the average correlation coefficient of each signal in each layer and select the signal with the largest average correlation coefficient as the reference signal. The specific calculation is to first calculate the correlation coefficient between every two signals in each layer, then average the correlation coefficients of each signal with all other signals. The signal with the largest average correlation coefficient is the reference signal for that layer. The purpose of selecting a reference signal is to use this signal as a benchmark during the threshold calculation and adjustment process, comprehensively consider the correlation between other signals and it, and achieve collaborative noise reduction for multi-source signals. During the selection process, signals with excessive correlation with the single pendulum wave reference signal will be excluded to avoid mistakenly selecting signals affected by single pendulum wave interference as reference signals, thereby ensuring the reliability and effectiveness of the reference signal.

[0140] Dynamic reference signal selection mechanism:

[0141] Noise frequency band analysis: Analyze the spectrum of various signals and single pendulum wave signals to determine the main noise frequency bands.

[0142] Correlation matrix calculation: Calculate the correlation coefficients of all signal pairs at each layer Matrix . =6, which means the total number of types of non-single pendulum wave reference signals (the total number of multi-source signals). That is the single pendulum wave reference signal.

[0143] Automatic selection mode: calculate the average correlation coefficient of each signal in each layer, and select the signal with the largest average correlation coefficient as the reference signal .

[0144] Manual Designation Mode: Operators can manually designate a reference signal based on the actual noise type and signal characteristics. For example, if a pipeline is known to have a potential leak, the AE signal can be manually designated as the reference signal, as it is sensitive to leak noise. This allows for more targeted processing of leak-related noise, improving noise reduction and fault diagnosis accuracy.

[0145] Threshold calculation and adjustment:

[0146] Base Threshold: Calculates the base threshold based on the noise standard deviation and signal length. This calculation takes into account both the noise intensity and signal length, providing a reasonable initial reference for threshold processing. The specific calculation formula is selected and adjusted based on actual conditions to ensure that the base threshold accurately reflects the approximate noise level in the signal.

[0147] Basic threshold calculation: , For the Tier Basic threshold of class signal; For the Tier The length of the signal-like coefficients.

[0148] Joint Adjustment: Threshold parameters are adjusted based on the correlation between signals and their differences from a single pendulum reference signal. After calculating the base threshold, the threshold is adjusted jointly based on the correlation coefficient between each signal and the reference signal, as well as the difference between each signal and the single pendulum reference signal. For signals with high correlation with the reference signal, the threshold is appropriately raised to avoid excessive noise reduction and loss of useful signals. For signals with large differences from the single pendulum signal, the threshold is also adjusted based on the specific situation to ensure that key signal characteristics are preserved while removing noise. This joint adjustment method allows the threshold to be more adaptable to the characteristics of different signals and noise conditions, achieving adaptive threshold processing.

[0149] Joint threshold adjustment: , For the Tier Class signal adjusted threshold; is the weight coefficient; is the correction coefficient related to the single pendulum wave.

[0150] Thresholding:

[0151] Soft thresholding is performed on the real and imaginary parts of the high-frequency complex detail coefficients. Soft thresholding is a commonly used signal denoising method that can remove noise while minimizing the loss of signal features. Specifically, for each high-frequency complex detail coefficient, the real and imaginary parts are evaluated separately. If the absolute value of the real or imaginary part is greater than the adjusted threshold, the threshold is subtracted from it and the result is multiplied by the sign function (when the value is greater than 0, the sign function value is 1; when the value is less than 0, the sign function value is -1; when the value is 0, the sign function value is 0). If the absolute value of the real or imaginary part is less than or equal to the threshold, it is set to zero. In this way, the high-frequency complex detail coefficients are processed to obtain the denoised high-frequency complex detail coefficients, effectively removing the noise component in the signal while retaining useful signal features.

[0152] Threshold processing uses a soft threshold function to process high-frequency complex detail coefficients:

[0153] ;

[0154] in, is a symbolic function, when hour, ;when hour, ;when hour, For high-frequency complex detail coefficients , perform soft threshold processing on its real part a and imaginary part b respectively. The specific process is as follows:

[0155] Compute the thresholding result of the real part:

[0156] ;

[0157] Calculate the thresholding result of the imaginary part:

[0158] ;

[0159] Get the high-frequency complex detail coefficient after noise reduction:

[0160] .

[0161] A high-frequency complex detail coefficient of the third layer of the vibration signal For example, assuming that the calculated adjusted threshold For the real part a=3, since |3|>2, according to the soft threshold function, For the imaginary part b=4, since |4|>2, The high-frequency complex detail coefficient after noise reduction is In the actual processing process, all high-frequency complex detail coefficients of each type of signal (pressure signal, flow signal, temperature signal, vibration signal, acoustic emission signal, strain signal) in each layer will be thresholded in the above way. In this way, while effectively suppressing noise, the useful features in the signal are retained as much as possible, and the denoising operation of the high-frequency complex detail coefficients of multi-source signals is completed, providing high-quality data for subsequent signal reconstruction. After threshold processing, the denoised high-frequency complex detail coefficients are obtained. , these high-frequency complex detail coefficients will be compared with the low-frequency complex coefficients that have not been thresholded. and Together, they are used in the subsequent dual-tree complex wavelet inverse transform to achieve signal reconstruction.

[0162] If there is identifiable low-frequency noise in the low-frequency complex coefficients (e.g., noise peaks detected through spectral analysis), or if the number of decomposition levels is high (e.g., greater than or equal to 3), and the deep low-frequency coefficients contain some noise components, a very low threshold is used, using either a soft or hard threshold function, while retaining the main coefficients. After denoising, the energy loss rate of the low-frequency complex coefficients must be less than 5% to avoid signal distortion. Accordingly, the denoised low-frequency complex coefficients are used in subsequent signal reconstruction.

[0163] In some embodiments, signal reconstruction:

[0164] After threshold processing, it is necessary to perform a dual-tree complex wavelet inverse transform on the noise-reduced coefficients, reconstruct the signal from the frequency domain back to the time domain, and perform a consistency check. The specific steps are as follows:

[0165] Inverse transform reconstruction:

[0166] The inverse dual-tree complex wavelet transform (ITT) is used to reconstruct the denoised high-frequency complex detail coefficients and the unprocessed low-frequency complex coefficients layer by layer. The IDT is the inverse process of the IDT. It restores the processed frequency-domain coefficients of the signal to the time-domain signal through operations corresponding to the decomposition process. During the layer-by-layer reconstruction process, each layer includes upsampling and filtering operations. Upsampling restores the length of the coefficients to the size of the previous layer to compensate for the information lost during the decomposition process. Filtering processes the coefficients through specific filters to restore the signal's time-domain characteristics, gradually restoring the signal to its original time series form. Through layer-by-layer inverse transform reconstruction, a preliminary reconstructed signal is ultimately obtained.

[0167] The specific steps are as follows:

[0168] Layer-by-layer reconstruction calculation: Starting from the highest decomposition layer L, gradually reconstruct towards the original signal layer. Layer to Take the reconstruction of the layer as an example:

[0169] Tree A signal reconstruction:

[0170] ;

[0171] in, It is an upsampling operation used to restore the length of the coefficient to the scale of the previous layer; For the The result of denoising the high-frequency detail coefficients of layer tree A. The upsampled coefficients are convolved with the corresponding filters to restore the low-frequency and high-frequency components of the signal.

[0172] Tree B signal reconstruction:

[0173] ;

[0174] in, For the The result of noise reduction on the high-frequency detail coefficients of layer tree B. Synthesize the reconstructed signal: Synthesize the signals reconstructed from tree A and tree B:

[0175] ;

[0176] in, and After layer-by-layer reconstruction, the real and imaginary signals of tree A and tree B are respectively obtained on the original signal scale (0th layer). Class Signal .

[0177] Consistency check:

[0178] The pressure signal and flow signal are verified using physical relationships such as the Bernoulli equation, with an error requirement of less than 5%. The Bernoulli equation describes the relationship between pressure, velocity, and height when a fluid flows in a pipe and is an important equation in fluid mechanics. When verifying the pressure signal and flow signal, according to the principle of the Bernoulli equation, the reconstructed pressure signal and flow signal are substituted into the equation for calculation, and the difference between the calculated results and the actual physical situation is compared. If the reconstructed signal does not meet the condition of an error of less than 5%, the reconstruction result is considered unreasonable, and there may be problems such as excessive noise reduction or improper parameter setting. In this case, it is necessary to return to S5 to adjust the threshold parameters, re-perform noise reduction processing and signal reconstruction until the reconstructed signal passes the consistency check to ensure that the reconstructed signal conforms to physical laws and can truly reflect the pipeline operation status.

[0179] Using the Bernoulli equation (in, 、 is the pressure value at different positions; 、 is the flow velocity at the corresponding position; is the fluid density; is the head loss), the reconstructed pressure signal and flow signals If the reconstructed signal does not meet the condition that the error of the equation is less than 5%, the reconstruction result is considered unreasonable and the threshold needs to be readjusted. Return to S5 to re-perform threshold processing and subsequent signal reconstruction operations to ensure that the reconstructed signal conforms to physical laws and can truly reflect the pipeline operation status.

[0180] In some embodiments, effect evaluation and optimization:

[0181] To ensure the effectiveness and reliability of the noise reduction method, it is necessary to comprehensively evaluate the final noise reduction signal and optimize the noise reduction process based on the evaluation results. The specific contents are as follows:

[0182] Objective indicator evaluation:

[0183] Signal-to-noise ratio (SNR): An improvement of greater than 10dB is required. The SNR is a key indicator of signal quality, representing the ratio of useful signal power to noise power. By calculating the SNR of the signal before and after noise reduction and comparing the magnitude of the improvement, you can visually assess the extent to which the noise reduction method has improved signal quality. If the SNR improves by greater than 10dB, it indicates that the noise reduction method has effectively suppressed noise, increased the proportion of useful signal in the signal, and significantly improved signal quality.

[0184] Mean Squared Error (MSE): This value must be less than a set threshold. The MSE measures the average difference between the denoised signal and the ideal clean signal. A smaller value indicates a closer approximation to the ideal clean signal, and a better noise reduction effect. In actual evaluations, a reasonable MSE threshold is set based on specific application requirements and signal characteristics. If the MSE of the denoised signal is less than this threshold, the noise reduction is considered satisfactory.

[0185] Peak Signal-to-Noise Ratio (PSNR): Evaluates signal peak error. While PSNR is often used to assess peak signal error, it's relatively rare in PE pipeline multi-source signal processing. It can provide another perspective on large signal errors, helping to assess the effectiveness of noise reduction algorithms on signal peaks and comprehensively evaluating the performance of noise reduction methods.

[0186] Feature Assessment:

[0187] Extract pipeline fault features (such as high-frequency leak features and low-frequency blockage features), with a retention rate exceeding 70%. Pipeline fault features are crucial for determining the presence and type of pipeline faults. Pipeline leaks typically produce high-frequency signatures in acoustic emission (AE) and vibration signals, such as a sudden increase in energy in the 100-500 Hz frequency band of AE signals. Pipeline blockages, on the other hand, exhibit specific variations in low-frequency bands like pressure and flow signals, such as abnormal fluctuations in the 0.5-5 Hz frequency band of the pressure signal. By analyzing the retention of these fault features in the denoised signal, it is possible to intuitively determine whether the denoising method has adversely affected fault diagnosis. If the fault feature retention rate exceeds 70%, it indicates that the denoising process has effectively removed noise while retaining key diagnostic information, ensuring the ability to identify pipeline faults.

[0188] Feedback optimization:

[0189] If an evaluation metric fails to meet expectations (e.g., an SNR improvement of less than 10dB, an MSE greater than a set threshold, or a fault signature retention rate below 70%), the feedback optimization mechanism automatically activates. First, the noise reduction parameters are adjusted specifically based on the specific metric that failed to meet the standard.

[0190] Threshold parameter adjustment: If the fault feature retention rate of a certain type of signal is low, it means that the useful signal may be overly suppressed during the threshold processing. In this case, the single pendulum wave correction coefficient corresponding to the signal will be reduced. , reduce the threshold adjustment amplitude caused by the correlation with the single pendulum wave, making the threshold processing more conservative; or reduce the threshold adjustment weight coefficient , reducing the impact of signal correlation on the threshold and avoiding excessive noise reduction. After adjusting the parameters, it will return to S5 to re-threshold processing, and complete subsequent signal reconstruction, effect evaluation and other operations in sequence until a satisfactory noise reduction effect is achieved.

[0191] Adjustment of the number of decomposition layers: If the overall noise reduction effect is not good, for example, the SNR improvement is not obvious and the MSE is large, it may be due to an unreasonable setting of the decomposition layer number of the dual-tree complex wavelet transform. If the number of decomposition layers is too small, the noise and useful signals cannot be fully separated; if the number of decomposition layers is too large, additional calculation errors may be introduced or the useful signal may be over-decomposed. At this time, an attempt will be made to adjust the number of decomposition layers L, such as increasing it from 3 layers to 4 layers, or reducing it from 4 layers to 3 layers, and then returning to S3 to re-perform the dual-tree complex wavelet transform decomposition. After that, a series of operations such as noise estimation, threshold processing, signal reconstruction and effect evaluation are carried out in sequence according to the process of S4-S7, continuously optimizing the noise reduction process until all evaluation indicators meet the expected standards, ensuring that the PE pipeline multi-source signal noise reduction processing based on this method can effectively suppress noise and accurately retain signal characteristics under various complex working conditions, providing reliable protection for the safe operation monitoring of PE pipelines.

[0192] In summary, multiple specific embodiments of the present invention are disclosed. Under the condition that there is no self-contradiction, the various embodiments can be freely combined to form new embodiments. That is, the embodiments belonging to the replacement schemes can be freely replaced but cannot be combined with each other; the embodiments that do not belong to the replacement schemes can be combined with each other, and these new embodiments also belong to the essential content of the present invention.

[0193] The above embodiments describe multiple specific implementations of the present invention, but those skilled in the art should understand that various changes or modifications can be made to these implementations without departing from the principles and essence of the present invention, but these changes and modifications are all within the scope of protection of the present invention.

Claims

1. A PE pipeline multi-source signal denoising method based on dual-tree complex wavelet transform is characterized by: include: S1. Multi-source signal acquisition and preprocessing: Collect pressure signals, flow signals, temperature signals, vibration signals, acoustic emission signals, and strain signals from PE pipelines, perform preliminary filtering and time synchronization processing to obtain pre-processed multi-source signals; collect single pendulum wave signals, perform frequency calibration and time stamping to obtain calibrated single pendulum wave signals; S2. Signal collaborative calibration; noise analysis of the calibrated single pendulum wave signal to obtain a single pendulum wave reference signal; time synchronization optimization of the pre-processed multi-source signal based on the single pendulum wave reference signal to obtain a time-synchronized multi-source signal; frequency comparison calibration of the time-synchronized multi-source signal and the single pendulum wave reference signal to obtain a calibrated multi-source signal; S3. Dual-tree complex wavelet decomposition: Perform dual-tree complex wavelet transform multi-layer decomposition on the calibrated multi-source signal and the single pendulum wave reference signal, respectively, to obtain low-frequency complex coefficients and high-frequency complex detail coefficients of each layer of the multi-source signal, and low-frequency complex coefficients and high-frequency complex detail coefficients of each layer of the single pendulum wave; S4. Noise level estimation: Estimating the noise standard deviation of the high-frequency complex detail coefficients of each layer of the multi-source signal, calculating the difference index between the high-frequency complex detail coefficients of each layer of the multi-source signal and the high-frequency complex detail coefficients of each layer of the single pendulum wave, performing correlation analysis between the signals, and obtaining the noise level estimation result; S5. Adaptive threshold noise reduction; Based on the noise level estimation result, a reference signal is selected, the basic threshold is calculated and a joint threshold adjustment is performed, and a soft threshold function is used to perform threshold processing on the high-frequency complex detail coefficients of each layer of the multi-source signal to obtain the high-frequency complex detail coefficients of each layer after noise reduction; S6. Signal reconstruction: Perform a dual-tree complex wavelet inverse transform on the high-frequency complex detail coefficients of each layer after denoising and the low-frequency complex coefficients of each layer of the multi-source signal, reconstruct the signal layer by layer, and obtain a preliminary reconstructed signal; perform a physical relationship-based consistency check on the preliminary reconstructed signal. If the check fails, return to S5 to adjust the threshold. If the check passes, obtain the final denoised signal. S7. Effect evaluation and optimization: Evaluate the final noise reduction signal using the signal-to-noise ratio, mean square error, and peak signal-to-noise ratio indicators. If the evaluation indicators do not meet expectations, adjust the threshold parameters or the number of decomposition layers and return to S3 for reprocessing until the evaluation indicators meet expectations.

2. The PE pipeline multi-source signal denoising method based on dual-tree complex wavelet transform according to claim 1 is characterized in that: The pressure signal is collected through a piezoresistive pressure sensor, the flow signal is collected through an electromagnetic flowmeter or an ultrasonic flowmeter, the temperature signal is collected through a thermocouple or a thermistor temperature sensor, the vibration signal is collected through an acceleration sensor, the acoustic emission signal is collected through an acoustic emission sensor, and the strain signal is collected through a strain gauge; a bandpass filter is used for preliminary filtering, and an interpolation algorithm is used for time synchronization.

3. The PE pipeline multi-source signal denoising method based on dual-tree complex wavelet transform according to claim 1 is characterized in that: A simple pendulum device is installed next to the pipeline. The swing of the simple pendulum follows the law of simple harmonic motion, and the swing signal of the simple pendulum is collected by a displacement sensor.

4. The PE pipeline multi-source signal denoising method based on dual-tree complex wavelet transform according to claim 1 is characterized in that: The dual-tree complex wavelet transform uses two parallel filter banks. The filter banks of tree A and tree B are designed to meet the analytical signal conditions. The number of decomposition layers is determined to be 3 to 5 according to the signal length and noise characteristics.

5. The PE pipeline multi-source signal denoising method based on dual-tree complex wavelet transform according to claim 1 is characterized in that: The noise standard deviation is estimated using the median estimation method. The difference index is calculated as the ratio of the sum of the absolute differences between the high-frequency complex detail coefficients of each layer of each signal and the high-frequency complex detail coefficients of each layer of the simple pendulum wave to the sum of the high-frequency complex detail coefficients of each layer of the simple pendulum wave. Correlation analysis calculates the correlation coefficient between signals and retains the estimated values ​​with absolute values ​​greater than 0.

5.

6. The method for denoising multi-source signals in PE pipelines based on dual-tree complex wavelet transform according to claim 1, characterized in that: The reference signal is selected in automatic selection mode or manual specification mode. The automatic selection mode calculates the average correlation coefficient of each signal in each layer and selects the signal with the largest average correlation coefficient as the reference signal; the joint threshold adjustment takes into account the correlation between signals and the difference from the single pendulum wave signal.

7. The method for denoising multi-source signals in PE pipelines based on dual-tree complex wavelet transform according to claim 1, characterized in that: The dual-tree complex wavelet inverse transform adopts the dual-tree structure corresponding to the decomposition, and reconstructs the high-frequency complex detail coefficients of each layer after denoising and the unprocessed low-frequency complex coefficients of each layer by upsampling and filtering. The consistency check uses the Bernoulli equation to check the pressure signal and flow signal, and the error is required to be less than 5%.

8. The method for denoising multi-source signals in PE pipelines based on dual-tree complex wavelet transform according to claim 1, characterized in that: The signal-to-noise ratio improvement requirement for indicator evaluation is greater than 10dB, and the mean square error is less than the set threshold.

9. The method for denoising multi-source signals in PE pipelines based on dual-tree complex wavelet transform according to claim 1, characterized in that: Time synchronization optimization uses the period of the single pendulum wave reference signal as a reference to perform interpolation calibration on sensor signals whose time deviation exceeds the set threshold; frequency comparison calibration adjusts the bandpass filter parameters of each signal based on the theoretical frequency of the single pendulum wave reference signal.

10. The method for denoising multi-source signals in PE pipelines based on dual-tree complex wavelet transform according to claim 4, characterized in that: For the high-frequency complex detail coefficients obtained by dual-tree complex wavelet transform using two sets of filters in parallel, soft threshold processing is performed on their real and imaginary parts respectively, and then the denoised high-frequency complex detail coefficients are synthesized.

Images