EMD (Empirical Mode Decomposition) and correlation analysis-based double-probe vortex shedding flowmeter anti-interference method

Abstract

The invention provides an EMD (empirical mode decomposition) and correlation analysis-based double-probe vortex shedding flowmeter anti-interference method, which relates to the technical field of signal processing, and is characterized in that a vortex shedding sensor signal sequence containing periodic vibration interference is decomposed into multi-scale IMF components through self-adaptive characteristics of an EMD method, and spectrum components of a flow signal and vibration noise are accurately separated. According to the method, aiming at the non-stable and multi-band coupling characteristics of double-probe signals, the vibration noise interference under the complex vibration interference of an industrial field is effectively solved, and the environmental adaptability of noise suppression is remarkably improved. According to the method, on the basis that the advantages of self-adaptive decomposition are reserved, the IMF component series is dynamically controlled, that is, only the IMF components of the first four stages are taken, redundant calculation caused by over-decomposition is avoided, and the real-time performance of the system is guaranteed. An upper probe signal is introduced to serve as a reference channel, interference components are removed through a correlation analysis method, the complexity of a blind source separation algorithm is reduced, and meanwhile the method has higher detection precision and stability.

Description

Technical Field

[0001] The present invention relates to the technical field of signal processing, and specifically to an anti-interference method for a dual-probe vortex flowmeter based on EMD and correlation analysis. Background Art

[0002] In the fields of industrial and agricultural production, the requirements for fluid metering performance are becoming increasingly strict. The vortex flowmeter, with its advantages of no mechanical moving parts, small pressure loss, wide range of measuring ranges, and the ability to be applicable to various media such as gases, liquids, and saturated steam, demonstrates excellent adaptability and reliability in industrial measurements and occupies an important position in the domestic and international flowmeter markets. The vortex flowmeter consists of a vortex sensor and a transmitter. The research object of the present invention is its transmitter part, including modules such as sensor signal acquisition, signal conditioning, digital signal processing system, and display control. The core principle of the vortex flowmeter originates from the "Karman vortex street" theory in fluid mechanics. When the fluid flows through a non-streamlined cylinder (i.e., a vortex generator) in the measuring pipeline, a periodic alternating vortex train is generated. When the longitudinal and transverse spacings of adjacent vortices satisfy a certain proportional relationship, a stable eddy current is formed. The piezoelectric element in the sensor probe generates a charge signal with the same frequency as the eddy current characteristics by picking up the eddy current characteristics. A large number of experiments have proven that within a certain range of Reynolds numbers, the frequency of this charge signal is proportional to the average velocity of the fluid in the pipeline, its amplitude is proportional to the square of the frequency, and is independent of the composition and viscosity of the measured medium.

[0003] Currently, digital signal processing algorithms are the mainstream trend in the field of vortex flow detection. Digital methods are usually more complex and require a large amount of computing resources. However, industrial instruments have strict requirements for real-time performance. Therefore, in practical applications, it is necessary to weigh whether the computing power of the microprocessor can support the implementation of complex algorithms. On the other hand, the interference signals with a wide frequency range generated by on-site mechanical vibrations overlap with the characteristic frequencies of the flow signals. Especially in low-flow velocity conditions, the vibration noise intensity exceeds the flow signal, significantly affecting the flow measurement. Therefore, while reducing the algorithm complexity and ensuring the real-time performance of the system, achieving high-precision frequency detection in a strong noise environment has become a research hotspot. Currently, the research on vortex flow signal processing methods includes two directions: frequency detection technology and anti-interference technology.

[0004] Frequency detection techniques mainly include the interpolation FFT method, the sparse Fourier transform (SFT) method, and the multiple signal classification (MUSIC) method, etc. Since the traditional FFT method will produce spectral leakage phenomenon during non-integer period sampling, it is usually necessary to introduce a window function to truncate the input signal. In addition, due to the limitations of the sampling frequency and sampling length, the "fence effect" will occur in discrete spectrum analysis, resulting in limited frequency resolution, which in turn affects the frequency detection accuracy. The SFT method realizes the characteristic analysis of the entire signal sequence by constructing an approximate compressed representation of the discrete Fourier transform, and processes the main frequency components in the signal spectrum. However, when processing non-stationary signals, due to the lack of time-frequency analysis ability, this method may not be able to accurately capture the instantaneous characteristics of the signal, so it is difficult to extract accurate flow information from complex eddy current signals. The MUSIC method is based on the eigen-decomposition of the autocorrelation matrix, and uses the orthogonality of the signal subspace and the noise subspace to perform high-resolution estimation of the signal, which can improve the frequency estimation accuracy of the signal to a certain extent. However, due to its large computational complexity, there are still certain difficulties in practical engineering applications. In contrast, the interpolation FFT method based on the ratio of Fourier coefficients solves the above problems well and can achieve precise correction of the signal frequency.

[0005] The anti-interference technology focuses on suppressing the interference of periodic vibration noise. To solve this technical problem, researchers have proposed various solutions: the Burg modern spectrum estimation method, the EMD scale filtering method, the Kalman filtering method, etc. Among them, the Burg algorithm can control the relative error of frequency estimation within 3% by constructing a quantitative relationship model between the vortex shedding frequency and the optimal model order. However, both the accuracy and the operation efficiency of spectrum estimation are strongly correlated with the model order, resulting in difficulty in determining the optimal model order for eddy current signals in different frequency bands through an adaptive mechanism. The EMD scale filtering method decomposes the eddy current signal by EMD to obtain IMF components in different frequency bands, and determines the effective signal frequency based on its energy distribution. However, this filter performs poorly when there are two or more dominant frequencies in the signal, and the measurement error is large. The improved Kalman filtering method improves the traditional KF algorithm by integrating fuzzy logic optimization and iterative parameter adjustment, and shows significant advantages in terms of dynamic tracking ability, environmental noise suppression effect, and computational efficiency. However, this method has extremely high computational complexity and is currently only limited to the simulation test stage, and it is difficult to test its application effect in a specific industrial environment.

[0006] At present, it is difficult to effectively balance the measurement accuracy, system stability and algorithm complexity of the signal processing method of the vortex flowmeter. This technical defect is particularly significant when there is periodic vibration interference in the industrial field, resulting in difficulty in extracting the frequency information of the flow signal. At the same time, due to the computational power of the hardware system, complex anti-interference algorithms cannot be implemented, and ultimately the real-time performance of the instrument is difficult to meet the actual needs, seriously restricting its performance optimization and engineering application. Summary of the Invention

[0007] Aiming at the deficiencies of the prior art, the purpose of the present invention is to propose an anti-interference method for a dual-probe vortex flowmeter based on EMD and correlation analysis, including:

[0008] Step 1: Collect the upper probe signal sequence x1(n) and the lower probe signal sequence x2(n) through a dual-probe sensor;

[0009] Step 2: Perform EMD decomposition on the lower probe signal sequence x2(n) to obtain the first-level IMF component c1(n), the second-level IMF component c2(n), the third-level IMF component c3(n), and the fourth-level IMF component c4(n);

[0010] Step 3: For each IMF component among the first-level IMF component c1(n), the second-level IMF component c2(n), the third-level IMF component c3(n), and the fourth-level IMF component c4(n), calculate the Pearson correlation coefficient between the IMF component and the upper probe signal sequence x1(n) to obtain the Pearson correlation coefficients of the four IMF components;

[0011] Step 4: Among the Pearson correlation coefficients of the four IMF components, determine the IMF component with the largest Pearson correlation coefficient value, and superimpose the other three IMF components except this IMF component to obtain the reconstructed signal sequence x(n);

[0012] Step 5: Adopt the rectangular window spectrum function in the spectrum centroid correction method, deduce and construct a correction equation with the frequency deviation amount as a variable, and process the reconstructed signal sequence x(n) based on the correction equation to obtain the corrected signal frequency f'.

[0013] Optionally, Step 2 specifically includes:

[0014] Step 2.1: Take the lower probe signal sequence x2(n) as the current signal sequence;

[0015] Step 2.2: Obtain all local extreme points of the current signal sequence, and all local extreme points include maximum points and minimum points;

[0016] Step 2.3: Adopt the cubic spline interpolation algorithm to construct the upper envelope curve e max (n) based on the maximum points, and construct the lower envelope curve e min (n) based on the minimum points;

[0017] Step 2.4: Calculate the mean curve m1(n) of the upper envelope curve e max (n) and the lower envelope curve e min (n);

[0018] Step 2.5: Calculate the difference between the current signal sequence and the mean curve m1(n) to obtain the first-level IMF component c1(n).

[0019] Step 2.6: Use the first-level IMF component c1(n) as the current signal sequence and return to execute Step 2.2. When executing to Step 2.5, obtain the second-level IMF component c2(n). Use the second-level IMF component c2(n) as the current signal sequence and return to execute Step 2.2. When executing to Step 2.5, obtain the third-level IMF component c3(n). Use the third-level IMF component c3(n) as the current signal sequence and return to execute Step 2.2. When executing to Step 2.5, obtain the fourth-level IMF component c4(n).

[0020] Optionally, Step 2.4 is specifically implemented through the following formula:

[0021] m1(n) = (e max (n) + e min (n)) / 2.

[0022] Optionally, Step 3 specifically includes:

[0023] For each IMF component among the first-level IMF component c1(n), the second-level IMF component c2(n), the third-level IMF component c3(n), and the fourth-level IMF component c4(n), calculate the mathematical expectation E(C) of the IMF component, and at the same time calculate the mathematical expectation E(X) of the upper probe signal sequence x1(n). Calculate the Pearson correlation coefficient P(X,C) between the IMF component and the upper probe signal sequence x1(n) according to E(C) and E(X). Perform the above calculations for each IMF component to obtain the Pearson correlation coefficients of the four IMF components.

[0024] Optionally, calculating the Pearson correlation coefficient P(X,C) between the IMF component and the upper probe signal sequence x1(n) according to E(C) and E(X) is specifically implemented through the following formula:

[0025]

[0026] where, x i represents the i-th value in the upper probe signal sequence x1(n), c i represents the i-th value of the IMF component, and N represents the number of sampling points when the dual-probe sensor collects the signal sequence.

[0027] Optionally, the formula of the rectangular window spectrum function W(ω) in Step 5 is expressed as:

[0028]

[0029] where ω is the angular frequency.

[0030] Optionally, in step 5, the reconstructed signal sequence x(n) is processed based on the calibration equation to obtain the calibrated signal frequency f', which is specifically implemented through the following formula:

[0031]

[0032] where k is the spectral line position of the spectral peak corresponding to the reconstructed signal sequence x(n), and f s represents the sampling frequency when the dual-probe sensor collects the signal sequence, and Δk is an intermediate representation used to simplify the formula, where Δk is represented by the following formula:

[0033]

[0034] where Y(k) is the amplitude of the k-th spectral line in the spectrum corresponding to the reconstructed signal sequence x(n), Y(k + 1) is the amplitude of the (k + 1)-th spectral line in the spectrum corresponding to the reconstructed signal sequence x(n), and Y(k - 1) is the amplitude of the (k - 1)-th spectral line in the spectrum corresponding to the reconstructed signal sequence x(n).

[0035] The beneficial effects produced by adopting the above technical solution are as follows:

[0036] Through the adaptive characteristic of the EMD method, the present invention decomposes the vortex street sensor signal containing periodic vibration interference into multi-scale IMF components, and accurately separates the spectral components of the flow signal and vibration noise. Aiming at the non-stationary and multi-band coupling characteristics of the dual-probe signal, this method effectively solves the vibration noise interference under complex vibration interference in the industrial field, and significantly improves the environmental adaptability of noise suppression. By optimizing the EMD termination condition, while retaining the advantages of adaptive decomposition, the present invention controls the number of IMF component series dynamically, that is, only taking the first four levels of IMF components, avoiding redundant calculations caused by over-decomposition, and ensuring the real-time performance of the system. By introducing the upper probe signal as a reference channel, the present invention calculates the Pearson correlation coefficient in the four-level IMF components by using the correlation analysis method, and obtains the component with the largest Pearson correlation coefficient among them and removes it. By reconstructing the signal with the remaining three components in the four-level IMF components, the complexity of the blind source separation algorithm is reduced, and the present invention has higher detection accuracy and stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 is a schematic flow chart of the EMD and correlation analysis fusion algorithm in the embodiment of the present invention;

[0038] Figure 2 is a schematic flow chart of the anti-interference method for a dual-probe vortex flowmeter based on EMD and correlation analysis in the embodiment of the present invention;

[0039] Figure 3Schematic diagram of the hardware system in the embodiment of the present invention;

[0040] Figure 4 Schematic diagram of the software system in the embodiment of the present invention;

[0041] Figure 5 Schematic diagram of the main monitoring program flow combining software and hardware in the embodiment of the present invention;

[0042] Figure 6 Experimental result diagrams of different flow points, different frequencies, and amplitude vibration interferences in the embodiment of the present invention. Among them, (a) is the maximum relative error result diagram of frequency detection for different flow points, different frequencies, and amplitude vibration interferences, and (b) is the variance result diagram for different flow points, different frequencies, and amplitude vibration interferences. Specific embodiments

[0043] The following combines the accompanying drawings and embodiments to further describe in detail the specific embodiments of the present invention. The following embodiments are used to illustrate the present invention but are not used to limit the scope of the present invention.

[0044] Aiming at the problems of the prior art, the anti-interference method of the dual-probe vortex flowmeter based on EMD and correlation analysis of the present invention aims to effectively eliminate the vibration interference components based on the output signal characteristics of the dual-probe vortex sensor through the collaborative processing and compensation mechanism of the dual-channel signals, so as to accurately extract the flow information. While ensuring high-precision flow detection, by optimizing the signal processing scheme, the algorithm complexity is reduced to meet the real-time requirements of the system, and the anti-interference performance of the system is significantly improved.

[0045] The present invention provides an anti-interference method for a dual-probe vortex flowmeter based on EMD and correlation analysis. The core idea is to perform EMD decomposition on the original upper probe signal to obtain several intrinsic mode functions representing different vibration modes, and then introduce the lower probe signal as a reference channel. Through correlation analysis, the frequency of the interference signal is accurately identified, the frequency component is suppressed, and the remaining components are reconstructed into a characteristic signal to effectively remove the interference components and restore the true characteristics of the signal. Finally, the frequency value of the flow signal is calculated through the spectrum analysis method. The overall scheme is as Figure 1As shown in the figure. The algorithm process is as follows: First, perform EMD decomposition on the lower probe signal sequence to obtain multiple levels of IMF components, including the first-level IMF component IMF1, the second-level IMF component IMF2, the third-level IMF component IMF3, and the fourth-level IMF component IMF4. These components represent the changes in the signal at different frequency scales. According to the IMF correlation analysis of the lower probe signals collected at different flow points and under different frequency vibration interferences, it is found that the first four modal components can cover the important frequency components in the sensor signal. Therefore, combined with the upper probe signal sequence, the first four levels of intrinsic mode functions are taken to calculate the correlation coefficient, and then feature extraction is performed. Specifically, the component with the largest correlation is selected and removed through the comparison of this coefficient, and the three obtained components are the results of feature extraction; then, modal reconstruction is performed on the remaining effective signal components to eliminate the influence of vibration noise; finally, the frequency value of this flow-dominant signal is calculated through a frequency detection method, which can be understood as correcting the reconstructed signal to obtain the corrected signal frequency.

[0046] The core algorithm module of the present invention consists of an anti-periodic vibration interference algorithm and a frequency detection algorithm. Among them, the anti-interference algorithm uses the above-mentioned EMD and correlation analysis fusion algorithm, and the frequency detection algorithm uses the energy center of gravity correction method (SCCM). According to the signal characteristics of the dual-probe sensor: the upper probe detection unit can only respond to vibration signals, while the lower probe detection unit synchronously senses vibration interference and vortex street flow signals. The time-frequency domain waveforms of the dual-probe signals collected under different working conditions in the industrial field, where the IMF components decomposed by the EMD method need to meet two basic criteria: First, the number of extreme points and zero-crossing points maintains a strict correspondence relationship, being equal or at most differing by one; Second, the zero-mean characteristic of the local extreme envelope, that is, the mean value of the upper and lower envelope lines formed by the extreme points approaches zero at any time point, ensuring that the component has time-domain symmetry. Specifically, combined with Figure 2 , the anti-interference method for a dual-probe vortex flowmeter based on EMD and correlation analysis provided by the present invention may include the following steps:

[0047] Step 1: Collect the upper probe signal sequence x1(n) and the lower probe signal sequence x2(n) through a dual-probe sensor;

[0048] Step 2: Perform EMD decomposition on the lower probe signal sequence x2(n) to obtain the first-level IMF component c1(n), the second-level IMF component c2(n), the third-level IMF component c3(n), and the fourth-level IMF component c4(n);

[0049] Step 2.1: Take the lower probe signal sequence x2(n) as the current signal sequence;

[0050] Step 2.2: Obtain all local extreme points of the current signal sequence, where the all local extreme points include maximum points and minimum points;

[0051] Step 2.3: Use the cubic spline interpolation algorithm to construct the upper envelope curve e max (n) based on the maximum points, and construct the lower envelope curve e min (n) based on the minimum points;

[0052] Step 2.4: Calculate the mean curve m1(n) of the upper envelope curve e max (n) and the lower envelope curve e min (n), which is specifically implemented by the following formula:

[0053] m1(n) = (e max (n) + e min (n)) / 2.

[0054] Step 2.5: Calculate the difference between the current signal sequence and the mean curve m1(n) to obtain the first-level IMF component c1(n);

[0055] Step 2.6: Take the first-level IMF component c1(n) as the current signal sequence and return to execute Step 2.2. When executing to Step 2.5, obtain the second-level IMF component c2(n). Take the second-level IMF component c2(n) as the current signal sequence and return to execute Step 2.2. When executing to Step 2.5, obtain the third-level IMF component c3(n). Take the third-level IMF component c3(n) as the current signal sequence and return to execute Step 2.2. When executing to Step 2.5, obtain the fourth-level IMF component c4(n).

[0056] Through the frequency-domain characteristic analysis of the seven-level IMF obtained by decomposing the actual lower probe signal, it is found that under different flow conditions, when a vibration interference with a frequency close to the flow signal frequency is applied, even if there are power frequency and other harmonic interferences, the IMF1-IMF4 components of the lower probe signal can still cover all the important frequency components of the signal, including the key features of the flow signal and the vibration interference. These components show significant energy concentration characteristics in both the time domain and the frequency domain, and can fully reflect the core features of the signal. In contrast, the modal components after IMF5 mainly contain the low-frequency components and noise of the signal, with low energy and limited contribution to the feature extraction of the flow signal. Therefore, the present invention realizes the extraction of the flow signal frequency only through the first four levels of components.

[0057] After obtaining the four-level IMF components, use the correlation analysis method to calculate the correlation coefficients of each component with the upper probe signal sequence respectively. Specifically, use the Pearson correlation coefficient as the quantization parameter for the correlation between the EMD decomposition result and the original signal. The specific steps refer to Step 3.

[0058] Step 3: For each of the first-level IMF component c1(n), the second-level IMF component c2(n), the third-level IMF component c3(n), and the fourth-level IMF component c4(n), calculate the Pearson correlation coefficient between the IMF component and the upper probe signal sequence x1(n) to obtain the Pearson correlation coefficients of the four IMF components;

[0059] For each of the first-level IMF component c1(n), the second-level IMF component c2(n), the third-level IMF component c3(n), and the fourth-level IMF component c4(n), calculate the mathematical expectation E(C) of the IMF component, and at the same time calculate the mathematical expectation E(X) of the upper probe signal sequence x1(n). Calculate the Pearson correlation coefficient P(X,C) between the IMF component and the upper probe signal sequence x1(n) according to E(C) and E(X), which is specifically implemented by the following formula:

[0060]

[0061] where, x i represents the i-th value in the upper probe signal sequence x1(n), and c i represents the i-th value of the IMF component, and N represents the number of sampling points when the dual-probe sensor collects the signal sequence.

[0062] Thus, the above calculations are performed for each IMF component to obtain the Pearson correlation coefficients of the four IMF components.

[0063] Among them, in the specific implementation process of the present invention, after obtaining E(C) and E(X), the covariance between the IMF component and the upper probe signal sequence x1(n) can also be calculated, which is specifically implemented by the following formula:

[0064]

[0065] The greater the covariance, the more significant the correlation degree between the two signal sequences; otherwise, the correlation is weaker. In addition, due to the dimensional difference, the value of the covariance will have a scale dependence. Therefore, in order to eliminate the interference of the dimension, it needs to be standardized, and thus the dimensionless Pearson correlation coefficient is deduced. Furthermore, combined with the covariance, it is further calculated through the formula for calculating P(X,C) to obtain the Pearson correlation coefficients of the four IMF components.

[0066] Step 4: Among the Pearson correlation coefficients of the four IMF components, determine an IMF component with the largest Pearson correlation coefficient value, and superimpose the other three IMF components except this IMF component to obtain the reconstructed signal sequence x(n);

[0067] Among them, the core mechanism of the spectral centroid correction method (SCCM) lies in analyzing the spectral characteristics of the normalized windowed processing, selecting two spectral lines adjacent to both sides of the center of the main lobe with a spacing exactly equal to the frequency resolution Δf, constructing a correction equation with the frequency deviation as the variable, obtaining the frequency correction value by solving this equation, and then using the spectral centroid correction method to correct the reconstructed signal sequence x(n). For specific reference, see step 5.

[0068] Step 5: Use the rectangular window spectral function in the spectral centroid correction method to derive and construct a correction equation with the frequency deviation as the variable, and process the reconstructed signal sequence x(n) based on the correction equation to obtain the corrected signal frequency f'.

[0069] Among them, the formula of the rectangular window spectral function W(ω) is expressed as:

[0070]

[0071] Among them, ω is the angular frequency.

[0072] Therefore, based on the rectangular window spectral function W(ω), the derivation and construction of a correction equation with the frequency deviation as the variable specifically include:

[0073] When ω = 2πn / N (n = ±1, ±2,...), W = 0, the main lobe spectral width is 4π / N, and the adjacent spectral line spacing is exactly equal to the main lobe width 2π / N. Therefore, there are two spectral lines within the main lobe. Assume they are the kth and the (k + 1)th. At this time, the window function:

[0074]

[0075] When N >> 1, 1 / N → 0, sin(πk) ≈ πk / N, and within the main lobe, W0(k) = Nsin(πk) / πk. Let Y = W0 / N = sin(πk) / πk. Based on the characteristic that the energy distribution center of the discrete spectrum of the rectangular window function converges to the geometric center of the main lobe: kY(k) + (k + 1)Y(k + 1) = 0, it can be deduced that

[0076]

[0077] The calculated center coordinate is the position of the corrected spectral line:

[0078]

[0079] Let Among them, Y(k + 1) takes the adjacent spectral line with a higher amplitude, that is, one of Y(k + 1) and Y(k - 1). That is, the correction amount is:

[0080]

[0081] In the formula, k is the spectral line position corresponding to the spectral peak, Y(k) is the amplitude of the k-th spectral line, and Y(k + 1) and Y(k - 1) are the amplitudes of the (k + 1)-th and (k - 1)-th spectral lines respectively. After calibration, the signal frequency can be obtained as follows:

[0082]

[0083] Thus, a calibration equation with the frequency deviation as a variable is obtained. Based on this, in step 5, the reconstructed signal sequence x(n) is processed according to the calibration equation to obtain the calibrated signal frequency f'. Combining with the specific reconstructed signal sequence x(n), it is specifically implemented through the following formula:

[0084]

[0085] where k is the spectral line position of the spectral peak corresponding to the reconstructed signal sequence x(n), and f s represents the sampling frequency when the dual-probe sensor collects the signal sequence. Δk is an intermediate representation used to simplify the formula, and Δk is represented by the following formula:

[0086]

[0087] where Y(k) is the amplitude of the k-th spectral line in the spectrum corresponding to the reconstructed signal sequence x(n), Y(k + 1) is the amplitude of the (k + 1)-th spectral line in the spectrum corresponding to the reconstructed signal sequence x(n), and Y(k - 1) is the amplitude of the (k - 1)-th spectral line in the spectrum corresponding to the reconstructed signal sequence x(n).

[0088] The anti-interference method for dual-probe vortex flowmeters based on EMD and correlation analysis provided by the present invention can be implemented based on a hardware system. The structural schematic diagram of the hardware system is as Figure 3, which mainly consists of a signal conditioning module, a power conversion module, a digital signal processing system, and a communication module. The signal conditioning module mainly completes the functions of amplifying and filtering sensor signals. It leads out dual-channel differential charge signals through the upper and lower probe signal lines, converts them into voltage signals through charge integrator 1 and charge integrator 2 to complete the preliminary signal amplification. Then, the second-stage signal amplification is carried out through programmable gain amplifier 1 and programmable gain amplifier 2. At the same time, the main control ARM can adjust the amplification multiples of the two programmable gain amplifiers by controlling the digital potentiometer. After that, the high-frequency noise mixed in the signal is filtered out through low-pass filter 1 and low-pass filter 2. Finally, the impedance between the signal and the digital signal processing system is effectively matched through voltage follower 1 and voltage follower 2 to reduce signal loss. The power conversion module includes two parts of voltage stabilizing and step-down modules for converting 24V to 5V and 5V to 3.3V, providing the required working voltages for operational amplifiers, digital potentiometers, digital signal processing systems, communication modules, etc. The reset circuit is used to monitor system anomalies, and the clock management module provides an accurate clock for the system to ensure the stable operation of the system. The digital signal processing system is used to implement the anti-interference method of the dual-probe vortex flowmeter based on EMD and correlation analysis. The digital signal processing system includes an ARM chip and other interactive interfaces. Other interactive interfaces include AD, GPIO, UART, and TIMER. The digital signal processing system is used to implement the sampling control of dual-channel signals, digital signal processing, flow calculation, LCD display, button control, pulse output, serial communication, and RS485 communication.

[0089] The anti-interference method of the dual-probe vortex flowmeter based on EMD and correlation analysis provided by the present invention can also be implemented based on a software system. The structural schematic diagram of the software system is as Figure 4 , adopting the modular design concept. Under the unified scheduling of the STM32F103RCT6 main control module and the coordination of the interrupt mechanism, each module jointly realizes the signal processing and flow integration of the vortex flowmeter, provides flexible human-computer interaction and communication functions, and constitutes a complete vortex flowmeter transmitter system. Among them, the initialization module is responsible for the initialization configuration of system peripherals and algorithm parameters after the system is powered on, specifically including system initialization, peripheral initialization, and algorithm initialization; the signal conditioning circuit module adjusts the amplification multiple of the programmable gain amplifier by adjusting the resistance value of the digital potentiometer; the dual-channel data acquisition module realizes the synchronous acquisition of dual-channel signals; the signal processing algorithm module executes the anti-interference method of the dual-probe vortex flowmeter based on EMD and correlation analysis; the serial communication module is used for parameter configuration and monitoring during system debugging; the pulse output module converts the algorithm processing result into the corresponding pulse frequency output. The human-computer interaction module realizes the real-time display and page switching of the flow measurement result through the liquid crystal display module and the button control module, and the RS485 module is used for the remote transmission of the industrial field test result. The watchdog and interrupt module monitors the system operation status to ensure the stable operation of the system.

[0090] The anti-interference method of the dual-probe vortex flowmeter based on EMD and correlation analysis provided by the present invention is combined with Figure 5 The main monitoring program flow of the present invention combined with software and hardware includes:

[0091] After the system is powered on and reset, the main monitoring first performs initialization, including system clock initialization, ADC calibration, GPIO port setting, timer configuration, digital potentiometer setting, and initialization of the LCD screen and communication port. Since the vortex flowmeter needs to refresh the flow information once within 1 second during actual operation, in order to ensure the real-time performance of the system, it is necessary to start a periodic timer to trigger the signal acquisition and processing tasks. After the initialization is completed, the program enters the main loop and waits for the timer interrupt. When the timer interrupt is triggered, the system starts the dual-channel signal acquisition module for dual-channel data acquisition, synchronously collects the output signals of the dual-channel vortex sensors through the on-chip AD, and stores the data in the buffer to determine whether the number of points of the collected data reaches the sampling point number. If the number of points of the collected data does not reach the sampling point number, continue to perform dual-channel data acquisition. If the number of points of the collected data reaches the sampling point number, turn off the ADC interrupt to prevent it from being repeatedly triggered and constantly overwriting the data in the current buffer. Subsequently, the collected lower probe signal is subjected to EMD decomposition to determine whether the decomposed components reach the component level. If the decomposed components do not reach the component level, the EMD decomposition is returned to be executed. If the decomposed components reach the component level, four-level IMF components are obtained, including the first-level IMF component IMF1, the second-level IMF component IMF2, the third-level IMF component IMF3 and the fourth-level IMF component IMF4. The correlation coefficients are calculated for them and the upper probe signal respectively, and the maximum correlation is eliminated. The remaining IMF components are reconstructed, and the frequency of the reconstructed signal is calculated to obtain its flow value, that is, the corrected signal frequency. After the calculation is completed, the system clears the collected data of this group in time for the next round of data collection. Afterwards, the calculated flow data is sent to the host PC through the serial port for remote monitoring and data recording. At the same time, the pulse output module generates a pulse signal of the corresponding frequency according to the instantaneous flow value for equipment calibration. In the main loop, the system detects the button status in real time, that is Figure 5Determine whether there is a key operation. When no key operation is detected, perform the key operation, adjust the pulse output, and refresh the LCD display. Specifically, call the key processing module to execute corresponding functions, such as switching the display interface, adjusting system parameters, etc. Then, the LCD display module updates the display content, refreshing the flow information, system status, and parameters, etc. When a key operation is detected, adjust the pulse output and refresh the LCD display. Finally, the system checks whether the timer has reached 1 second. If not, continue to wait; if so, clear the timer and start the next count, entering the next round of data acquisition, signal processing, and display update process. The entire system runs in a loop under the efficient scheduling of the main monitoring program until the device is turned off or the watchdog module detects an abnormality and triggers a system reset.

[0092] Through the adaptive characteristics of the EMD method, the present invention decomposes the vortex sensor signal containing periodic vibration interference into multi-scale IMF components, accurately separating the spectral components of the flow signal and vibration noise. Aiming at the non-stationary and multi-band coupling characteristics of the dual-probe signal, this method effectively solves the vibration noise interference under complex vibration interference in the industrial field, significantly improving the environmental adaptability of noise suppression. By optimizing the EMD termination condition, while retaining the advantages of adaptive decomposition, by dynamically controlling the number of IMF component series, that is, only taking the first four levels of IMF components, redundant calculations caused by over-decomposition are avoided, ensuring the real-time performance of the system. By introducing the upper probe signal as a reference channel, the present invention calculates the component with the largest Pearson correlation coefficient among the four-level IMF components using the correlation analysis method, removes it, and reconstructs the signal through the remaining three components among the four-level IMF components, reducing the complexity of the blind source separation algorithm while making the present invention have higher detection accuracy and stability.

[0093] Using the anti-interference algorithm of the present invention, experiments are carried out at four typical flow rates (20 m 3 / h, 40 m 3 / h, 80 m 3 / h, 120 m 3 / h, 200 m 3 / h). At each flow rate, a vibration interference close to the frequency of the flow signal is applied. Since the ratio λ N of the vibration interference amplitude to the flow signal amplitude under actual working conditions usually ranges between 0.5 and 2.0, in order to comprehensively evaluate the influence of vibration interference on detection accuracy and stability, vibration interferences at four proportionality coefficients of 0.5, 1.0, 1.5, and 2.0 are taken for experiments, and the maximum relative error of frequency estimation when different frequency vibration interferences are applied at each flow rate is tested (that is, the case where the vibration interference is close to the center frequency of the flow signal), as shown in Figure 6 (a), where the abscissa is the flow rate Q v, the vertical axis is the maximum relative error of frequency estimation Err, and the variance of frequency detection at this time, as shown in Figure 6 (b) in, where the horizontal axis is the flow point Q v , and the vertical axis is the variance of frequency detection σ 2 . The experimental results show that the maximum value of the frequency estimation error gradually increases with the increase of the flow rate. At the flow point of 20 m 3 / h, when the proportionality coefficient λ N of the vibration interference to the amplitude of the flow signal is 2.0, the frequency estimation error and variance reach the maximum value. At the flow point of 200 m 3 / h, when the proportionality coefficient λ N of the vibration interference to the amplitude of the flow signal is 0.5, the minimum value is obtained. Overall, the present invention can maintain high precision and stability under all test conditions. The relative error of frequency estimation always remains at 0.615%, and the variance does not exceed 3.1×10 -3 . This result fully meets the requirements of the system for precision and stability, and at the same time verifies that the anti-vibration interference algorithm can exhibit excellent robustness under different flow points and vibration interference conditions, and can effectively meet the frequency estimation requirements under complex working conditions.

[0094] On this basis, the self-developed dual-probe anti-vibration vortex flowmeter was calibrated based on the industrial field calibration platform. The results show that the entire system has a relative indication error better than 1% and a repeatability better than 0.33% within a 10:1 range of measurement, meeting the industrial grade 1.0 instrument standard. The instrument has rich functions and stable hardware operation, and has certain industrial practical value.

[0095] The above description is only the preferred embodiment of the present disclosure and the description of the applied technical principles. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by the specific combination of the above technical features, but also covers other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above inventive concept. For example, the technical solutions formed by mutually replacing the above features with the technical features (but not limited to) having similar functions disclosed in the embodiments of the present disclosure.

Claims

1. A method for anti-interference of a dual-probe vortex flowmeter based on EMD and correlation analysis, characterized in that, Including: Step 1: Acquire the upper probe signal sequence x1(n) and the lower probe signal sequence x2(n) through a dual-probe sensor; Step 2: Perform EMD decomposition on the lower probe signal sequence x2(n) to obtain the first-level IMF component c1(n), the second-level IMF component c2(n), the third-level IMF component c3(n), and the fourth-level IMF component c4(n); Step 3: For each IMF component among the first-level IMF component c1(n), the second-level IMF component c2(n), the third-level IMF component c3(n), and the fourth-level IMF component c4(n), calculate the Pearson correlation coefficient between the IMF component and the upper probe signal sequence x1(n) to obtain the Pearson correlation coefficients of the four IMF components; Step 4: Among the Pearson correlation coefficients of the four IMF components, determine the IMF component with the largest Pearson correlation coefficient value, and superimpose the other three IMF components except this IMF component to obtain the reconstructed signal sequence x(n); Step 5: Adopt the rectangular window spectrum function in the spectrum centroid correction method, deduce and construct a correction equation with the frequency deviation amount as the variable, and process the reconstructed signal sequence x(n) based on the correction equation to obtain the corrected signal frequency f'.

2. The anti-interference method of the dual-probe vortex flowmeter based on EMD and correlation analysis according to claim 1, characterized in that Step 2 specifically includes: Step 2.1: Take the lower probe signal sequence x2(n) as the current signal sequence; Step 2.2: Obtain all local extreme points of the current signal sequence, and the all local extreme points include maximum points and minimum points; Step 2.3: Using the cubic spline interpolation algorithm, construct the upper envelope curve e max (n) based on the maximum points, and construct the lower envelope curve e min (n); Step 2.4: Calculate the upper envelope curve e max (n) and the mean curve m1(n) of the lower envelope curve e min (n); Step 2.5: Calculate the difference between the current signal sequence and the mean curve m1(n) to obtain the first-level IMF component c1(n); Step 2.6: Take the first-level IMF component c1(n) as the current signal sequence and return to execute Step 2.

2. Execute until Step 2.5 to obtain the second-level IMF component c2(n). Take the second-level IMF component c2(n) as the current signal sequence and return to execute Step 2.

2. Execute until Step 2.5 to obtain the third-level IMF component c3(n). Take the third-level IMF component c3(n) as the current signal sequence and return to execute Step 2.

2. Execute until Step 2.5 to obtain the fourth-level IMF component c4(n).

3. The anti-interference method of the dual-probe vortex flowmeter based on EMD and correlation analysis according to claim 2, wherein, Step 2.4 is specifically implemented through the following formula: m1(n) = (e max (n) + e min (n)) / 2。 4. The anti-interference method of the dual-probe vortex flowmeter based on EMD and correlation analysis according to claim 1, characterized in that, Step 3 specifically includes: For each IMF component among the first-level IMF component c1(n), the second-level IMF component c2(n), the third-level IMF component c3(n), and the fourth-level IMF component c4(n), calculate the mathematical expectation E(C) of the IMF component, and at the same time calculate the mathematical expectation E(X) of the upper probe signal sequence x1(n). Calculate the Pearson correlation coefficient P(X,C) between the IMF component and the upper probe signal sequence x1(n) according to E(C) and E(X). The above calculations are performed for each IMF component to obtain the Pearson correlation coefficients of the four IMF components.

5. The anti-interference method of the dual-probe vortex flowmeter based on EMD and correlation analysis according to claim 4, wherein, Calculate the Pearson correlation coefficient P(X,C) between the IMF component and the upper probe signal sequence x1(n) according to E(C) and E(X), which is specifically implemented through the following formula: where x i represents the i-th value in the upper probe signal sequence x1(n), c i represents the i-th value of the IMF component, and N represents the number of sampling points when the double probe sensor collects the signal sequence.

6. The anti-interference method of the double-probe vortex flowmeter based on EMD and correlation analysis according to claim 1, wherein The formula of the rectangular window spectrum function W(ω) in Step 5 is expressed as: Among them, ω is the angular frequency.

7. The anti-interference method of the double-probe vortex flowmeter based on EMD and correlation analysis according to claim 1, wherein, In step 5, the reconstructed signal sequence x(n) is processed based on the correction equation to obtain the corrected signal frequency f', which is specifically implemented through the following formula: where k is the spectral line position of the spectral peak corresponding to the reconstructed signal sequence x(n), and f s represents the sampling frequency when the dual-probe sensor collects the signal sequence, and Δk is an intermediate representation used to simplify the formula, where Δk is expressed by the following formula: Among them, Y(k) is the amplitude of the k-th spectral line in the spectrum corresponding to the reconstructed signal sequence x(n), Y(k + 1) is the amplitude of the (k + 1)-th spectral line in the spectrum corresponding to the reconstructed signal sequence x(n), and Y(k - 1) is the amplitude of the (k - 1)-th spectral line in the spectrum corresponding to the reconstructed signal sequence x(n).

Images