A signal processing method for a coriolis mass flowmeter based on quadrature demodulation

Abstract

The application designs a Coriolis mass flowmeter signal processing method based on orthogonal demodulation; an orthogonal demodulation method CH-QD based on correlation operation and Hilbert transform is proposed, first, autocorrelation operation is carried out on any one sensor signal, and the obtained autocorrelation signal is used as the same frequency cosine reference signal of the sensor signal; the autocorrelation signal is subjected to Hilbert transform, and the obtained Hilbert transform signal is used as the same frequency sine reference signal of the sensor signal; according to the same frequency and orthogonal relationship between the autocorrelation signal and the Hilbert transform signal thereof, the instantaneous frequency of the signal is calculated by complex number and trigonometric function operation; the two sensor signals are respectively subjected to sine and cosine reference signal demodulation, the center frequency of the signal is moved to the double frequency of the original frequency, and the low frequency component carrying the signal phase information is contained in the signal; the high frequency component is filtered out through a multi-stage low pass filter, and the low frequency component is retained; the low frequency component is used to extract the phase difference of the two signals, and the high-precision density, mass flow and other information of the fluid are provided.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of mass flow measurement and signal processing, and particularly relates to a Coriolis mass flowmeter signal processing method based on quadrature demodulation. BACKGROUND

[0002] With the continuous improvement of economic level, the rapid development of various industries and the increasing research and development efforts of high-precision instruments and meters, the Coriolis mass flowmeter (CMF for short, hereinafter referred to as the Coriolis flowmeter) is widely favored in many fields such as petroleum and chemical industry, pharmaceutical manufacturing, food processing, energy metering, environmental protection and transportation, etc. due to its unique advantages of being able to directly measure the accurate mass flow and density of fluid and other physical quantities under non-contact conditions. When the Coriolis flowmeter is working normally, the measuring tube is driven by the exciter to vibrate at the natural frequency, and when fluid flows through the measuring tube, the Coriolis force F c caused by the Coriolis effect will be generated on the inlet side and outlet side of the measuring tube, which is proportional to the mass flow and has opposite directions The phase difference between the two vibration signals picked up by the sensors at both ends of the measuring tube is proportional to the mass flow Q m of the fluid flowing through the measuring tube, while the vibration frequency f0 of the measuring tube is inversely proportional to the fluid density p, and the proportional coefficient K of the flowmeter is also independent of the physical properties of the fluid and only related to the parameters of the flowmeter itself. Therefore, the performance of the signal processing method of the transmitter directly determines the important performance indicators such as the measurement accuracy and repeatability of the Coriolis flowmeter.

[0003] At present, the mainstream frequency detection algorithms for Coriolis flowmeters include DFT method and its derivative algorithms, digital zero-crossing method, Hilbert method, adaptive lattice notch filter method, etc. The performance difference of these algorithms is small, and the algorithm accuracy basically meets the requirements of instruments. Moreover, the phase difference of the two sensor signals of the Coriolis flowmeter is small, and its measurement accuracy is more easily affected by algorithm performance and other factors. Therefore, most of the relevant research teams at home and abroad focus on the research of phase difference detection algorithms, and have successively proposed a variety of phase difference detection algorithms suitable for Coriolis flowmeters, mainly including:

[0004] (1) Digital zero-crossing detection algorithm: the digital zero-crossing detection algorithm records the time of signal zero-crossing, and performs data interpolation near the zero-crossing point to measure the time interval between the two zero-crossing points, thereby simultaneously detecting and tracking the frequency and phase difference of the signal. This algorithm has small amount of calculation and fast response speed, but it only uses the information of signal zero-crossing point, and when the signal is mixed with interference signal, the fluctuation and deviation of the detected zero-crossing point are large, which leads to the detection error of the algorithm.

[0005] (2) Frequency domain transform method: The frequency domain transform method transforms the finite length signal from time domain to frequency domain, and detects the signal parameters in the frequency domain, which can effectively suppress the interference signal. However, the frequency leakage problem caused by the non-integer period truncation of the time domain signal reduces the accuracy of the algorithm, and the traditional Fourier transform method has a large amount of calculation. Therefore, researchers at home and abroad have successively proposed many frequency domain transform methods to improve the accuracy and reduce the calculation amount. The Micro Motion Company in the United States proposed a Coriolis flowmeter signal processing method based on the discrete Fourier transform (DFT) method. However, when the non-integer period sampling is performed, the calculation accuracy of the DFT method cannot meet the instrument requirements. Therefore, the company proposed a coarse measurement, fine measurement and frequency tracking scheme, but the key technology of the scheme was not disclosed. After that, many schemes use window function method or interpolation processing method to suppress the influence of frequency spectrum leakage. The window function method can increase the width of the main lobe and reduce the frequency resolution, while suppressing the frequency spectrum leakage phenomenon. It is often used to correct the signal spectrum and has achieved certain results, but it has not fundamentally solved the problem of frequency spectrum leakage. In order to improve the calculation accuracy and real-time performance, Hefei University of Technology proposed to use an adaptive lattice notch filter filtering method for signal frequency tracking, and combined with the use of DTFT algorithm with sliding window considering negative frequency (SDTFT) to calculate the phase information of the signal. When calculating the Fourier coefficients of the signal, the influence of the negative frequency component is considered, and a recursive algorithm is used, thereby improving the calculation accuracy and shortening the convergence process. However, the algorithm needs to know the signal frequency in advance, so the frequency calculation error will introduce a secondary error to the algorithm. Moreover, when the signal frequency changes, the adaptive lattice filter needs to converge to the new frequency value again, so there is a large error in the phase measurement of the convergence process.

[0006] (3) Traditional orthogonal demodulation algorithm: The traditional orthogonal demodulation algorithm (traditional QD method) first needs to generate two sinusoidal and cosine reference signals with the same frequency as the sensor signal. The two reference signals are used to demodulate the two sensor signals respectively, and then the high-frequency components are filtered out through a low-pass filter. The frequency and phase difference of the sensor signal are detected according to the information of the sensor signal retained in the low-frequency component. The algorithm principle is simple, has strong harmonic and random noise suppression ability, and can detect frequency offset in a short time. However, it relies on the known signal frequency to generate the same frequency sine and cosine reference signals, which not only requires additional frequency detection, but also when there is a slight deviation between the frequency estimation result and the true frequency, the deviation of the reference signal will gradually accumulate with the increase of the signal period, which will undoubtedly add additional error to the phase difference detection. Therefore, the traditional QD method is sensitive to the performance of the frequency detection algorithm and the signal frequency fluctuation, and the accuracy of the algorithm depends on the design of the low-pass filter.

[0007] Therefore, it can be seen that the existing Coriolis flowmeter signal processing methods have inherent defects in the algorithm itself, measurement accuracy, need to know the signal frequency in advance, and are susceptible to interference. SUMMARY

[0008] In order to overcome the shortcomings and deficiencies of the prior art, the application discloses a Coriolis mass flowmeter signal processing method based on orthogonal demodulation, and proposes a CH-QD orthogonal demodulation method based on correlation operation and Hilbert transform, which effectively solves the problems of limited measurement accuracy, the need to know the signal frequency, and the sensitivity to the performance of the frequency detection algorithm.

[0009] A Coriolis mass flowmeter signal processing method based on orthogonal demodulation, specifically comprising the following steps:

[0010] Step 1: Set the preprocessed signals of the Coriolis flowmeter sensor signals after FIR filtering as x1(n) and x2(n); the specific formula is as follows:

[0011]

[0012] Wherein, A1 and A2 are the amplitudes of the signals, f0 is the resonance frequency of the measuring tube, f s is the sampling frequency, θ1 and θ2 are the phases of the signals, n=1, 2, 3...N, and N is the number of signal points collected;

[0013] Step 2: For a sinusoidal power signal x(t) with a period of T, the autocorrelation operation is formula (2), and the autocorrelation signal R xx (τ) obtained will retain the frequency information of the original signal and lose the phase information; according to this characteristic, autocorrelation is performed on any one of the preprocessed signals to obtain a cosine reference signal x3(n) with the same frequency as the sensor signal;

[0014]

[0015]

[0016] Wherein, T is the signal period, A is the signal amplitude, θ is the signal phase, ω is the signal angular frequency, τ is the time offset, f0 is the resonance frequency of the measuring tube, f s is the sampling frequency;

[0017] Step 3: Since the Hilbert transform is defined in the time domain as the convolution of x(t) and , according to the time domain convolution theorem, it is equivalent to multiplying a full-pass filter with an amplitude-frequency characteristic |H(ω)|=1 and a phase-frequency characteristic , therefore, the Hilbert transform is performed on the cosine signal x3(n) to generate a same-frequency sinusoidal reference signal x4(n);

[0018]

[0019] Step 4: Construct a complex signal E(n) from the autocorrelation signal x3(n) and the Hilbert transform signal x4(n): the constructed complex signal E(n) is as follows:

[0020]

[0021] Step 5: The complex signal E(n) and its time-shifted conjugate signal E * (n-1) are subjected to mathematical operations of formulas (6), (7) and inverse trigonometric function operations to detect the real-time frequency f0(n) of the signal, n being a signal sampling point;

[0022] The step 5 is specifically as follows:

[0023]

[0024]

[0025] Step 6: Take the two preprocessed signals x1(n) and x2(n) as frequency-modulated signals, and demodulate the two preprocessed signals with the above-mentioned same-frequency cosine and sine reference signals, wherein the demodulated signal of the preprocessed signal x1(n) is x q1 (n), and the demodulated signal of the preprocessed signal x2(n) is x q2 (n):

[0026]

[0027]

[0028] Step 7: At this time, the demodulated signals x q1 (n) and x q2 (n) contain low-frequency components and 2 times high-frequency components, and are subjected to multi-stage IIR low-pass filtering to filter out high-frequency (4πf0 / f s n+θ1) components to obtain the same-direction component I1 and the quadrature component Q1 of the preprocessed signal:

[0029]

[0030] The filter coefficients LP1 and LP2 are designed by formulas (11), (12), wherein fs is the sampling frequency and fc is the cutoff frequency; when filtering the kth sampling data, the filter output y1(k) of the input x(k) is calculated according to formula (13);

[0031]

[0032] LP2 = 1 - LP1 (12)

[0033] y1(k) = LP1 y(k-1) + LP2 x(k) (13)

[0034] The orthogonal demodulation signal is filtered by using the scheme of filtering step by step by using four independent IIR low-pass filters, wherein the cut-off frequencies of the first two filters are selected as f c1 = 0.5f0, and the cut-off frequencies of the last two filters are selected as f c2 = 0.1f0; the first two times of filtering eliminate high-frequency components, and the last two times of filtering further eliminate high-frequency components and filter out residual low-frequency noise, so as to achieve better filtering effect;

[0035] Step 8: the other preprocessed signal x2(n) is demodulated and low-pass filtered by using step 6 and step 7, so as to obtain its same-direction component I2 and orthogonal component Q2:

[0036]

[0037] Step 9: the same-direction component I and the orthogonal component Q of the phase difference of the two sensor signals are obtained by using formula (15) and (16), and the real-time phase difference of the signal is calculated according to formula (17) so as to extract the time difference Δt of the two sensor signals;

[0038]

[0039]

[0040]

[0041]

[0042] Step 10: the real-time frequency and time difference of the sensor signal of the above CH-QD detection are sorted by using sorting and truncation filtering, and the parameters of 20% before and after are selected as singular values and discarded, so as to improve the precision of the quality flow detection; and the instantaneous quality flow detected per second is filtered by using 5-point sliding average filtering, so as to avoid the large amplitude mutation of the instantaneous flow, while the response performance of the instrument can still be ensured; according to the time difference of the two signals, the calibrated instrument coefficient is used to measure the real-time quality flow.

[0043] The present application has the beneficial technical effects:

[0044] The signal processing method has high precision, stability and anti-noise performance, does not need to know the signal frequency in advance, can realize joint estimation of frequency and phase difference, and enables the Coriolis flowmeter transmitter to meet the requirements of industrial level. Compared with a phase difference detection algorithm which depends on signal frequency input and is sensitive to frequency detection algorithm, the application does not need to know the signal frequency in advance, does not depend on high-quality frequency estimation, and avoids secondary error introduced by frequency input to phase difference detection. Compared with a frequency and phase difference detection algorithm which depends on signal periodicity parameter information, the sensor signal can be fully utilized, the signal real-time frequency, phase difference and time difference are detected at each sampling point, and the detection precision and dynamic response speed are higher. The filter parameter design of the low-pass filter link is simple and effective, the traditional online design scheme by means of Matlab does not need to be used, and the multi-stage filtering scheme can realize good suppression of high-frequency harmonic noise and random noise, and further enhance the anti-noise performance of the algorithm. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 is a block diagram of the CH-QD algorithm of the embodiment of the application;

[0046] Figure 2 is a hardware system block diagram of the transmitter of the embodiment of the application;

[0047] Figure 3 is a software system block diagram of the transmitter of the embodiment of the application;

[0048] Figure 4 is a flowchart of the main monitoring program of the transmitter of the embodiment of the application;

[0049] Figure 5 is a waveform diagram of the actual field two-way sensor signal of the embodiment of the application;

[0050] Figure 6 is a frequency spectrum diagram of the actual field two-way sensor signal of the embodiment of the application;

[0051] Figure 7 is a phase spectrum diagram of the actual field two-way sensor signal of the embodiment of the application;

[0052] Figure 8 is a cosine reference signal waveform diagram of the embodiment of the application;

[0053] Figure 9 is a sine reference signal waveform diagram of the embodiment of the application;

[0054] Figure 10 is the average relative error of the phase difference detection result under different phase differences of the embodiment of the application;

[0055] Figure 11 is the average MSE of the phase difference detection result under different signal-to-noise ratios of the embodiment of the application. Detailed Implementation

[0056] The present invention will be further described below with reference to the accompanying drawings and embodiments;

[0057] This invention provides an orthogonal demodulation algorithm based on correlation operation and Hilbert transform (CH-QD method for short) as a signal processing method for Coriolis flowmeters. The algorithm block diagram is shown below. Figure 1 As shown, firstly, autocorrelation is performed on any one sensor signal to obtain an autocorrelation signal that serves as a cosine reference signal of the same frequency as the sensor signal. Then, the autocorrelation signal is subjected to a Hilbert transform to obtain a Hilbert transform signal that serves as a sine reference signal of the same frequency as the sensor signal. Based on the same frequency and orthogonality relationship between the autocorrelation signal and its Hilbert transform signal, the instantaneous frequency of the signal is calculated using complex number operations and trigonometric function operations. After demodulation of the two sensor signals by the sine and cosine reference signals respectively, the center frequency of the signal is shifted to twice the original frequency, while the signal contains low-frequency components carrying phase information. High-frequency components are filtered out by a multi-stage low-pass filter, retaining the low-frequency components. Finally, the phase difference between the two signals is extracted using the low-frequency components, and combined with the transmitter's instrument coefficient, high-precision information such as fluid density and mass flow rate is provided in real time.

[0058] The hardware system block diagram of the transmitter of the present invention is as follows: Figure 2 As shown, the system includes a signal conditioning and AD sampling module, a digital signal processing and human-machine interaction module, a drive module, and a temperature compensation module. The signal conditioning and AD sampling module includes an amplification and filtering circuit 1, an analog-to-digital converter 1 (ADC1), an amplification and filtering circuit 2, and an analog-to-digital converter 2 (ADC2). It mainly amplifies and filters the analog signals representing the vibration characteristics of the measuring tube output by the primary instrument's sensor 1 and sensor 2, and performs A / D conversion under the control of the ARM. The digital signal processing and human-machine interaction module is a core board circuit built around the ARM, including an ARM chip, external SRAM, and interactive interfaces. It is used for sensor signal sampling control, digital signal processing, mass flow detection, LED indication, LCD display, current output, pulse output, serial communication, and other system peripheral configuration. The drive module is controlled by a separate ARM chip and is used to output a drive signal that drives the primary instrument's measuring tube to oscillate continuously, creating an environment that can generate Coriolis force in the measuring tube and maintain the Coriolis flow meter in a stable working state. The temperature compensation module includes an amplification and filtering circuit 3 and an analog-to-digital converter 3 (ADC3), which is used to collect ambient temperature information provided by the temperature sensor signal to compensate for the mass flow.

[0059] The transmitter software system block diagram of the present invention is as follows: Figure 3As shown, the whole system software design is complicated, the modular design method is adopted, the subprogram of specific function is combined into corresponding function module, the function module is called by the main monitoring program or called by each other, meanwhile, the execution of the main program is interrupted by the interrupt module to process special events. The software design mainly includes initialization module, watchdog module, interrupt module, data acquisition module, signal processing algorithm module, pulse output module, serial communication module, human-computer interaction module and the like, each module completes the system function under the general scheduling of the main monitoring program, realizes the real-time acquisition of signals, the real-time display and output of quality flow and the like information, and constitutes a complete Coriolis flowmeter transmitter system.

[0060] The flow chart of the transmitter main monitoring program of the application is as shown in Figure 4 The specific steps include the following: after the system is powered on and reset, the main monitoring calls the initialization module to complete the initialization setting of the system, peripherals and algorithms; then, the sampling frequency of AD sampling is controlled by the timer 0, and the synchronous AD sampling is started to collect the signals of the instrument sensor 1 and sensor 2 once; the timer 1 is started to time 1s, and the quality flow and the like information of the current time are refreshed every second; when the number of collected data points reaches the preset number N, the signal processing algorithm module is called to sequentially pre-process, detect the frequency, detect the phase difference, sort and cut off the filter and slide the average filter of the sampling signal, and finally calculate the quality flow and the like information; subsequently, the pulse output module and the serial communication module are called, the corresponding pulse quantity is sent according to the detected instantaneous flow value, and the fluid information is sent to the host computer or other equipment through SCI communication; during the working of the transmitter, the background continuously inquires whether the user presses the keyboard, if yes, the corresponding user operation processing is carried out, such as modifying the system display setting, instrument parameter setting, instrument zero point correction, saving flow information, inquiring historical data, viewing time and date and the like, subsequently, the display module is called to refresh the information displayed on the LCD; whether the timer 1 timing time reaches 1s is checked, if not, the waiting is continued, if the 1s time arrives, the AD sampling link is returned to, and the next round of data sampling, signal processing and the like operation is restarted, and the cycle is continuously repeated until the user turns off the instrument or the watchdog module detects the abnormal situation and resets.

[0061] The core of the application lies in the signal processing algorithm module, wherein the signal pre-processing adopts the finite impulse response (FIR) band-pass filter to filter the two-way sensor signals of the instrument collected, the waveforms, frequency spectrum and phase spectrum of the sensor signals in the actual field are respectively as shown in Figure 5 、 Figure 6 、 Figure 7As shown, according to the characteristics of the signal base frequency set, the passband center frequency of the filter is set as the inherent frequency of the measuring tube to filter out the harmonic interference, power frequency interference and random noise interference mixed in the signal, and to improve the signal-to-noise ratio of the signal. Frequency detection and phase difference detection are performed by using the CH-QD algorithm proposed in the application to extract signal information. The mass flow detection part mainly sorts, truncates and filters the detection results of the CH-QD algorithm, calculates the fluid density, instantaneous mass flow, cumulative mass flow and other fluid information in combination with the instrument coefficient, and performs temperature compensation on the instantaneous mass flow according to the collected temperature information.

[0062] A Coriolis mass flowmeter signal processing method based on quadrature demodulation, specifically comprising the following steps:

[0063] Step 1: set the preprocessed signals of the Coriolis flowmeter sensor signals after FIR filtering as x1(n) and x2(n); the specific process is as follows:

[0064]

[0065] Wherein, A1 and A2 are the amplitudes of the signals, f0 is the resonance frequency of the measuring tube, f s is the sampling frequency, θ1 and θ2 are the phases of the signals, n=1, 2, 3...N, and N is the number of signal points collected;

[0066] Step 2: for a sinusoidal power signal x(t) with a period of T, the autocorrelation operation is formula (2), and the autocorrelation signal R xx (τ) obtained will retain the frequency information of the original signal and lose the phase information; according to this characteristic, autocorrelation is performed on any one of the preprocessed signals to obtain a cosine reference signal x3(n) with the same frequency as the sensor signal;

[0067]

[0068]

[0069] Wherein, T is the signal period, A is the signal amplitude, θ is the signal phase, ω is the signal angular frequency, τ is the time offset, f0 is the resonance frequency of the measuring tube, f s is the sampling frequency;

[0070] Step 3: since the Hilbert transform is defined in the time domain as the convolution of x(t) and , according to the time domain convolution theorem, it is equivalent to multiplying a full-pass filter with an amplitude-frequency characteristic |H(ω)|=1 and a phase-frequency characteristic in the frequency domain, therefore, Hilbert transform is performed on the cosine signal x3(n) to generate a same-frequency sinusoidal reference signal x4(n);

[0071]

[0072] The frequency of the Coriolis flowmeter sensor signal fluctuates in a base frequency range, and the base frequency inherent to different structures of primary instruments is different, so most quadrature demodulation algorithms need to estimate the sensor signal frequency in advance with the help of an additional frequency detection algorithm, which on the one hand increases the complexity of the signal processing method, and on the other hand, the error and convergence process of the frequency detection algorithm will cause a difference between the generated reference signal and the same frequency signal actually required by the quadrature demodulation algorithm. Since the phase difference between the two sensor signals of the Coriolis flowmeter is between 0.01-4°, and the phase difference under the zero state is even smaller, a slight deviation of the reference signal will cause a phase shift of the subsequent quadrature demodulation signal, which will undoubtedly add a secondary error to the phase difference estimation.

[0073] The same frequency cosine reference signal generated by the application and the same frequency cosine reference signal generated according to the frequency input are respectively as shown in Figure 8 、 Figure 9 When the number of signal periods is small, the difference between the two reference signals is not obvious. But with the increase of the number of signal periods, the same frequency cosine reference signal generated according to the frequency input will be affected by the deviation between the input frequency and the real frequency, and the deviation is in a state of gradual accumulation and cannot be ignored. The reference signal generated by the application does not depend on the frequency input, and the frequency of the generated reference signal is consistent with the original signal, which effectively avoids the secondary error introduced by the input frequency.

[0074] Step 4: A complex signal E(n) is constructed from the autocorrelation signal x3(n) and the Hilbert transform signal x4(n): the constructed complex signal E(n) is as follows:

[0075]

[0076] Step 5: The complex signal E(n) and its time-shift conjugate signal E*(n-1) are subjected to mathematical operations of formulas (6), (7) and inverse trigonometric function operations, to detect the real-time frequency f0(n) of the signal, n being a signal sampling point;

[0077] The step 5 is specifically as follows:

[0078]

[0079]

[0080] Step 6: The two preprocessed signals x1(n) and x2(n) are regarded as frequency modulation signals, and the above-mentioned same frequency cosine reference signal is used to demodulate the two preprocessed signals, respectively, wherein the demodulated signal of one preprocessed signal x1(n) is x q1 (n), and the demodulated signal of the other preprocessed signal x2(n) is x q2(n):

[0081]

[0082]

[0083]

[0084] Step 7: At this time, the demodulation signal x q1 (n), x q2 (n) contains low frequency components and 2 times high frequency components, which are filtered by a multi-stage IIR low-pass filter to filter out high frequency (4pif0 / f s n+q1) components to obtain the in-phase component 11 and the quadrature component Q1 of the preprocessed signal:

[0085]

[0086] Because the design of the filter directly affects the accuracy of the quadrature demodulation, and considering the complexity and real-time performance of the filtering calculation, the present application selects a multi-stage infinite impulse response (IIR) filter scheme for low-pass filtering. Although the phase of the IIR filter presents a nonlinear characteristic between the frequency, the order of the IIR filter is much smaller than that of the FIR filter with linear phase, and the nonlinear delay of the two-way same frequency signals through the same IIR filter is consistent, so it does not affect the calculation result of the phase difference. The filter coefficients LP1 and LP2 are designed by formulas (11) and (12), where fs is the sampling frequency and fc is the cutoff frequency. When filtering the kth sampling data, the filter output y1(k) of the input x(k) is calculated according to formula (13);

[0087]

[0088] LP2=1-LP1 (12)

[0089] y1(k)=LP1·y(k-1)+LP2·x(k) (13)

[0090] The cutoff frequency f c The selection of the cutoff frequency f c When the cutoff frequency f c is selected to be relatively high, the delay of the filter is low, but the output is sensitive to noise and the phase difference fluctuation is large; when the cutoff frequency f c is selected to be relatively low, the output is relatively stable and the phase difference fluctuation can be significantly reduced, but the delay of the filter is high. Through multiple simulation tests, the best filtering scheme is selected, and finally the present application adopts a scheme of four independent IIR low-pass filters for step-by-step filtering of the quadrature demodulation signal, wherein the cutoff frequencies of the first two filters are selected as f c1 =0.5f0, and the cutoff frequencies of the last two filters are selected as f c2= 0.1f0. The first two filters can eliminate high frequency components to a great extent, and the last two filters can further eliminate high frequency components and residual low frequency noise to achieve better filtering effect.

[0091] Step 8: The other preprocessed signal x2(n) is demodulated and low-pass filtered by using step 6 and step 7 to obtain its in-phase component I2 and quadrature component Q2:

[0092]

[0093] Step 9: The in-phase component I and quadrature component Q of the phase difference of the two sensor signals are obtained by using formulas (15), (16), and the real-time phase difference of the signals is calculated according to formula (17) Thus, the time difference Δt of the two sensor signals is extracted:

[0094] I = Q1·Q2 + 11·12

[0095]

[0096]

[0097]

[0098]

[0099] Step 10: The real-time frequency and time difference of the sensor signals detected by the above CH-QD are sorted by using sorting and truncation filtering, and the parameters of the first and last 20% are selected as singular values and discarded, thereby improving the accuracy of the mass flow detection; and the instantaneous mass flow detected per second is subjected to 5-point sliding average filtering, thereby avoiding large amplitude mutation of the instantaneous flow while still ensuring the response performance of the instrument; according to the time difference of the two signals, the calibrated instrument coefficient is used to measure the real-time mass flow.

[0100] In industrial field, the actual output signal of the Coriolis flowmeter sensor is disturbed by power frequency, resonance harmonic and random noise. Therefore, this paper uses MATLAB to generate a sine signal containing quadratic, cubic harmonic, 50Hz power frequency disturbance and random noise to simulate the field sensor signal, and the signal model is defined as follows:

[0101]

[0102] Where θ1 and θ2 are signal phases, e1(n) and e2(n) are disturbance terms, containing power frequency, quadratic, cubic harmonic with relative amplitude of 10%, and Gaussian white noise with adjustable signal-to-noise ratio, and the sampling frequency f s= 10 kHz, signal base frequency f0= 205.5 Hz, amplitude 1 V, sampling point number N = 8000.

[0103] In order to verify the accuracy of the method, 500 simulation experiments were carried out on the method and several existing mainstream algorithms based on the signal model (19) with different phase differences under 30 dB random noise, and the average relative error of the phase difference detection results is shown in Table 1. Figure 10 According to the results, the relative errors of the five algorithms decrease with the increase of the phase difference and remain stable when the phase difference increases to a certain value, and the relative error of the CH-QD algorithm proposed in the application is obviously smaller than that of the other algorithms, and the relative error is less than 0.04%, which can meet the accuracy requirement of the Coriolis flowmeter for phase difference measurement.

[0104] In order to verify the anti-noise performance of the method, 500 independent simulation tests were carried out on the method and several existing mainstream algorithms based on the signal model (19) with different signal-to-noise ratios SNR, and the average mean square error MSE of the phase difference detection results is shown in Table 2. Figure 11 According to the results, the MSE of the CH-QD algorithm proposed in the application is smaller than that of the other algorithms under different signal-to-noise ratios, and the accuracy and stability under low signal-to-noise ratio conditions are obviously better than those of the other algorithms, which sufficiently proves that the method has good anti-noise performance.

Claims

1. A signal processing method for a Coriolis mass flowmeter based on quadrature demodulation, characterized by, Specifically comprising the following steps: Step 1: Set the pre-processed signal of the Coriolis flowmeter sensor signal after FIR filtering as and ; Step 2: For a sinusoidal power signal x(t) with a period of T, perform autocorrelation operation on it, and the autocorrelation signal obtained is will retain the frequency information of the original signal and lose the phase information; according to this characteristic, autocorrelation is performed on any one preprocessed signal to obtain a cosine reference signal with the same frequency as the sensor signal ; Step 3: Since the Hilbert transform is defined in the time domain as the convolution of x(t) with , according to the time domain convolution theorem, it is equivalent in the frequency domain to multiplying by an all-pass filter with an amplitude frequency characteristic of and a phase frequency characteristic of , so a Hilbert transform is performed on the cosine reference signal , which is co-frequency with the sensor signal, to produce a co-frequency sine reference signal ; Step 4: Constructing the complex signal E(n) from the cosine reference signal at the same frequency as the sensor signal and the same frequency sine reference signal Constructing the complex signal E(n) Step 5: complex signal E(n) and its time-shift conjugate signal After mathematical operation and inverse trigonometric function operation, the real-time frequency of the signal is detected n is the signal sampling point; Step 6: two preprocessed signals are mixed and considered as frequency modulation signals, the two preprocessed signals are demodulated by using the above-mentioned same frequency cosine and sine reference signals, wherein one preprocessed signal is demodulated by using the cosine reference signal, and the other preprocessed signal is demodulated by using the sine reference signal the demodulated signals are , ; Step 7: At this time, the demodulated signal , contains low-frequency components and 2 times frequency high-frequency components, and the high-frequency components are filtered out by multi-stage IIR low-pass filtering to obtain the co-directional component and the quadrature component of the preprocessed signal and ;​ Step 8: using the result of step 6 and step 7 to process another path of pre-processed signal Demodulating and low-pass filtering to obtain its co-directional component and quadrature component ; Step 9: Calculate the in-phase component I and the quadrature component Q of the phase difference of the two sensor signals, and calculate the real-time phase difference of the signals , thereby extracting the time difference of the two sensor signals ; Step 10: The real-time frequency and time difference of the sensor signal detected by the CH-QD detection method of orthogonal demodulation is sorted by using the sorting truncation filter, and the parameters of the former and latter 20% are selected as singular values and discarded, so as to improve the precision of the mass flow detection; And the instantaneous mass flow detected per second is subjected to 5-point sliding average filtering, so as to avoid the large amplitude mutation of the instantaneous flow, while still being able to guarantee the response performance of the instrument; according to the time difference of the two signals, the calibrated instrument coefficient is used to measure the mass flow in real time.

2. A signal processing method for a Coriolis mass flowmeter based on quadrature demodulation as recited in claim 1, characterized by, Step 1 is specifically as follows: ； wherein and is the amplitude of the signal, is the resonance frequency of the measuring tube, is the sampling frequency, and is the phase of the signal, n = 1,2,3...N, N is the number of collected signal points.

3. The signal processing method for a Coriolis mass flowmeter based on quadrature demodulation as recited in claim 1, characterized in that, Step 2 correlation operation is formula and get the cosine reference signal with the same frequency as the sensor signal Specifically: ； ； where T is the signal period, A is the signal amplitude, and Θ is the signal phase, is the signal angular frequency, and τ is the time offset, is the resonance frequency of the measurement tube, is the sampling frequency.

4. The method of claim 1, wherein the method further comprises: The generation of the in-phase sinusoidal reference signal is described in step 3 Specifically: 。 5. The method of claim 1 wherein the method further comprises the step of, The complex signal E(n) constructed in step 4 is as follows: 。 6. The method of claim 1 wherein the method further comprises the step of, The step 5 mathematical operation and inverse trigonometric function operation are specifically as follows: ； 。 7. The method of claim 1 wherein the method further comprises the step of, The step 6 signal is , Specifically: ； 。 8. The method of claim 1 wherein the method is a signal processing method for a Coriolis mass flowmeter based on quadrature demodulation. The step 7 co-directional component and quadrature component : ； The filter coefficients LP1 and LP2 are designed by equations (11), (12), where fs is the sampling frequency, fc is the cut-off frequency, and the filtered output of the input x(k) is calculated according to equation (13) when filtering the kth sampling data ; ； ； ； The orthogonal demodulation signal is filtered by a scheme of four independent IIR low-pass filters filtering step by step, wherein the cut-off frequencies of the first two filters are selected as , and the cut-off frequencies of the last two filters are selected as ; the first two times of filtering eliminate high-frequency components, and the last two times of filtering further eliminate high-frequency components and filter out residual low-frequency noise, so as to achieve better filtering effect.

9. The method of claim 1 wherein the method further comprises the step of, The step 8 co-directional component and quadrature component : 。 10. The method of claim 1 wherein, The step 9 co-directional components I, quadrature components Q and real-time phase difference Specifically: ； ； ； 。

Images