The hottest vortex flow signal processing based on

Vortex flow signal processing based on spectral analysis

Abstract: Based on the simulation and analysis of the influence of non integer period sampling on spectral analysis method, the actual vortex flow signal processing method based on spectral analysis is discussed

key words: non integer period sampling; Spectrum analysis; Vortex signal processing based on spectrum analysis

Lin min

(non professionals of medical instrument should not operate and debug the equipment at will. Once the machine has an abnormal state, mention college, Shanghai Science & Technology University, Shanghai 200093, China)

Abstract: Spectrum analysis for vortex signal processing was dis c ussed on the basis of simulating and analyzing the effect of sampling in non co mplete periods on spectrum arithmetic.

Key words: sampling in noncomplete periods; spectrum analysis; Vortex signal processing

1 Introduction

the core problem of vortex flow signal processing is to expand the lower range limit of vortex flowmeter in low frequency band. Trying to solve this problem by using spectral analysis is one of the current research hotspots. References [1] to [3] used simulation methods to compare the periodic spectrum method based on FFT and the maximum entropy spectrum method based on Burg, and concluded that FFT algorithm requires more sampling data, which is suitable for suppressing low-frequency deterministic noise; The Burg algorithm needs less data and is suitable for suppressing random noise. However, they used the ideal waveform generated by the signal generator for spectral analysis, and did not calculate the spectrum of the actual vortex signal. In addition, the spectrum analysis is realized on the basis of full period sampling of the signal. However, due to the unknown signal frequency, it is impossible for the actual system to realize full period sampling of the vortex output signal. Therefore, full period sampling is only an ideal situation, which is almost impossible to realize in the actual system

in view of this, before applying spectral analysis to the actual flow signal processing, it is necessary to simulate and analyze the influence of non integer period sampling on the results of spectral analysis and processing through MATLAB software, so as to reduce the blindness in development and shorten the research cycle

2 simulation calculation of non integer period sampling

assume that the output signal of the flowmeter is

signal frequency fsig=20hz. Start sampling from time t=0, and the total sampling time remains unchanged. Then when φ= 360 ° · K, k=0, 1, 2,... Is full cycle sampling. In order to compare the influence of non integer period sampling on FFT spectrum analysis and Burg entropy spectrum analysis, we take n=1024 points, perform FFT calculation on the basis of 64 integer periods, and take n=128 points, perform Burg calculation on the basis of 16 integer periods, and make the offset angle φ= 0 °, 5 °, 10 °,..., 180 °, that is, the influence of non integer periodic sampling on the power spectrum peak P and frequency estimation F of the two algorithms is simulated and calculated in steps of 5 °. Now, the two groups of calculation results are drawn as Figure 1 and Figure 2 respectively for more intuitive analysis. In the figure, the abscissa represents the signal frequency value f (Hz); The ordinate represents the power spectral density value p (w/hz) amplified by 1000 times

comparing the calculation results of the two groups, it is found that:

I FFT spectral analysis has the maximum spectral peak P when the deflection angle is 0 °, that is, the whole period sampling, and the corresponding frequency value f is also the most accurate, which is 20Hz; With the increase of deflection angle, the peak value and frequency of spectrum decrease. The peak value p of the Burg spectrum is not the largest when the deflection angle is 0 °, but the corresponding frequency value f is the most accurate, also 20Hz; There is no obvious regular relationship between the spectral peak P and the deflection angle

Ⅱ. FFT spectrum analysis with the increase of deflection angle, the frequency value corresponding to each deflection angle decreases approximately linearly, that is, the deflection angle difference is equal, and the corresponding frequency difference is also approximately equal. But there is no such rule in Burg spectrum analysis

Ⅲ. The more the number of sampling periods, the less the influence of non integer period sampling on these two spectral analysis. For example, the frequency error of FFT spectrum analysis when sampling 64 cycles is 1/2 when sampling 32 cycles, 1/4 when sampling 16 cycles, and 1/8 when sampling 8 cycles. That is, the detected frequency error increases approximately linearly with the linear reduction of the number of sampling cycles. The error obtained by FFT spectrum analysis when it is biased 180 ° on the basis of 64, 32, 16 and 8 full cycle samples is 0.77%, 1.54%, 3.03% and 5.88% respectively

Ⅳ. The peak P of power spectrum obtained by FFT spectrum analysis is accurate; The frequency value corresponding to the spectral peak is accurate. The peak value p of power spectrum obtained by Burg spectrum analysis is inaccurate and irregular; But the frequency value corresponding to the spectral peak is also accurate

Ⅴ. The signal frequency value f is related to the number of sampling periods. As long as the number of cycles is large enough, the measurement error caused by non integer period sampling becomes very small and can be ignored

therefore, considering the above points and referring to the research conclusions of references [1] to [3], this paper chooses to process the actual vortex flow signal based on FFT spectral analysis

3 actual vortex flow signal processing

3.1 vortex flow signal acquisition system

the acquisition system is composed of two parts: vortex flow signal acquisition device and pc-6333 multi-function module in module out interface card that can be installed in PC. The acquisition device is composed of water pipe, LUGB vortex flowmeter of Tianjin Instrument Factory, admacse electromagnetic flowmeter, y90s-2 three-phase asynchronous electric market competitiveness of Hudong electric company, further improving the motivation and water pump, MicroMaster420 frequency converter of Siemens company, water tank and other parts (Figure 3). In the figure, the direction of the arrow indicates the direction of water flow

the working process of the acquisition device is: start the frequency converter, set the display frequency of the frequency converter, and the motor rotates at a certain speed under the control of the frequency converter to drive the water pump to work and pump water from the water tank. The water flows through the electromagnetic flowmeter to display the instantaneous flow percentage, and then returns to the water tank after being detected by the vortex flowmeter. The output analog signal of vortex flowmeter is converted into digital signal by pc-6333 analog in analog out interface card, and then sent to PC for further processing

3.2 vortex flow signal processing and analysis based on spectral analysis

set the working frequency of the water pump to 50Hz, 47.5hz, 45Hz,... Through the frequency converter, and gradually reduce in steps of 2.5hz. When it is lower than 35Hz, the water pump motor will automatically and gradually stop rotating. Corresponding to each frequency point, 2048 points of data are sampled respectively. The 2048 point sampling data is loaded into the matlab program, and the approximate whole period FFT calculation is carried out through the sptool toolbox. In order to facilitate the comparison and improve the calculation accuracy, at each pump frequency point, we take two sections of sampling data in different ranges, each section is approximately sampled in the whole period, and then calculate the average FFT value of the two sections, as shown in Figure 4. The abscissa represents the frequency average value f (Hz) obtained by spectral analysis; The ordinate represents the average flow value (%) displayed by the electromagnetic flowmeter at the frequency point of each pump. Each detection parameter has its range selection

it is found from Figure 4:

(1) the average FFT value of the whole cycle calculated at each pump frequency point decreases with the gradual reduction of the value of "pump frequency/flow"

(2) according to the user manual, the fluid frequency range that LUGB vortex flowmeter can detect is 13.191hz ~ 131.91hz, which is the frequency detection range when the flowmeter uses the traditional circuit threshold method to process vortex signals. When the "pump frequency/flow" is 35hz/60.58%, the vortex signal frequency obtained by using the spectrum analysis method is 11.0726hz, which is about 2Hz lower than the lower limit frequency detected by the traditional circuit threshold method. From this point of view, the spectrum analysis method is superior to the traditional circuit threshold method in dealing with vortex signals, especially when measuring low velocity sections

(3) the sampling frequency selected in this experiment is 1kHz, and the number of sampling points is 2048. Therefore, if the frequency resolution is

then the relative measurement error of the system at low flow rate is

. It can be seen that if the measurement accuracy is expected to be 0.45%, the resolution should be below 0.05Hz when the signal frequency is unchanged, and the sampling frequency should be reduced to below 100Hz when the number of sampling points is unchanged, which cannot meet Shannon's sampling theorem. If the sampling frequency remains unchanged and the resolution is below 0.05Hz, the number of sampling points needs to be increased to more than 20480 points. However, increasing the number of points will increase the amount of data storage, increase the amount of calculation and increase the calculation time, which will reduce the real-time performance of the system. When the sampling frequency and sampling points are fixed, the lower the signal frequency is, the greater the measurement error is. In this regard, the method of setting the sampling frequency in sections is used to meet the requirements of calculation accuracy and real-time performance of the system at the same time. Frequency segmentation requires frequent switching of sampling frequency, which makes it difficult to achieve signal acquisition and truly meet the real-time performance of the system. Therefore, this method cannot fundamentally solve the problem

4 experimental conclusions and work prospects

at low flow rates, especially when the signal frequency is below 10Hz, the vortex signal and noise signal almost overlap, and even the amplitude of the noise is slightly larger than that of the vortex signal. At this time, if the vortex flow signal is processed by power spectrum analysis method, how can the noise that is likely to be obtained be smeared with plaster mortar? In fact, the peak value of the spectrum is higher than the amplitude of the signal spectrum, which will mistake the noise frequency for the vortex signal frequency. From this point of view, it is very difficult to reach the lower range limit of the extended vortex flowmeter at low flow rate simply by using the spectrum analysis method

however, using spectrum analysis can well show the frequency distribution characteristics of the signal, and can initially provide the frequency of the vortex signal, which is ready for the further accurate detection of the vortex signal frequency, and provides the research foundation and direction. On the basis of spectrum analysis, we have designed an expert system for vortex signal threshold processing, and have made some progress. Next, we will continue to use artificial intelligence and expert system methods to process vortex flow signals in combination with spectral analysis, in order to make a breakthrough in expanding the lower range of vortex flowmeter

references

[1] Xu Kejun, et al Comparison of spectral analysis methods of vortex flowmeter output signal [J] Journal of Hefei University of technology, 1994, 17 (2):

[2] Xu Kejun, et al A signal processing system of vortex flowmeter with spectrum analysis function based on DSP [J] Journal of instrumentation, 2001, 22 (3)

[3] Huang Zhiyun, Xu Kejun Spectrum correction method and application based on FFT [J] Journal of instrumentation, 2001, 22 (3):

[4] Xu Kejun, et al Common technology in automatic detection and instrumentation [M] Beijing: Tsinghua University Press, 2000 (end)