CN113176448B

CN113176448B - Conductivity detection method and system based on double conductivity sensors

Info

Publication number: CN113176448B
Application number: CN202110301689.1A
Authority: CN
Inventors: 逄博; 黄希; 官宇翔; 崔莉
Original assignee: Institute of Computing Technology of CAS
Current assignee: Institute of Computing Technology of CAS
Priority date: 2021-03-22
Filing date: 2021-03-22
Publication date: 2023-06-27
Anticipated expiration: 2041-03-22
Also published as: CN113176448A

Abstract

The invention provides a conductivity detection method and system based on a double conductivity sensor. The invention utilizes the response characteristics of the two conductivity sensors, carries out a data fusion method in a self-adaptive filter mode, estimates the system parameters of the two conductivity sensors, and compensates the output results of the two conductivity sensors. The method has the technical effects that the response delay of the seven-electrode conductivity sensor and the drift amplitude of the open four-electrode conductivity sensor can be estimated, and a high-precision measurement result is obtained.

Description

Conductivity detection method and system based on double conductivity sensors

Technical Field

The invention relates to the technical field of sensor signal processing, in particular to a technology for simultaneously measuring conductivity by using two sensors and performing signal processing on two groups of measurement data so as to obtain high-precision conductivity.

Background

Conductivity (Conductivity), temperature (Temperature), depth (Depth), i.e. thermal salt Depth sensor (CTD), are the most basic and important sensors in marine research, and these parameters are not only suitable for marine environment and ecological research, but also provide background compensation parameters for other sensors.

Traditional high performance CTD products such as sbe911+, SBE49, etc. can support salinity measurement accuracy of ± 0.003psu, and sampling rates above 10 Hz. The method mainly adopts an electrode type conductivity sensor with a channel, and the conductivity sensor obtains a response speed which is rapid, stable and nearly consistent with a temperature sensor through a high-speed constant-current water pump, so that the error of salinity conversion is reduced. However, because the water pump consumes a lot of power, it is difficult to use the water pump for a long time on a submerged buoy with limited energy, an underwater glider and other platforms. Meanwhile, the CTD without the water pump newly developed in recent years has different defects, such as slow water sample updating, long response time and instability of the self-brushing CTD sensor; the open CTD is prone to drift and low in accuracy. Therefore, there is a need for a high precision, fast response, low drift, low power CTD.

In recent years, a series of CTD sensors without water pumps are internationally developed, and a self-brushing or open type design and a related data processing method are adopted, so that higher response speed and dynamic measurement accuracy are realized with lower power consumption. For example, the U.S. wuz hall institute and the british national ocean center propose to use semi-open and open conductivity sensors, the electrodes of which are in direct contact with seawater, the water sample is updated rapidly, a faster response speed can be obtained without the need for a water pump, and fig. 1 is a graphical representation of an open conductivity sensor. However, due to direct contact with external seawater and its own trough structure, material aging and contaminant accumulation are easily caused, so that drift is generated, and sensor accuracy is reduced; the self-brushing Teledyne Underway CTD and Idronaut 320+ctd are self-brushing type, which are obtained by compensating the amplitude and the phase of the acquired data by using a constant linear response model of the self-flushing conductivity cell at a fixed arrangement speed, so that the measurement accuracy is obviously improved, and fig. 2 is a self-brushing type conductivity sensor physical diagram, but the self-brushing type CTD has great limitations in application, such as slow response speed, instability and the like. These new sensors have not yet been able to reach the level of sbe911+ completely in terms of accuracy.

Disclosure of Invention

Aiming at the demands of actual application on the precision, stability and response time of the conductivity parameters and the defect of overhigh power consumption of the existing sensor with the water pump, a self-brushing type and open type conductivity sensor cooperative working mode is adopted, and the fused conductivity measurement data has the advantage of high response speed of the open type conductivity sensor and the advantages of high measurement accuracy and strong anti-drifting capability of the self-brushing type conductivity sensor through data fusion based on the self-adaptation filter, so that the high-precision, fast response and low-power consumption measurement of the conductivity parameters are realized.

Aiming at the defects of the prior art, the invention provides a conductivity detection method based on a double conductivity sensor, which comprises the following steps:

step 1, conducting conductivity detection on liquid to be detected through a self-brushing type conductivity sensor and a linear response model thereof to obtain first conductivity, and conducting conductivity detection on the liquid to be detected through a drift ratio of constants of an open type conductivity sensor and a conductivity cell thereof to obtain second conductivity;

step 2, the second conductivity is used as an input signal to be input into the adaptive filter, the filtering result of the filter is compared with the first conductivity, and the coefficient of the adaptive filter is adjusted according to the difference value of the filtering result and the first conductivity;

and 3, circularly executing the step 1 and the step 2 until the difference value is converged, storing the current adaptive filter, updating the drift ratio according to the coefficient of the current adaptive filter to obtain a calibration ratio, and detecting the conductivity of the liquid to be detected by adopting the open conductivity sensor and the calibration ratio to obtain a third conductivity. The drift ratio is calibrated through the steps, and then the relatively more accurate third conductivity is obtained as the measurement result of the liquid to be measured.

The conductivity detection method based on the dual conductivity sensor comprises the following steps:

step 11, the self-brushing conductivity sensor actually measures the conductivity to obtain the first conductivity through a linear response model with low-pass and delay characteristics;

and step 12, multiplying the actual measured conductivity of the open conductivity sensor by the drift ratio of the constant of the conductivity cell, and then superposing the measured random error to obtain the second conductivity.

The conductivity detection method based on the double conductivity sensor further comprises the following steps:

and step 4, performing cross-correlation on the third conductivity and the first conductivity to obtain the delay of the self-brushing conductivity sensor relative to the open conductivity sensor so as to perform phase compensation on the first conductivity to obtain a fourth conductivity. I.e. by calibrating and compensating the delay of the open conductivity sensor by the above steps, a relatively more accurate fourth conductivity is obtained as a measurement result of the liquid to be measured.

and 5, judging whether the difference between the fourth conductivity and the third conductivity is larger than a threshold value, if so, adopting the third conductivity as a measurement result of the liquid to be measured, otherwise, adopting the fourth conductivity as the measurement result of the liquid to be measured.

The conductivity detection method based on the double conductivity sensors is characterized in that the self-brushing conductivity sensor and the open conductivity sensor are respectively provided with a plurality of electrodes for measuring conductivity.

The invention also provides a conductivity detection system based on the double conductivity sensor, which comprises:

the module 1 is used for conducting conductivity detection on the liquid to be detected through the self-brushing type conductivity sensor and a linear response model thereof to obtain first conductivity, and conducting conductivity detection on the liquid to be detected through the drift ratio of the constant of the open type conductivity sensor and the constant of the conductivity cell of the open type conductivity sensor to obtain second conductivity;

the module 2 is used for inputting the second conductivity as an input signal to the adaptive filter, comparing the filtering result of the filter with the first conductivity, and adjusting the coefficient of the adaptive filter according to the difference value of the filtering result and the first conductivity;

and the module 3 is used for circularly calling the module 1 and the module 2 until the difference value is converged, storing the current adaptive filter, updating the drift ratio according to the coefficient of the current adaptive filter to obtain a calibration ratio, and detecting the conductivity of the liquid to be detected by adopting the open conductivity sensor and the calibration ratio to obtain a third conductivity.

The conductivity detection system based on dual conductivity sensors, wherein the module 1 comprises:

the module 11 is configured to obtain the first conductivity by passing the conductivity actually measured by the self-brushing conductivity sensor through a linear response model with low-pass and delay characteristics;

the module 12 is configured to multiply the actual measured conductivity of the open conductivity sensor by the drift ratio of the conductivity cell constant and then superimpose the measured random error to obtain the second conductivity.

The conductivity detection system based on the double conductivity sensors further comprises:

and the module 4 is used for carrying out cross-correlation on the third conductivity and the first conductivity to obtain the delay of the self-brushing conductivity sensor relative to the open conductivity sensor so as to carry out phase compensation on the first conductivity to obtain the fourth conductivity.

and a module 5, configured to determine whether the difference between the fourth conductivity and the third conductivity is greater than a threshold value, and if so, use the third conductivity as a measurement result of the liquid to be measured, otherwise use the fourth conductivity as a measurement result of the liquid to be measured.

The conductivity detection system based on the double conductivity sensors, wherein the self-brushing conductivity sensor and the open conductivity sensor are respectively provided with a plurality of electrodes for measuring conductivity.

The advantages of the invention are as follows:

(1) According to the method, the time delay of the self-brushing conductivity sensor and the drift coefficient of the open conductivity sensor are estimated through self-adaptive filtering, and the two sensors are respectively compensated, so that the measurement results with high precision, low response time and low drift influence can be obtained.

(2) Compared with the existing method, the method does not need to additionally increase a water pump component, and saves the volume and the power consumption of the sensor.

Drawings

FIG. 1 is a schematic diagram of an open conductivity cell;

FIG. 2 is a schematic diagram of a self-brushing band-pass conductivity cell;

FIG. 3 is a schematic diagram of a system architecture;

FIG. 4 is a schematic diagram of a linear system H(s) model of sensor A;

FIG. 5 is a flow chart of data processing;

FIG. 6 is a graph of the conductivity x (t) in a simulated real environment;

FIG. 7 is an analog sensor A signal y _A And sensor B signal y _B ；

FIG. 8 is a graph comparing the signals of sensors A and B with the actual conductivity x signal;

FIG. 9 shows the adaptive filter output signal and the sensor A signal y _A A comparison chart;

FIG. 10 is a graph of adaptive filter output error versus time;

FIG. 11 is a graph of DC gain of an adaptive filter over time;

FIG. 12 is a graph showing the error between the DC gain of the adaptive filter and the k value over time after being subjected to moving average;

FIG. 13 is a graph of adaptive filter coefficients over time;

FIG. 14 is a graph showing the output compensated signal y _{A_fix} And y _{B_fix} ；

FIG. 15 is a graph comparing the compensated signal with the input signal and the original signal.

Detailed Description

The key technical point of the invention is that the response characteristics of the two conductivity sensors are utilized, the data fusion method is carried out by a self-adaptive filter mode, the system parameters of the two conductivity sensors are estimated, and the output results of the two conductivity sensors are compensated. The method has the technical effects that the response delay of the seven-electrode conductivity sensor and the drift amplitude of the open four-electrode conductivity sensor can be estimated, and a high-precision measurement result is obtained.

Specifically, in the invention, a self-brushing 7-electrode conductivity sensor (hereinafter referred to as sensor A) with a channel, which has high stability, but has a relatively slow response time and is related to factors such as flow rate and temperature, is adopted; an open 4-electrode conductivity sensor (hereinafter referred to as sensor B) having a response time of almost 0 but a large drift amount achieves the objectives of sensor accuracy (0.003 mS/cm), response time (70 mS), and month drift (0.002 mS/cm/month) by means of data fusion of an adaptive filter, and a flow chart is shown in fig. 3.

First, the principle of conductivity measurement will be briefly described, with conductivity sigma proportional to the conductance (1/R _s )，

R _s For solution impedance, the proportionality coefficient κ, the conductivity cell constant, will drift over time depending on the geometry of the electrodes and the geometry and size of the water volume in the vicinity of the electrodes.

In accordance with the above principle, two data fusion methods of conductivity sensors are described in detail below:

step 1 sensor A and sensor B simultaneously measure sea water conductivity

If the conductivity in the actual environment is x (t), the output value y of the sensor A _A It can be seen that x (t) is obtained by passing a linear response model H(s) with low pass and delay characteristics, and H(s) varies with flow rate and temperature. Meanwhile, since the sensor a has no drift, the dc gain of H(s) is 1, and the model thereof is shown in fig. 4.

While for sensor B, which is in direct contact with external seawater, there is no delay, sensor B will drift due to changes in conductivity cell constant caused by material aging or contaminant attachment. This amount of drift is proportional to the measured value. Therefore, it can be considered that the output value y _B The method comprises the following steps:

y _B (t)＝k*x(t)+n(t)

where k is the drift ratio of the conductivity cell constants, which varies slowly with time, when k=1, the sensor has no drift; n (t) is the random error of its measurement, typically gaussian white noise, and its standard deviation is the resolution of sensor B.

Thus, if we can pass y _A And y is _B Data comparison of (2) to determine k and to compare y _B Gain compensation is performed to obtain a fused measurement x (t) without response delay and drift:

from the above analysis we can construct a linear transformation H(s) that is band stationary and causally compliant so that y _B The value after this linear transformation is as close to y as possible _A . The desired model of H(s) is H(s) =h (s)/k. From the above analysis, the DC gain of H(s) is 1/k. That is, as long as we can find H(s), we can calculate k and calculate x (t) according to the above formula, achieving our goal.

Step 2 construction of an adaptive Filter

The adaptive filter is used to construct a linear system between two sets of correlated digital signals. The working principle of the method is that an FIR filter H (z) is constructed to filter an input signal x (n); and comparing the output result y (n) of the filter with the expected signal d (n), and adjusting the coefficient k of the FIR filter according to the difference e (n) of the output result y (n) and the expected signal d (n), so that the difference e (n) is minimum after a period of filtering. And according to causal relation of signals, y _B Can be used as input signal, y _A Can be used as the desired signal. When the difference e (n) reaches the system accuracy requirement<0.003 mS/cm), we can consider that the current adaptive filter is locked and successfully constructs a linear system H(s), and can calculate the k value for y from this linear system _B And compensating. The self-adaptive filter has a self-adaptive coefficient adjusting function, and is particularly suitable for describing models and parameters of H(s), k and the like which are slowly transformed along with time in a sensor.

Wherein the H(s) is derived from a laplace transform, which is mainly used for analyzing continuous time signal processing. H (Z) is derived from the Z-transform, which is an important tool for analyzing the problem of linear time-invariant discrete-time systems. I.e. the laplace transform is dedicated to analysing analog signals, the Z transform is dedicated to analysing digital signals. FIR is a digital filter, which performs arithmetic processing on discrete digital signals, but cannot directly process continuous signals, and is therefore denoted by H (z).

Step 3 Compensation of Sensors A and B

Obtaining a compensated sensor B signal y according to the k value obtained by the adaptive filtering _B-fix If the adaptive filter is not locked, because the drift of sensor B is slowIn the process, the drift ratio does not change greatly in a short time, so that the last calculated drift ratio can be used for calculation.

The compensated sensor B signal y is then used _B-fix And sensor A signal y _A The cross correlation analysis is a concept in signal analysis, and represents the degree of correlation between two time sequences, generally, the position where the correlation between two signals is largest corresponds to the delay between two signals, and after the delay of sensor a is obtained, the phase compensation of sensor a is performed. In order to reduce the calculation amount of the correlation, the center of gravity of the finite impulse response of the adaptive filter is calculated, a window for calculating the correlation between the sensors A and B is taken around the center of gravity to calculate the correlation, and the delay between the two signals is obtained by the method of the correlation calculation method _A Performing translation to obtain a value y close to the original signal x _A-fix 。

Step 4 comparing and outputting data

The compensated seven-electrode sensor A has the advantage of smaller noise, and under the condition that the measured values of the two compensated sensors are similar, the data y of the seven-electrode sensor A should be calculated _A-fix And outputting. However, due to the characteristics of the seven-electrode sensor, such as delay and low pass, the high frequency amplitude of the sensor is attenuated when the environmental signal changes rapidly, and the two sensor measurement values after compensation have larger deviation, we should switch to the data y of the four-electrode sensor B after compensation _B-fix And outputting. The original standard of data switching is the difference value of the measured values of the two sensors after compensation, and we can also further study the switching method by combining the actual data, and even add a machine learning and judging method.

In order to make the above features and effects of the present invention more clearly understood, the following specific examples are given with reference to the accompanying drawings.

In practice, there are many cases where continuous measurement of data is required for a longer period of time, however, the measurement environment changes considerably over a longer period of time, and the sensor a cannot be regarded as a time-invariant system. Therefore, in practice, the data is processed in a segmented manner, and here we consider only the data in a period of time, and assume that the sensor a and the sensor B are both time-invariant linear systems in this period of time. The data processing procedure by matlab simulation is briefly described as follows, and a flowchart is shown in fig. 2.

First, we set the sampling frequency of the system to 25Hz, pass a gaussian white noise through an IIR low-pass filter with a bandwidth of 0.01Hz, and translate and amplify the noise to form a value transformed between 28 and 32, so as to virtualize the conductivity x (t) in the actual environment, and the effect is shown in fig. 6.

At the same time, we use an IIR low-pass filter with bandwidth of 0.1Hz to virtualize the delay and low-pass characteristics of sensor A to generate y _A . Taking the k value of the sensor B as 0.999, namely 0.1 percent of drift amount, adding Gaussian white noise with standard deviation of 0.001mS/cm, and generating y _B As shown in fig. 7. With partial enlargement of FIG. 7, it can be seen that y _A There is a delay of about 0.06 minutes, i.e. a delay of 3.6 seconds, relative to x because it is mainly affected by the delay characteristics, there is little change in signal shape, and y _B The relative x is attenuated by the drift-affected amplitude and is noisy, as in fig. 8.

Next, we construct an LMS adaptive filter of order 2000, step 0.0000001, to y _A To be the desired signal, y _B For the input signal, adaptive filtering is performed and the filter dc gain (i.e. the sum of all coefficients) after each filtering is recorded. Filter output values y (n) and y _A Comparison of the values as shown in FIG. 9, it can be seen that after 0.4 minutes, the value of y (n) is substantially the same as y _A And overlap. And it can also be seen from the value of e (n) of fig. 10 that after 0.3033 minutes the output y (n) of the adaptive filter is equal to the desired signal y _A The error of (2) is already less than 0.002m/cm and the lock-in state is entered.

At this time, as shown in fig. 11, it can be seen that the dc gain of the adaptive filter is always 1.001 after the filter is lockedNear jitter (1.001=1/0.999, i.e. 1/k), but the jitter value is relatively large. From the characteristic of the sensor B, the k value is relatively stable in a short time, so that it is not necessary to directly correct the k value by using the gain after each adaptive filtering, but the average value in a period of time can be used, or the gain of the filter can be subjected to low-pass filtering, and the correction is performed after the jitter error is eliminated. For the data in the graph we will average the average value from 1/k after the 1 st minute every 10 minutes, as can be seen in FIG. 12, y after compensation _B The drift amount of the (C) is reduced to 0.005% from the original 0.1%, and the drift amount is reduced to about 0.0015mS/cm, so that the target precision can be achieved. This error can be further reduced if a larger time window is used.

While gain compensation for the four-electrode sensor is completed, we can also estimate and compensate for the delay of the 7-electrode sensor based on the coefficients of the ADP filter. Wherein the final state adaptive filter coefficients (i.e. the impulse response of the linear system) are shown in fig. 13. The main response is concentrated within 10 seconds, and the gravity center position is located at about 2.6 seconds. We match y with 1000 sampling points (40 seconds of data) within the range of 0-5.2 seconds with the center of gravity as the center _A And y is _B And y when the cross-correlation coefficient value is maximum _A As the response delay of its sensor, for y _A And performing time shift compensation. Calculated y _A The delay was 81 samples, i.e. 3.24 seconds.

Compensated y _A 、y _B Comparison of the conductivity x with the real environment As shown in FIG. 14, the compensated curve y _{A_fix} And y is _{B_fix} There is a greater improvement than before compensation, which in the figure is substantially coincident with x. Further partial enlargement, as can be seen in FIG. 15, y after time compensation _{A_fix} Almost completely coincides with the x signal, and y _{B_fix} Also perturbed around x, indicating that both the time delays and gains of sensor a and sensor B are well compensated, respectively.

The above embodiments are only application cases of the present invention, and in the actual use process, related personnel may make various modifications and changes within the scope of the technical idea of the present invention, for example, select different adaptive filters to perform system identification, change the output selection policy in the data fusion portion, and so on.

The following is a system example corresponding to the above method example, and this embodiment mode may be implemented in cooperation with the above embodiment mode. The related technical details mentioned in the above embodiments are still valid in this embodiment, and in order to reduce repetition, they are not repeated here. Accordingly, the related technical details mentioned in the present embodiment can also be applied to the above-described embodiments.

Claims

1. A conductivity detection method based on a dual conductivity sensor, comprising:

and 3, circularly executing the step 1 and the step 2 until the difference value is converged, storing the current adaptive filter, updating the drift ratio according to the coefficient of the current adaptive filter to obtain a calibration ratio, and detecting the conductivity of the liquid to be detected by adopting the open conductivity sensor and the calibration ratio to obtain a third conductivity.

2. The conductivity detection method based on a dual conductivity sensor according to claim 1, wherein the step 1 comprises:

3. The dual conductivity sensor based conductivity detection method according to claim 1, further comprising:

and 4, cross-correlating the third conductivity with the first conductivity to obtain the delay of the self-brushing conductivity sensor relative to the open conductivity sensor so as to perform phase compensation on the first conductivity to obtain a fourth conductivity.

4. The dual conductivity sensor based conductivity detection method according to claim 3, further comprising:

5. The conductivity detection method based on dual conductivity sensor according to claim 1, wherein the self-brushing type conductivity sensor and the open type conductivity sensor have a plurality of electrodes for measuring conductivity, respectively.

6. A conductivity detection system based on dual conductivity sensors, comprising:

7. The conductivity detection system based on dual conductivity sensors according to claim 6, wherein the module 1 comprises:

8. The dual conductivity sensor based conductivity detection system according to claim 6, further comprising:

9. The dual conductivity sensor based conductivity detection system according to claim 8, further comprising: and a module 5, configured to determine whether the difference between the fourth conductivity and the third conductivity is greater than a threshold value, and if so, use the third conductivity as a measurement result of the liquid to be measured, otherwise use the fourth conductivity as a measurement result of the liquid to be measured.

10. The dual conductivity sensor based conductivity sensing system according to claim 6, wherein the self-brushing conductivity sensor and the open conductivity sensor have a plurality of electrodes for measuring conductivity, respectively.