Magnetic flowmeter with improved processing

Abstract

A magnetic flowmeter includes at least one coil configured to generate a magnetic field within a process fluid flow. A pair of electrodes is configured to detect an electromotive force within the process fluid flow in response to the magnetic field. A measurement circuit is operably coupled to the pair of electrodes and configured to provide an indication of the detected electromotive force. A processor is coupled to the measurement circuit and configured to receive the indication of the detected electromotive force. The processor is configured to obtain a sequence of indications of the detected electromotive force within a time interval and generate a plurality of electromotive force sample sets by selecting non-continuous indications of the detected electromotive force. Each of the plurality of electromotive force sample sets is processed by a signal processing engine to provide an electromotive force sample output. The processor is configured to combine the electromotive force sample outputs from each electromotive force sample set to generate a process fluid flow output. On the other hand, the processor is configured to obtain a sequence of indications of the detected electromotive force within a time interval and generate a plurality of data streams from the sequence of indications of the detected electromotive force. The processor is also configured to perform signal processing on each data stream to generate a plurality of stream outputs and generate a process fluid flow output based on a combination of the plurality of data stream outputs. In yet another aspect, the processor is configured to calculate a median absolute deviation for each data stream to provide each data stream output.

Description

Background Art

[0001] A magnetic flowmeter (or magnetic meter or direct reading frequency meter) measures flow by applying Faraday's law, an electromagnetic effect. A magnetic flowmeter excites one or more coils by passing an excitation current through a field winding, which generates a magnetic field across an electrically isolated conductive process fluid stream. An electromotive force (EMF) is generated by the flowing process fluid passing through the magnetic field. This induced voltage (potential) induced both across the fluid and relative to the rest of the process fluid can be easily measured by one or more conductive electrodes in contact with the flowing process fluid. The volume flow rate is proportional to the cross-sectional area of ​​the flow tube and the flow velocity. The flow velocity is proportional to the electrode voltage potential (EV), which is proportional to the induced magnetic field strength (B). Assuming that the induced magnetic field strength is proportional to the applied magnetic field (H), the magnetic field is directly related to the magnitude of the excitation current. Therefore, a direct correlation is provided between the measured electrode voltage potential and the indicated volume flow rate. Technical Field

[0002] Magnetic flowmeters are used in a variety of conductive and semi-conductive fluid flow measurement environments. In particular, the flow of water-based fluids, ionic solutions, and other conductive fluids can be measured using magnetic flowmeters. In addition, magnetic flowmeters are often used with fluids that may contain solids, such as pulp used in papermaking processes. Therefore, magnetic flowmeters can be found in water treatment facilities, beverage and sanitary food production, chemical processing, high-purity pharmaceutical manufacturing, and hazardous and corrosive fluid processing facilities. However, some environments are more susceptible to signal noise. When using a magnetic flowmeter in such a noisy environment, a magnetic flowmeter that provides a better response to signal noise will improve the accuracy of the flow output. Summary of the invention

[0003] A magnetic flowmeter includes at least one coil configured to generate a magnetic field within a process fluid flow. A pair of electrodes is configured to detect an electromotive force within the process fluid flow in response to the magnetic field. A measurement circuit is operably coupled to the pair of electrodes and configured to provide an indication of the detected electromotive force. A processor is coupled to the measurement circuit and configured to receive the indication of the detected electromotive force. The processor is configured to obtain a sequence of indications of the detected electromotive force within a time interval and generate a plurality of electromotive force sample sets by selecting non-continuous indications of the detected electromotive force. Each of the plurality of electromotive force sample sets is processed by a signal processing engine to provide an electromotive force sample output. The processor is configured to combine the electromotive force sample outputs from each electromotive force sample set to generate a process fluid flow output. On the other hand, the processor is configured to obtain a sequence of indications of the detected electromotive force within a time interval and generate a plurality of data streams from the sequence of indications of the detected electromotive force. The processor is also configured to perform signal processing on each data stream to generate a plurality of stream outputs and generate a process fluid flow output based on a combination of the plurality of data stream outputs. In another aspect, the processor is configured to calculate a median absolute deviation for each data stream to provide each data stream output. BRIEF DESCRIPTION OF THE DRAWINGS

[0004] Figure 1 is a schematic diagram of an environment in which a magnetic flow meter according to an embodiment of the present invention is useful.

[0005] Figure 2 is a block diagram of a magnetic flowmeter according to an embodiment of the present invention.

[0006] Figure 3A and Figure 3B is a graph illustrating the noise experienced by the magnetic flowmeter at simulated flow rates.

[0007] Figure 4 is a block diagram of improved magnetic flowmeter signal processing according to an embodiment of the present invention.

[0008] Figure 5 is a flow chart of a method for processing raw flow data in a magnetic flow meter according to an embodiment of the present invention.

[0009] Figure 6 is a graph illustrating noise limits expressed as a percentage of flow, comparing the output of a single signal processing engine to the output of a time-mixed signal processing engine, according to an embodiment of the present invention.

[0010] Figure 7 is a graph illustrating the improvement of the signal processing engine with noise from 10% pulp stock.

[0011] Figure 8is a graph illustrating the raw flow seen in a typical magnetic flow meter installation.

[0012] Fig. 9 is a comparison of the rolling MAD (with a 0.5 second window) and the rolling standard deviation of a window with the same time duration.

[0013] Fig.10 is a flow chart of a method of processing electrode EMF samples to provide a flow output in a magnetic flow meter according to an embodiment of the present invention. DETAILED DESCRIPTION

[0014] Figure 1 A typical environment 100 for a magnetic flow meter 102 is illustrated. The magnetic flow meter 102 is coupled to a process pipeline (diagrammatically illustrated as line 104), which in turn is coupled to a control valve 112. The magnetic flow meter 102 is configured to provide a flow output regarding a process fluid flowing through the pipeline 104 in the process. Examples of such fluids include slurries and liquids in chemical, pulp, pharmaceutical, and other fluid processing plants.

[0015] The magnetic flow meter 102 includes an electronics housing 120 connected to the flow tube 108. The magnetic flow meter 102 output is configured to be transmitted to a controller or indicator over a relatively long distance via a process communication connection 106. In a typical process plant, the communication connection 106 can employ a digital communication protocol, an analog communication signal, or a combination thereof. The same or additional process information can be provided via wireless communication, pulse width or frequency output, or discrete input / output (DI / DO). The system controller 110 can display flow information for an operator and provide control signals through the process communication connection 106 to control the process using a control valve (such as valve 112).

[0016] Figure 2 1 is a block diagram of a magnetic flowmeter 102, to which embodiments of the present invention are particularly applicable. The magnetic flowmeter 102 measures the flow of an electrically conductive process fluid flowing through a flow tube 108. A coil 122 is configured to apply an external magnetic field in the fluid flow in response to an applied excitation current from a coil driver 130. An EMF sensor (electrode) 124 is electrically coupled to the fluid flow and provides an EMF signal output 134 related to the EMF generated in the fluid flow due to the applied magnetic field, flow velocity, and noise to an amplifier 132. An analog-to-digital converter 142 provides a digitized EMF signal to a microprocessor 148 of a flow meter electronics 143.

[0017] The microprocessor 148 may be configured to provide digital signal processing functionality related to the EMF output 134 by hardware, software, or a combination thereof, in order to provide an output 152 related to the flow rate. In addition, as will be described in more detail below, the signal processing can provide improved noise suppression. The microprocessor 148 may include or be coupled to a memory 150 containing instructions that, when executed by the microprocessor 148, provide process fluid flow rate output calculations and improved noise reduction according to embodiments described herein.

[0018] Microprocessor 148 calculates the fluid flow rate through flow tube 108 based on the relationship between EMF output 134 and flow rate as described in the application of Faraday's law:

[0019]

[0020] Where E may be a signal output 152 related to the EMF output 134, V is the velocity of the fluid, D is the diameter of the flow tube 108, B is the strength of the induced magnetic field in the process fluid, and k is a proportionality constant. The microprocessor 148 calculates the flow rate of the process fluid using the velocity and the measured magnetic field or coil current according to known techniques. A digital-to-analog converter 158 is coupled to the microprocessor 148 of the flow meter electronics 143 and generates an analog transmitter output 160 coupled to the communication connection 106. The digital communication circuit 162 may generate a digital transmitter output 164. The analog output 160 and / or the digital output 164 can be coupled to a process controller or monitor as desired.

[0021] Magnetic flowmeters are often used in applications with high solids levels because the flowmeter does not require any type of flow obstruction to provide its measurement. When such solids strike the measuring electrodes of the magnetic flowmeter, they generate a significant amount of electrical noise. This noise can distort the measurement waveform and cause significant errors in the flow output. In extreme cases, the error can be as high as 100% of the actual flow value. Additionally, these problems are exacerbated by the fact that modern designs are subject to reduced permissible limits on coil drive current.

[0022] It has been determined that much of the noise experienced by magnetic flow meters is relatively nonlinear. Typically, there will be significant spikes in the raw flow measurement EMF measured by the analog-to-digital converter 142 across the electrodes 124 that may be many times the actual flow value. Since the noise spikes are not symmetrical about the average value, the damping filter will not be able to properly filter them.

[0023] Figure 3A and 3B is a graph illustrating the noise experienced by a magnetic flow meter with the simulated flow rate fixed at 3 ft / sec. Figure 3AA significant negative spike is shown, followed by a significant positive spike at approximately 4 seconds. The negative peak is over -40 ft / sec, while the positive peak is over +40 ft / sec. Figure 3B The corresponding damping flow output shown in shows a clear deformation of the damping flow value at about 4 seconds.

[0024] According to an embodiment of the present invention, a new signal processing technique for a magnetic flowmeter is provided. This signal processing can be performed within the microprocessor of the magnetic flowmeter, or can be performed in any suitable logic or circuit coupled to the microprocessor. Typically, the raw flow data stream (separate EMF measurements at both ends of the electrode at the maximum sampling rate) is used to create separate data streams that are randomly or at least non-sequentially mixed over a fixed time period. When the flow data is mixed in time, it will separate each sample from its time neighborhood. This separation in time helps to more effectively identify spikes so that they can be correctly removed. The data streams are then recombined to provide an overall average, which will help minimize flow errors.

[0025] Figure 4 2 is a block diagram of improved magnetic flowmeter signal processing according to an embodiment of the present invention. It can be seen that the raw flow measurement value (indicated by reference numeral 200) is provided to the time mixer 202 and is also provided to the signal processing engine 226. The time mixer 202 buffers the samples within a certain scanning time period (e.g., 0.250 seconds), and when the buffer is filled, the time mixer 202 creates multiple data streams of flow data. In one example, the time mixer 202 creates five different flow data streams, each of which is randomly organized. For example, if 100 samples are collected during 0.25 seconds, each stream will see all samples due to time mixing, but in a different order than other streams. Those skilled in the art will also understand that embodiments of the present invention can be practiced using non-random techniques in which the buffering site selection ensures that non-continuous or continuous sites are used. However, in a preferred embodiment, random buffering sites are used.

[0026] like Figure 4 As shown in FIG. 1 , the time mixer 202 outputs a plurality of new traffic streams 206, 208, 210, 212, and 214. Each of these traffic streams is provided as input to a corresponding signal processing engine 216, 218, 220, 222, and 224. Each corresponding signal processing engine is used to classify the raw traffic samples provided by the time mixer 202 and provide a moving average. Each signal processing engine evaluates the traffic samples based on a continuously updated limit value around the moving average.

[0027] In one embodiment, these limits are continuously updated based on the median absolute deviation (MAD). MAD is a mathematically known calculation that is robust against outliers. It is generally believed that MAD is more resistant to outliers in a data set than standard deviation. In standard deviation, the distance from the mean is squared, and in median absolute deviation, the deviation of a small number of outliers is insignificant. MAD is calculated by first calculating the median of a set of samples. Then the deviation of each sample from the median is calculated. Finally, the median of the sample deviations is calculated to provide MAD. For example, consider the data (2, 2, 4, 4, 8, 12, 18). Its median value is 4. The absolute deviations relative to 4 are (2, 2, 0, 0, 4, 8, 14), and their median value is 2 (because the absolute deviations after sorting are (0, 0, 2, 2, 4, 8, 14)). Therefore, the median absolute deviation of this data is 2.

[0028] MAD is used to dynamically measure the amount of noise in the stream and adjust the limit accordingly. As described above, MAD is similar to standard deviation, but performs well in processing nonlinear noise spikes. Each signal processing engine 216, 218, 220, 222 and 224 provides its output to a corresponding damping filter, which can average the continuous output from each signal processing engine to provide a time-damped output. Damping can also include any other suitable technology or calculation. For example, one or all of the damping filters can include a low-pass filter.

[0029] Although Figure 4 Each of the signal processing engines 216, 218, 220, 222, and 224 is illustrated as providing its output to a corresponding individual damping filter, but it is also expressly contemplated that the outputs from the signal processing engines may be averaged together and then fed to the signal damping filters. Furthermore, while the signal processing settings for each of the signal processing engines may all be the same, they may be different if desired. The different signal processing settings will minimize the impact of having suboptimal signal processing settings. Additionally, note that the values ​​of each traffic flow from the corresponding signal processing engine may be sorted and the middle value among the flows may be used. This will further eliminate flows that react poorly to noise. Figure 4 As can be seen in Figure 2, the outputs of the various damping filters are ultimately fed to a block 230 which averages the filter values ​​of the flow output and provides a modified flow output value 232. Alternatively, the outputs from the various damping filters may be combined in a different manner rather than simply averaging them together. Certain statistical measures (e.g., standard deviation, peak-to-peak range, and / or interquartile range) may be used to exclude weaker signal processing engines or to assign more weight to stronger signal processing engines.

[0030] Figure 5 3 is a flow chart of a method for processing raw flow data in a magnetic flow meter according to an embodiment of the present invention. The method 300 starts at box 302, where a value labeled as a clipping noise limit is set equal to a noise factor multiplied by a median absolute deviation (MAD). At box 304, a quantity "outlier noise limit" is set equal to an outlier factor multiplied by the median absolute deviation. Next, at box 306, a difference is generated by subtracting a moving average from a new input. At box 308, it is determined whether the new input is within the clipping noise limit set in box 302. If so, control is transferred to box 310, where the new value is set equal to the new input and the original value is used. Then, control is transferred down to box 312, where the new value is added to the moving average. However, if the new input is not within the clipping noise limit at box 308, control is transferred to box 314, where it is determined whether the time period for which the new input exceeds the clipping noise limit is greater than the maximum spike width time. This maximum peak width time may be set by the manufacturer or may be a user selectable setting in the magnetic flow meter. In any case, if the new input exceeds the clipping noise limit for a time greater than the maximum peak width time, control is again transferred to box 310 where the new value is set equal to the new input and the original value is used where control is transferred to box 312.

[0031] If the time that the new input exceeds the clipping noise limit is not greater than the maximum spike width time at block 314, control passes to block 316, where it is determined whether the new input exceeds the outlier noise limit. If so, control passes to block 318, where the new value is set equal to the moving average, and control passes to block 312. However, if it is determined at block 316 that the new input does not exceed the outlier noise limit, control passes to block 320, where it is determined whether the new input is greater than the clipping noise limit. If so, control passes to block 322, where the new value is set equal to the clipping noise limit, and control passes to block 312. However, if it is determined at block 320 that the new input is not greater than the clipping noise limit, control passes to block 324, where the new input is set equal to the negative clipping noise limit, and control passes to block 312. As described above, at block 312, the new value is added to the moving average, and then at block 326, the moving average is returned.

[0032] about Figure 5 The described method can be summarized as follows. If the new sample is within the clipping noise limit, then the sample can be used as is. However, if the new sample is outside the clipping noise limit but within the outlier limit, then the sample will be clipped at the clipping noise limit. If the sample is outside the outlier limit, then the moving average is used. Although the Figure 5The described clipping is considered hard clipping, but one could also consider using soft clipping, which reduces the gain as samples get further away from the moving average. Another option is to automatically adjust the limit to keep a certain number of samples.

[0033] Figure 6 is a graph illustrating noise limits expressed as a percentage of flow rate comparing the output of a single signal processing engine to the output of a time-mixed signal processing engine according to an embodiment of the present invention. Figure 6 As shown, the flow rate is still set to 3 ft / s. The vertical line shows the relationship between the flow rate and the noise limit setting (expressed as a percentage of the flow rate). Therefore, it can be seen that using time mixing for multiple signal processing engines will greatly reduce the noise signal, thereby significantly improving the noise suppression of the entire flow output.

[0034] Figure 7 is a graph illustrating the improvement of the signal processing engine with noise from 10% pulp stock. The dashed line illustrates the flow measurement using a simple 2 second damping technique, while the solid line illustrates the flow measurement using the time mixer signal processing routine. The standard deviation and peak to peak range of the signal processing are 1 / 18.5 and 1 / 36.3 of that of the straight damped filter, respectively. Figure 7 In Figure 1, the flow rate of the 10% pulp stock is still set at 3 ft / sec. Pulp stock is an example of a flowing medium with significant solids content, which can create noise when such solids impact the electrodes. It can be seen that at approximately 100 seconds, the spike in the 2 second damped flow signal exceeds 10 ft / sec. In contrast, the time mixed signal processing routine remains at 3 ft / sec. This is a significant improvement in the signal processing of the flow meter in such a high solids handling environment.

[0035] Figure 8 is a graph illustrating the raw flow seen in a typical magnetic flow meter installation. Fig. 9 is for Figure 8 The raw flow values ​​shown in , rolling MAD (with 0.5 second window) and rolling standard deviation with the same window time. Comparison of rolling MAD and standard deviation shows that MAD is much more stable than standard deviation for magnetic flowmeter signal processing. For normally distributed noise, the standard deviation should be 1.4826 times the MAD. Fig. 9 The mean of the standard deviation in the flow rate is 10.7 times the peak-to-peak range of approximately 80.05 ft / sec. The range of the MAD is 2.14 ft / sec. In applications where shocks in the flow stream cause spikes in the flow signal, the standard deviation is very poor. As you can see, compared to Fig. 9The two signals shown in , can provide useful information about the properties of the process fluid flow. Therefore, in some embodiments, the median absolute deviation can also be used to detect when solids or chemicals are present. In addition, the median absolute deviation can even be used to correlate the amount of solids for a specific process.

[0036] Fig.10 4 is a flow chart of a method for processing electrode EMF samples to provide flow output in a magnetic flowmeter according to an embodiment of the present invention. Method 400 begins at box 402, where a sequence of EMF samples is obtained from electrode 124. The sequence can be up to a defined time period, such as 0.5 seconds or a variable time period. Next, at box 404, a single EMF sample is stored. In one embodiment, they are stored in a buffer in a manner that maintains their sequence. At box 406, multiple data sets are created from the stored EMF samples, where non-continuous samples are used. In one embodiment, five different data sets are used. However, those skilled in the art will recognize that more or fewer data sets may be used according to the embodiments described herein. The manner in which non-continuous samples are used may be to extract samples from random locations in the buffer, as shown in reference numeral 408, or in any other suitable manner, as shown in reference numeral 410.

[0037] At block 412, each data set is processed to obtain a signal processed output for each data set. The signal processing may employ a median absolute deviation as shown at reference numeral 414, a standard deviation as shown at reference numeral 416, or other suitable processing as shown at reference numeral 418. Finally, at block 420, the different outputs from each processed data set are combined to provide a process flow output value. As described above, this may be a simple average of the various outputs, or may include assigning a weight to each data set based on the particular type of processing performed on that data set, as well as a statistical aspect or parameter of the emf sample.

[0038] Although the above signal processing is generally done independently of flow percentage, it is also contemplated that the signal processing routine may use flow percentage to adjust the noise limit because, to some extent, noise is considered to scale with flow.

[0039] Although the present invention has been described with reference to preferred embodiments, those skilled in the art will recognize that changes may be made in form and detail without departing from the spirit and scope of the present invention. For example, as described above, median absolute deviation is the preferred technique for embodiments of the present invention. However, it is also expressly contemplated that time mixing may be used in conjunction with standard deviation or other statistical or signal processing techniques.

Claims

1. A magnetic flowmeter, comprising: at least one coil configured to generate a magnetic field within a process fluid flow; a pair of electrodes configured to detect an electromotive force within the process fluid flow in response to the magnetic field; a measurement circuit operatively coupled to the pair of electrodes and configured to provide an indication of the detected electromotive force; and a processor coupled to the measurement circuit and configured to receive an indication of the detected EMF, the processor configured to obtain a sequence of indications of the detected EMF within a time interval and to generate a plurality of EMF sample sets by performing a plurality of non-sequential mixing of the sequence of indications of the detected EMF using a time mixer to select non-contiguous indications of the detected EMF, wherein each of the plurality of EMF sample sets is processed by a signal processing engine to provide an EMF sample output, and wherein the processor is configured to combine the EMF sample outputs from each EMF sample set to generate a process fluid flow output.

2. The magnetic flowmeter according to claim 1, wherein: The processor is configured to randomly select discontinuous indications of detected electromotive force.

3. The magnetic flowmeter according to claim 1, wherein: The time mixer includes a buffer configured to receive a sequence of indications of detected electromotive forces.

4. The magnetic flowmeter according to claim 3, wherein: The time mixer is configured to generate each of a plurality of sets of electromotive force samples by selecting content at a random location in the buffer.

5. The magnetic flowmeter according to claim 1, wherein: The signal processing engine includes a plurality of signal processing engines, each signal processing engine being configured to receive a corresponding set of electromotive force samples.

6. The magnetic flowmeter according to claim 5, wherein: The plurality of signal processing engines perform the same signal processing on each corresponding set of electromotive force samples.

7. The magnetic flowmeter according to claim 5, wherein: The plurality of signal processing engines perform different signal processing on each corresponding set of electromotive force samples.

8. The magnetic flowmeter according to claim 1, wherein: The signal processing engine is configured to determine a median absolute deviation for each set of electromotive force samples.

9. The magnetic flowmeter according to claim 1, wherein: The processor is configured to generate the process fluid flow output based on an average of the electromotive force sample outputs.

10. The magnetic flowmeter according to claim 9, wherein: The average is a weighted average.

11. The magnetic flowmeter according to claim 1, wherein: The processor is configured to generate the process fluid flow output by applying at least one low pass filter on the EMF sample output.

12. A magnetic flowmeter, comprising: at least one coil configured to generate a magnetic field within a process fluid flow; a pair of electrodes configured to detect an electromotive force within the process fluid flow in response to the magnetic field; a measurement circuit operatively coupled to the pair of electrodes and configured to provide an indication of the detected electromotive force; and a processor coupled to the measurement circuit and configured to receive the detected indication of the electromotive force, the processor configured to: obtain a sequence of the detected indication of the electromotive force within a time interval; perform a plurality of mixing of the sequence of the detected indication of the electromotive force by using a time mixer to generate a plurality of data streams from the sequence of the detected indication of the electromotive force and perform signal processing on each data stream to generate a plurality of data stream outputs; and generating a process fluid flow output based on a combination of the plurality of data stream outputs.

13. The magnetic flowmeter according to claim 12, wherein: The processor is configured to calculate a median absolute deviation for each data stream to provide a per data stream output.

14. The magnetic flowmeter according to claim 12, wherein: The processor is also configured to provide an indication of solids in the process fluid flow based at least in part on a median absolute deviation.

15. The magnetic flowmeter according to claim 14, wherein: The indication of solids is an indication of the amount of solids in the process fluid stream.

16. A method of providing a process fluid flow output in a magnetic flow meter, the method comprising: obtaining a series of electromotive force measurements from a pair of electrodes of the magnetic flowmeter; performing a plurality of mixing on the series of electromotive force measurement values ​​to randomly distribute the electromotive force measurement values ​​into a plurality of electromotive force sample sets; processing each of the plurality of sets of electromotive force samples to provide a plurality of electromotive force outputs; and The multiple electromotive force outputs are combined to produce a process fluid flow output.

17. The method according to claim 16, wherein: Processing each of the plurality of sets of electromotive force samples includes calculating a median absolute deviation for each set.

18. The method according to claim 17, further comprising: An indication of solids in the process fluid stream is generated based at least in part on the median absolute deviation for each set.

19. The method according to claim 16, further comprising: The series of electromotive force measurements are stored in a buffer of the time mixer.

20. The method according to claim 16, wherein: Processing each of the plurality of sets of electromotive force samples includes applying different signal processing to each of the plurality of sets of electromotive force samples.

21. The method according to claim 16, wherein: Processing each of the plurality of sets of electromotive force samples includes clipping the flow samples to reduce noise.

22. The method according to claim 21, wherein: Clipping involves replacing outliers with the clipped values.

Images