JPH0238995B2 - TOTSUHENNOIZUNOJOKYOHOSHIKI - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a method for removing sudden noise, and is useful for signal processing of electromagnetic flowmeter converters.

In physical quantity measuring devices, various noises are superimposed on input signals from sensors, so it is necessary to perform signal processing to remove noise and output signals. Among various types of noise, noise caused by sudden changes in DC potential, that is, sudden change noise, is difficult to distinguish from changes in physical quantities, and therefore sufficient noise removal is not necessarily performed. This will be explained regarding an electromagnetic flowmeter.

FIG. 1 shows an electromagnetic flowmeter in which a converter is constructed using a microcomputer. In Figure 1,
1 is an excitation coil, 2 is a measurement pipe, and the excitation coil 1 is designed to generate a magnetic field that is perpendicular to the fluid flow direction in the pipe 2 and the mounting direction of the electrodes 3a and 3b in the pipe 2. It is located. 4 is a differential amplifier with high input impedance, and 5 is an analog-to-digital converter (hereinafter referred to as
(abbreviated as ADC), 6 is a control calculation unit (hereinafter referred to as
7 is a memory, 8 is an output port (I/O), 9 is a digital-analog converter (hereinafter abbreviated as DAC), 10 is a constant current source, 11 is the output of a differential amplifier, 12 is a bus line.

Each switch from SWa to SWd in Figure 1 is
Opening/closing is controlled via the output port 8 based on commands from the CPU 6. Since the excitation coil 1 is connected to the constant current source 10 via these switches SWa to SWd, the current flowing through the excitation coil 1 is
The commands from the CPU 6 result in a waveform as shown in FIG. 2a, for example.

The output 11 of the differential amplifier 4 generally includes a signal proportional to the flow rate (same as the current waveform in Figure 2a), magnetic flux change noise (Figure 2b), and DC noise generated by electrochemical factors between the electrodes ( Figure 2c) Commercial frequency noise caused by commercial power supply (2nd
Figure d) was thought to have been mixed. Conventionally, the amplifier output 11 is sampled at timings such as A ₁ , B ₁ , C ₁ , D ₁ . . . shown in Fig. 2a, and the above and noise components are removed by time-series data calculation. A method has been considered. In other words, focusing on the fact that the noise components of ~ have regularity, the CPU 6 reads the amplifier output 11 at each timing via the ADC 5 and the bus line 12 and stores it in the memory 7, for example, using the formula (1) Sn=a・V(An)+b・V(Bn) +c・V(Cn)+d・V(Dn) ...Equation (1) However, a to d are constants, and V(An) to V(Dn) are the data at each timing. Perform calculations. If the constants a to d in equation (1) are chosen to be appropriate values determined based on the regularity of the noise, the calculation result is
Sn is only for the signal.

However, recently, as the excitation frequency has been lowered and measurement accuracy has improved, sudden changes in direct current noise due to the effects of slurry in the fluid, which had not received much attention in the past, have become a problem. This kind of sudden noise has a short pulse shape of several milliseconds to one second, but its amplitude is several times to several tens of times that of a normal signal. FIG. 2e shows an example of the waveform of sudden noise. Because of this property, the above-described noise removal method was not effective against sudden noise.

Therefore, a sudden noise removal method called a so-called window processing method has been considered (Japanese Patent Application No. 57-80130 (Japanese Unexamined Patent Publication No. 58-197597): Title of invention: ``Sudden noise removal method''). In other words, a new input signal value S is compared with a time-series average value M of past input signal values,
In the case of an input signal value of |M−S|>α (where α is a constant), it is assumed that sudden noise is superimposed, and instead of this input signal value
is a constant), that is, limit calculation is performed, and after the input signal value of M−S>α continues for a certain number of times or for a certain period of time, or after S−M>α
After the input signal value continues for a certain number of times or for a certain period of time, the above limit calculation is stopped and the output value is calculated using the input signal value S itself.

The idea behind this method of removing sudden noise is that sudden noise is short-lived, but if a signal such as a flow rate suddenly changes, the change continues for a certain amount of time. For example, in the case of an electromagnetic flowmeter, if the input signal suddenly changes, the input signal is monitored for a certain period of time, say 1 second, and if the input signal remains unchanged during that time, then this change is not due to noise and is due to the flow rate. It is determined that this is due to a sudden change in the value, and the limit calculation operation of the window processing method is canceled. Therefore, even if the flow rate changes suddenly due to the opening and closing of a valve in a plant, for example, it will not be mistaken as sudden noise. In other words,
Pulse-like sudden noise can be effectively removed without impairing the response speed to sudden changes in the signal component.

By the way, the above-mentioned "sudden noise removal method"
(Patent Application No. 57-80130), the following problems remained. In other words, large random noise or 1/f noise (noise equivalent to flow noise in an electromagnetic flowmeter) that occurs frequently and continuously and has a large value.
If the noise is superimposed on the signal and the noise occurs continuously for a relatively long time, the limit calculation operation may be stopped even though the noise is superimposed. In this case, in the conventional method, it took a relatively long time from when the limit calculation operation was stopped until it was restarted. This is |M−S|<α after |M
This is because the limiting operation is started when the condition -S|>α is satisfied. Therefore, (1) If noise continues even after the limit calculation operation has stopped, spike-like output fluctuations will occur due to the noise between the time the limit calculation operation stops and the time it restarts.

(2) The output before the limit calculation operation is stopped and the output after it is restarted are significantly shifted.

SUMMARY OF THE INVENTION In view of the above-mentioned circumstances, it is an object of the present invention to provide a sudden noise removal method in which limit calculation is promptly restarted. The present invention, which achieves this object, performs limit calculation under the condition that, after the limit calculation operation is stopped, the polarity of the deviation between the average value M and the new input signal S is reversed, in other words, M-S>α. Immediately after stopping, S-M>0 or S-M>
After stopping the limit calculation under the condition α, M−S>
The basis of the technical idea is that the limit calculation is restarted even if the limit becomes 0. This is because continuous random noise or 1/f
Whereas noise reverses its polarity in a short period of time,
This is because we focused on the fact that the direction of change (polarity) of a change in a physical quantity (flow velocity in the case of an electromagnetic flowmeter) continues for a certain period of time.

Hereinafter, a case where the present invention is applied to an electromagnetic flowmeter will be described.

(i) In FIGS. 1 and 2, normally, the flow rate signal value is calculated using four or less data of signals V (An) to V (Dn) during one cycle.
For example, if we calculate the signal value Sn using equation (1),
The time-series signal values Sn- ₁ , Sn- ₂ ...S ₁ are stored in the memory 7, M=Sn+Sn- ₁ +...+S ₁ /n...The CPU 6 calculates the average value M using the formula (2), and outputs the current from the DAC 9. or output as voltage.

(ii) CPU6 is a signal value calculated every cycle
Compare Sn and the average value M immediately before it, |M-Sn|>α...Formula (3) However, if α is a constant and is set to, for example, 5% of the set span, then S′n= M+β...Equation (4) However, β is a constant, for example, 5% of the set value
value.

The average value M' is calculated as follows: M'=S'n+Sn- ₁ +...+S ₁ /n = M+β+Sn- ₁ +...+S ₁ /n (5), and the current and voltage are output. That is, when a sudden noise as shown in FIG. 2e occurs, the signal value Sn in that cycle is discarded as an abnormal value, and the average value M up to the previous cycle is used. This eliminates the effects of sudden noise.

(iii) On the other hand, if formula (3) holds true, determine whether M-S>α...Equation (6) or S-M>α...Equation (7), and then calculate Equation (6). and Equation (7). Check the consecutive number of times (also continuous time) for each state.

If formula (6) or formula (7) holds true a certain number of times or for a certain period of time, the process in (ii) above is canceled and only the process in (i) is performed, regardless of the magnitude of the signal value. . This prevents the response speed from becoming slow due to the process (ii) when the flow rate changes suddenly due to opening and closing of a valve, etc.

(iv) Furthermore, on the condition that equation (7) holds true continuously, the equation
Immediately after the limit calculation shown in (5) is canceled, the limit calculation is performed if a signal such as M-Sn + ₁ > 0...Equation (8) is input, or if Equation (6) is continuously satisfied. Immediately after the release, if a signal such as Sn+ ₁ -M>0 (Equation (9)) is input, the limit calculation in (ii) is immediately restarted. In other words, the limit calculation starts even when the polarity of the deviation between the average value M and the input signal Sn+ ₁ is reversed after the limit calculation is canceled. At this time, the time until the limit calculation is restarted on the condition that the polarity of the deviation is reversed is calculated using equation (3) (|M
−Sn|>α) is shorter than the time required to start limit calculation on the condition that (the reason will be described later). Therefore, if random noise or the like continues to be superimposed even after the limit calculation is canceled in (iii), the limit calculation is immediately performed and the noise is removed. As a result, spike-like output fluctuations occur less frequently, and even if output fluctuations occur, the amount of the fluctuations is small. Furthermore, the amount of output shift between before the limit calculation is canceled and after it is restarted becomes smaller, and the conventional problem is solved. Note that the limit calculation restarted on the condition that Equations (8) and (9) hold, will be restarted on the condition that M-S>0...Equation (10) or S-M>0...Equation (11) continues. It will be canceled. After the release, the process returns to (i)→(ii)→(iii)→(iv).

Here, we will compare and examine the time required for restarting the limit calculation between the case where the polarity of the deviation between M and S is reversed and the case where the condition is that Equation (3) is satisfied. For example, when M-S>α changes to S-M>α (assuming that random noise etc. is continuous at this time), |M-S| becomes |M-S|>α → |M -S｜＜α→｜M−S｜
>α (resuming limit calculation), while M
The polarity of the deviation between and S changes from M-S>0 to S-M>0 (resuming limit calculation). Therefore, restarting limit calculation by reversing the polarity takes less time than starting limit calculation by establishing equation (3).

(v) On the other hand, when there is no reversal of the deviation polarity of M and S and |M-Sn|<α...Equation (12) is satisfied, the sudden change in flow rate has ended, so (iii) Stop processing and try again (i)→
Return to the process of (ii)→(iii).

The above is an explanation of the present invention based on Figures 1 and 2, and for convenience, the constant current three-level excitation shown in Figure 2a is shown as the excitation waveform, but this method can be applied regardless of the excitation method of the electromagnetic flowmeter. . In short, it can be applied to any process in which signals from sensors such as electrodes are processed to obtain an output proportional to a physical quantity such as a flow rate.

Next, FIG. 3 shows an example of an electromagnetic flowmeter that implements the method of the present invention without using a microcomputer. In Fig. 3, 13 is a synchronous rectifier circuit, and this synchronous rectifier circuit 13 and an excitation switching switch are connected.
SWa to SWd are controlled by a timing circuit 14. The output b of the synchronous rectifier circuit 13 normally becomes the flow rate output signal a via the switch _SW1 . On the other hand, the two comparators 15a and 15b and the OR gate 1
5c, if |signal a−signal b| is larger than a predetermined window value α, the output c of one of the comparators and the output d of the OR gate 15c are set to “H” (high). level)
The switch is activated by the output of the AND gate 18.
SW ₁ is turned off and the output signal a is sent to capacitor 19.
is held. In this case, if it is a sudden noise, the outputs c and d of the window comparator 15 immediately become "L" (low level), the switch SW ₁ is turned on again, and the output b of the synchronous rectifier circuit is output as usual. . However, for example, if the condition a-b>α continues for a long time, one of the counters 16a is activated by the output c of the comparator 15a to count the pulses from the timing circuit 14, and after a certain period of time, the carry e is Flip-flop 1 to be output
7 is set and switch SW ₁ is closed. In other words, it is determined that the noise is not sudden change noise, and sudden changes in flow rate can be measured. In the case of ba>.alpha., if this continues for a long time, the other counter 16b starts to operate, the flip-flop 17 is similarly set, and the switch _SW1 is closed. Furthermore, for example, if the condition a-b>α continues for a long time and the condition becomes b-a>0 immediately after the switch SW ₁ is closed due to the carry e of the counter 16a, the comparator 1
The output f of the flip-flop 5d becomes "L", and this output f reaches the reset terminal R of the flip-flop 17 via the inverter 15e, the AND gate 20a, and the OR gate 21. Therefore, flip-flop 1
7 is reset and switch _SW1 is turned off.
Similarly, the state where b-a>α is a long-duration switch
Immediately after SW ₁ is closed, when a-b > 0, the output f of the comparator 15d becomes "H", and this signal reaches the reset terminal R via the AND gate 20b and the OR gate 21, and as a result, the switch SW ₁
is turned off.

As described above, according to the present invention, sudden noise can be removed without reducing the response speed to changes in physical quantities.

[Brief explanation of drawings]

1 and 3 are block diagrams of the electromagnetic flowmeter, and FIG. 2 is a waveform diagram for explaining the operation. In the drawing, 1 is an excitation coil, 2 is a measurement pipe, and 3
a and 3b are electrodes, 5 is an analog-to-digital converter, 6 is a control arithmetic unit, 7 is a memory, 13 is a synchronous rectifier circuit, 14 is a timing circuit, 15 is a window comparator, 15a, 15b and 15d are comparators,
16a and 16b are counters, 17 is a flip
Flop, 19 is a hold capacitor, 20a,
20b is AND gate, 21 is OR gate, SW ₁
is a switch.

Claims (1)

[Claims] 1 In a converter that calculates an output value related to a physical quantity from time-series input signal values, the difference between the average value M of input signal values within a certain time-series preceding period and the new input signal value S If the absolute value |M-S| is larger than the predetermined value α, that is, |M-S|>α, M+β is used instead of S as the new input signal value.
(However, β is a constant) to calculate the output value, and after the input signal value of M-S>α continues for a certain number of times or for a certain period of time, or S-M
After the input signal value >α continues for a certain number of times or for a certain period of time, the above limit calculation is stopped and the output value is calculated using the input signal value S itself, and M−S>α
Immediately after stopping limit calculation on the condition that input signals such as Immediately after M-S
A sudden noise removal method characterized in that limit calculation is restarted when an input signal >0 is input.