JPH01199146A - Method of detecting gas - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Application of the Invention] The present invention relates to gas detection by a gas sensor, and particularly to noise processing. Further, the present invention can be used for detecting gas using a gas sensor, such as controlling an air conditioner, detecting gas leaks, or preventing gas explosions.

[Prior Art] As is well known, gas sensor signals contain a lot of noise. For example, in the case of a gas leak, a signal due to a temporal increase in gas concentration, such as due to the use of an organic solvent, is noise. Furthermore, in the case of air conditioning control, signals caused by local increases in gas concentration or temporal temperature fluctuations of sensors are noise.

To remove noise, it is known to process the sensor signal with an integrating circuit and extract only the signal that has passed through the integrating circuit (Utility Model Publication No. 57-38,619).By the way, the size and duration of the noise are known. depends on the usage environment and is difficult to process in a regular manner. Therefore, if the duration of noise is detected using some method and gas detection conditions are determined based on this, noise and significant signals can be accurately distinguished. Furthermore, the delay time required to distinguish between noise and significant signals can be kept to a minimum.

[Problem to be solved by the invention] An object of the present invention is to learn the duration of noise and remove noise more efficiently.

[Configuration of the Invention] In the present invention, it is determined whether or not it is noise based on the duration of a sensor signal, and the duration of the noise is detected and stored. Then, gas detection conditions are changed based on the stored duration.

The basic characteristic of noise is its short duration. Therefore, an appropriate upper limit is fixed, and signals with durations shorter than this are treated as noise. In addition to noise, the other six characteristics are low signal strength. Therefore, noise may be determined by taking this feature into consideration. In order to identify noise, for example, an appropriate threshold value may be prepared, and the duration time may be extracted from a comparison between the threshold value and the sensor signal.

The duration of noise can be learned, for example, as follows. Set an appropriate interval and extract the maximum value of the duration of the noise during that interval. If detection is continued for a long time, noise with a length close to the upper limit will occur, so the stored maximum value is decreased each time the section is changed, and is used as the initial value for the next section. Alternatively, instead of measuring the actual duration of the noise, an estimated value for the duration of the noise is prepared. Next, the number of noises that are longer than the estimated value in a certain section is counted, and the estimated value is corrected based on this count value. In this method, the estimated value is corrected so that the frequency of noise that is longer than the estimated value falls within the target range. Of course, if gas detection continues only for a short period of time, there is little possibility that the duration will diverge, and such processing is unnecessary. Once the duration of the noise is learned, it can be used, for example, as follows.

(1) Sampling N sensor signals at intervals ΔT,
A weighting factor an (n=1 to N) is added to this to obtain a cumulative average. Here, if N·ΔT is made sufficiently larger than the duration T of the noise, the influence of the noise will not appear on the arithmetic average. Then, the detection interval ΔT is changed based on the duration of the sampled noise.

(2) Compare the duration of the sensor signal with the duration T of the noise, and extract only the signals whose duration is longer than T.

(3) Store the duration and peak value of the noise.

Detection is performed using this as the lower limit, and signals exceeding either the duration or the peak value are extracted.

(4) Integrate the noise signal during the noise duration T. A signal exceeding this integrated value is determined to be a significant signal.

[Nomenclature] In this specification, the sensor signal is shown as increasing with the gas to be detected. However, depending on the type of sensor, signal processing method, etc., the opposite result may occur. In this case, the meanings of signal increase and decrease may be interpreted in the opposite manner to this specification.

[Example] Example 1 A first example is shown in FIGS. 1 to 5. Each of the embodiments takes as an example the control of air conditioning by detecting a gas with low turbidity.

FIG. 1 schematically shows the relationship between the duration T of a sensor signal and the frequency of the signal. Tl in the figure is the lower limit value for a significant signal, and signals longer than this should be treated as significant signals. This length is determined depending on the application, for example, about 200 minutes for gas leak detection, and about 2 minutes for air conditioning control. On the other hand, the peak length of the noise is at a point lower than this. According to experiments conducted by the inventors, in the case of air conditioning control, the duration of noise is mainly 10 seconds or less.

Further, in the case of gas leak detection, there is a lot of noise for about 1 to 5 minutes. Therefore, a cutoff is provided between the peak of the length of the noise and the lower limit value TI for the length of the significant signal signal, and signals below the cutoff are discarded. The cutoff position is
Correct by learning the duration of the noise. In this way, the noise determination time can be shortened from TI to the cutoff point.

Note that whether it is noise or not can also be determined from the strength of the signal. Therefore, a large signal may be treated as a significant signal regardless of its duration.

In this case, it is sufficient to perform the discrimination based on the duration only for signals that are relatively weak and close to the level for determining whether or not they are noise.

FIG. 2 shows the sampling principle in this embodiment.

N sensor signals sampled at intervals T are fl~
fn, add a weighting factor to this, and average it.

Here, it is assumed that N-ΔT=m-Tmax (1). Note that m>1, and Tma represents the maximum duration of noise. The influence of noise on the average value decreases as m increases, and due to the balance between detection accuracy and detection delay time, m
Decide. Then, by changing ΔT or N according to T maX, the influence of noise can be suppressed. For example, m may be about 3.

FIG. 3 schematically shows an embodiment. This embodiment includes initialization, a sampling subroutine, a noise width learning subroutine, and a T ma
It consists of four blocks for x and xmin change processing. When detection starts, data is initialized. For example, let the initial value of f l-f n be the sensor signal V at that time
Also, the initial value of the additive average value X of r I-f' n is N-V, and the initial value of the minimum value xmin of the additive average value X is N-
Let it be V. Furthermore, an initial value of the sampling interval ΔT is determined. Here, all the weighting factors for f1 to fn are set to l, but this may be changed depending on the value of n. For example, it may be made large for the signal rn near the present, and may be made small for the signal rn in the past.

Next, at every sampling interval ΔT, a transition is made to a sampling subroutine by interrupt processing.

Add a new sensor signal V to fn, and add the original f to f+-fn-+.
, ~fn and modify X by 1/N. Also, compare X and xmin, and set the minimum value of X to xlI+in. Further, by comparing X and xmin, atmospheric contamination is detected.

The learning subroutine for the noise width, that is, the noise duration, is processed separately from the sampling subroutine that is performed intermittently. Prepare an appropriate noise detection level and calculate the duration of the noise by comparing it with this level. That is, a case where a signal higher than the detection level lasts only for a period lower than the upper limit TI is considered to be noise.

Then, the maximum value of the duration is set as T max, and ΔT is changed according to equation (1).

When the interval has elapsed, in order to prevent xmin and ΔT from diverging, these values are changed and the reception is continued to the next interval.

For example, assume that the new T max is 0.8 times the original value and xnin is 1.05 times the original value. These constants can be changed arbitrarily depending on the purpose.

FIG. 4 shows the circuit configuration of the embodiment. In the figure, 2 is a suitable power source, 4 is a gas sensor using a change in resistance value of a metal oxide semiconductor 6, and 8 is a heater thereof. Here, the sensor 4 used had a resistance value change of 5 nOt. In addition to these sensors, there are also catalytic combustion gas sensors that detect gas from the heat of combustion of combustible gas in a catalytic oxidation catalyst, or sensors that use the electromotive force of solid electrolytes such as ZrO or antimonic acid. Any gas sensor can be used, such as.

lO is the load resistance of the sensor 4. Let the voltage V applied to the load resistor 10 be the signal of the sensor 4.

12 is a microcomputer for signal processing, for example 4
A one-chip microcomputer with 8 bits to 8 bits is used. 14 is an A/D converter for A/D converting the sensor signal V, and 16 is an arithmetic and logic unit. 18 is a clock circuit; 20 is a RAM for storing the sensor signal V;

22 is a RAM for storing the sensor signal sequence (f, to fn), and 24 is a RAM for storing X and xmin.

26 is a RAM for storing ΔT and T max. Further, 28 is a timer counter that stores the output signal K, and 30 is a timer counter, which includes a count value C3 for the duration of noise, a count value Cp for the time for determining the sampling interval, and a count value for the time for changing the interval. Count C5. 32 is the load air conditioner,
Use ventilation fans, air purifiers, etc. The load 32 is operated by the contamination detection signal K.

Moving to FIG. 5, the operation of the embodiment will be explained.

When the power is turned on, the device waits for 2 minutes, for example, and initializes the data. For example, the initialization may be performed by setting the sensor signal V at that time to the initial value of the signal string (L-fn), and
Let the initial value of be NV. Also, the initial value of ΔT is ΔT9
, Tff1ax is assumed to have an initial value of N·ΔT/+a. Next, counting is started with C1 set to 0, and counting of C5 is also started. Hereinafter, "starting counting" means starting counting after setting the count value to 0.

Thereafter, sampling processing is performed every time ΔT elapses.

(2) is read, the past signal strings (f, to fn) are shifted to the past side one by one, and V is substituted for the latest signal fn.

Next, the average of these additions is defined as the digital sensor signal X. Note that N has a plurality of values, preferably at least 3 or more. Compare X with the minimum value xmin, and
If it is smaller than in, assign X to xmin. Also X
When X exceeds J-xn+1n (J is the detection factor for air pollution, here 1.3), it is determined to be pollution and the pollution detection signal K is set, and when X is less than J-xmin, the pollution detection signal K is reset. do. load 3 according to the value of J.
Operate 2. Another feature of this subroutine is that although the average value of N points is used, detection is performed at every sampling interval ΔT.

The noise width learning process is performed, for example, at intervals of 2 seconds.

Let the noise detection level be, for example, J −xmin, and V
When exceeds J-xIlii, counting of CI starts.

Setting of the noise detection level is arbitrary. Next, when V falls below the noise detection level, a signal with a length below the upper limit TI is extracted and compared with the original Tll1ax, and the maximum value of the noise duration is set as TIIIax.

Based on this Tll1ax, the sampling interval Δ
Define T. Note that here, the same level J-v is used at the start of counting the noise duration C3 and at the end of counting.
Although IIlin was used, the two levels may be different.

It is detected from the count value of C3 that an appropriate interval has elapsed, and Tmax'Pxmin is corrected to be the initial value for the next interval.

Embodiment 2 Figures 6 and 7 show a modification of the first embodiment. In this example, Tff1ax was not measured directly;
Use appropriate estimates. Then, the number d of noises whose length exceeds the estimated value Tll1ax is counted for each section.

At the time of changing the section, this count value d determines the next T
Determine wax and ΔT. For example, change T with d=1
max and ΔT are maintained, and when d=0, T wax and ΔT are decreased by 20%. For d=2, T+oax and ΔT are 20
% and increases by 40% when d is 3 or more.

A new microcomputer 72 is shown in FIG. The difference is that a counter 74 for counting d is provided.

FIG. 6 shows the noise width learning subroutine of this embodiment. The noise duration C1 is determined from the time when (2) exceeds the noise detection level J-xmin.

From the comparison between C3 and Twax, the number of noises whose length exceeds the estimated value of Tmax is counted as d, and the value of d is checked every time the interval passes. The progress of the section is C8≧T
s (2) (T3 is the length of the section). And at this point, according to the above conditions, T m
Change ax and ΔT. The point where d is calculated without directly calculating T wax, and the value of d is used to calculate ΔT at each interval change.
This embodiment is the same as the first embodiment except that max is modified.

Embodiment 3 FIGS. 8 to 1O show an embodiment in which noise is discriminated based on the maximum value of the noise and the maximum value of the length of the noise. As shown in Fig. 10, upper limits TI and PI are set for the length and magnitude of the noise, and within this range T m
The maximum value Pff1ax of the magnitude of ax and noise is determined. and the duration is less than or equal to T max and the magnitude is P
Signals below max are treated as noise.

FIG. 9 shows a new microcomputer 92. In the figure, 94 is a RAM for storing the minimum value Vmin of the sensor signal V, 96 is a counter for counting the time signals CI and C1, and 98 is a RAM for storing P max and its intermediate value P. .

Returning to FIG. 8, in this embodiment, the sensor signal V at each point in time is used without averaging using the signal sequence rl-fn. Accordingly, counting of C3 becomes unnecessary.

Furthermore, noise sampling and contamination detection are processed in the same subroutine (detection subroutine).

Also, as in the previous embodiment, the minimum value V of the sensor signal in the section
Focusing on min, contamination is detected from V>J-Vmin (3). Then, when the state of equation (3) continues for a period longer than TIIIax, or when V>Pmax-VIIlin (4), the contamination signal K is set.

In the initialization process, after appropriately setting the initial value of P max as Pl, the minimum value Vm of the sensor signal in an interval of length T is set.
Sample in. Each time a sensor signal is read, the process moves to a detection subroutine and detects that the state of V>J-Vmin has exceeded Tmax or that V has changed to P.
Since the increase has exceeded maX-vIIlin, the detection signal K is set to 1. At the same time, a signal below the upper limit TI is treated as noise, and its maximum duration is set as T max.

Furthermore, for signals whose duration is less than or equal to TI and whose magnitude is less than or equal to PI, the maximum value is extracted. This maximum value is Vmin
Divide by P max. This has two upper bounds TI and p
This means sampling the maximum value of the duration and magnitude of the noise for those less than or equal to t. A signal that exceeds either of these values is determined to be a significant signal.

When changing the section, T
max is 0.8 times the original, P wax is also 0.8 times the original, V
IIlin is 1.05 times the original value.

Although both P IIIax and T max are used here, it goes without saying that only the maximum noise duration T max may be processed.

Figures 11 and 12 show an embodiment in which noise signals are integrated and noise and significant signals are discriminated based on this integrated value.

The sensor signal is compared with the noise detection level J·V sin to determine noise, and the portion exceeding the detection level is integrated (integrated value S).

The maximum value S max of this integrated value is determined for an appropriate section, and a signal having an integrated value greater than or equal to S max is determined to be a significant signal. This is equivalent to learning the integrated value of the noise signal and performing noise discrimination based on this.

Returning to FIG. 11, the operation of this embodiment will be explained.

The main change is that the detection subroutine has been changed from FIG. 8 to that shown in FIG. 11. The main loop in Figure 8 is
Use 5IIlax instead of P wax, and P 1ll
If the initial value P of ax is changed to the initial value S of Sll1ax, it can be directly applied to the one shown in FIG.

(2) counts the time C3 during which time exceeds 11/min, and integrates the difference between V and J-Vmin to obtain S.

If CI is less than or equal to upper limit T1 and S is less than or equal to upper limit Sl, it is considered noise, and the maximum value of S is defined as S max.

The count value of S is also used to detect contamination, and if the integrated value is greater than or equal to Smax, it is determined to be contamination, and a detection signal K is set.

In each of the embodiments, the absolute value of the sensor output is used, but instead of the absolute value, a time differential value of the sensor output or the like may be used. Alternatively, a time integral value of the sensor output or the like may be used. The embodiments shown in FIGS. 11 and 12 are examples that use time integral values.

Next, each embodiment is aimed at controlling air conditioning, and places particular emphasis on detecting the generation of low-concentration gas.

For this reason, gas was detected using xmin, Vmin, etc. as a reference, taking into account the effects of changes in the sensor over time, temperature and humidity, and the like. However, when it is important to detect the absolute value of the gas concentration, such as when detecting a gas leak, it is preferable that the reference value be a fixed value rather than being varied along the sensor output. Also, the period for changing T max, ΔT, etc. is not about 1 to 4 hours in the case of gas leak detection;
It is preferable to set it as about 1 week to 1 month.

[Effects of the Invention] In this invention, noise is removed through learning the duration of the noise.

[Brief explanation of the drawing]

Figures 1 and 2 are characteristic diagrams of the first embodiment, Figure 3 is a flowchart of the main parts of the first embodiment, Figure 4 is its circuit diagram,
FIG. 5 is a flowchart showing details of the operation. FIG. 6 is an operation flowchart of the second embodiment, and FIG. 7 is a block diagram of the main part of the circuit. FIG. 8 is an operation flowchart of the third embodiment, FIG. 9 is a block diagram of the main part of the circuit, and FIG. 1O is a characteristic diagram thereof. Figure 11 is the fourth
The operation flowchart of this embodiment is shown in FIG. 12, and its characteristic diagram is shown in FIG. In the figure, 4 gas sensor, 12.72.92 microcomputer, 14
A/D converter 16 arithmetic logic unit.

Claims (6)

[Claims] (1) In a gas detection method using a gas sensor, a sensor signal that lasts only a short time is determined to be noise, the duration of the noise is memorized, and the gas detection conditions are changed depending on the duration of the noise. gas detection method. (2) A gas detection method according to claim 1, characterized in that the duration of the noise is detected from a comparison between a sensor signal and a noise detection level. (3) In the method according to claim 1, a sensor signal is sampled at time intervals, a weighting factor is added to the sampled sensor signal, and a plurality of weighting factors are integrated,
A gas detection method characterized by detecting gas based on this integrated signal and changing the sampling time interval depending on the duration of the noise. (4) A gas detection method according to claim 1, characterized in that a signal that lasts for the duration of noise or longer is defined as a significant signal, and gas is detected based on this significant signal. (5) The method according to claim 1, characterized in that the duration of the noise and the magnitude of the noise are detected and stored, and based on this, it is determined whether the sensor signal is a significant signal or not. gas detection method. (6) A gas detection method according to claim 1, characterized in that an integrated value of noise signals is detected and stored, and a sensor signal exceeding this is determined to be a significant signal.