JPH0618466A - Gas sensor - Google Patents

Abstract

PURPOSE:To eliminate the random fluctuation of gas sensor signals, and detect gas quickly and accurately by correcting the signals obtained from local behavior, using the wide range features of the signals. CONSTITUTION:A differentiation circuit 16 detects gas generated due to smoking, via the single or double differentiation of a gas sensor signal. A time constant used for the single or double differentiation is, for example, approximately between 2 to 10 seconds. Also, a time interval is shortened within a range where gas can be identified from a noise signal for quickly detecting gas due to smoking. A noise filter 18 detects a random noise having a small time constant without any directivity from a gas sensor signal V, generated by wind or the like from a window or a door. A counter 20 is added, for example, with a value of one, when DELTAV obtained with the circuit 16 is equal to or above a detection threshold D, while being subtracted, for example, by a value of two at DELTAV less than D. The target value of the counter 20 is determined by the signal of the filter 18, and made to increase according to a noise rise. When the counter 20 counts values up to the target, an air cleaner 12 is caused to operate.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a gas detection device, and more particularly to control of an air purifier of an automobile and detection of gas in real time in the presence of noise.

[0002]

2. Description of the Related Art It has been proposed to control the air conditioning of automobiles by using a metal oxide semiconductor gas sensor such as SnO2. The target of detection is the gas accompanying smoking, and in order to correct the drift due to the ambient temperature and humidity of the gas sensor, changes over time, etc., the envelope of the gas sensor signal can be extracted and smoking can be detected by the deviation from the envelope. Proposed.

However, the inventors have encountered the following problems in controlling the air purifier of a vehicle. 1) Drift occurs in the gas sensor signal due to the operation of the air conditioner cooler and heater, the number of passengers in the car, and so on. The drift is large and is quicker than the fluctuation of the temperature and humidity of the outside air, and it is difficult to distinguish it from smoking by detecting the envelope. For this reason, especially when a large number of passengers board a car,
The gas sensor signal increases and the air purifier operates. 2) When running with the car window open, the signal from the gas sensor will generate significant noise. In particular, when traveling with two front and rear windows opened, a signal larger than that due to smoking is generated. The noise problem occurs even when the door of a car is opened, and when the door is opened, strong and large random noise is generated. Changes in the sensor signal when the door is open are difficult to distinguish from changes in the sensor signal due to smoking.

As a related prior art, Japanese Patent Publication No. 3-14133 proposes to obtain an envelope of a gas sensor signal and detect a gas by a deviation from the envelope. Further, Japanese Unexamined Patent Publication (Kokai) No. 3-271,020 discloses a problem of an air cleaner for an automobile by envelope detection.
It has been reported that the air purifier malfunctions when the door is opened causing significant noise.

A basic problem with air purifiers for automobiles is to detect smoking quickly. Rapid detection of smoking requires the extraction of small changes in the gas sensor signal, which is inconsistent with the noise and drift problems from windows and doors.

Here, the prior art is shown for the air cleaner of the automobile. However, such a problem also applies to, for example, control of an air purifier for a living room or when gas is detected in real time in the presence of noise.

[0007]

SUMMARY OF THE INVENTION A basic object of the present invention is to provide a gas detection device having a high S / N ratio against noise.

[0008]

The gas detector of the present invention comprises means for detecting gas from the local behavior of the gas sensor signal,
A means for extracting the global characteristics of the gas sensor signal from the past global behavior of the gas sensor signal, and for correcting the gas detection signal obtained from the local behavior with the global characteristics of the gas sensor signal. And means are provided. Further, the gas detecting device of the present invention differentiates the gas sensor signal by at least one or more steps to detect the gas, a means for detecting a random fluctuation of the gas sensor signal, a signal of the gas detecting means and a random fluctuation. It is characterized in that means for controlling the air conditioner is provided by combining the signals of the detecting means of. The gas detection device of the present invention further comprises means for differentiating the gas sensor signal by at least one or more levels to detect gas, means for detecting a sharp drop in the gas sensor signal, and detection of a sharp drop in the gas sensor signal. After that, a means for correcting the gas detection by the gas detection means so as to suppress the gas detection is provided.

[0009]

According to the present invention, the global characteristics of the gas sensor signal are extracted from the past global behavior of the gas sensor signal. Extracting the global characteristics is to evaluate the degree of noise in the atmosphere in which the gas sensor is placed, and in the case of automobile air conditioning, for example, the open state of windows and doors, the drift state after closing windows and doors, This means classifying the atmosphere into stable states after the end of drift. It also means that a signal that reflects the background gas concentration, humidity, etc. can be obtained and used, for example, to compensate for the non-linearity of the gas sensor signal. The signal of the gas sensor is generally non-linear with respect to the gas concentration, but if the rough influence of the background gas concentration and humidity is known from the global characteristics, the non-linearity can be approximately compensated. For example, for the average value of the sensor signal over a long time,
It is assumed that there is basically no background effect. Next, the ratio or the difference of the gas sensor signal in the vicinity of the present shows the change of the background gas concentration and the humidity from the long time average. Therefore, the non-linearity can be approximately corrected from this ratio or difference. By extracting the global characteristics in this way, it is possible to improve the S / N of the detection or to compensate the non-linearity of the sensor.

The present invention will be described by taking the case of air-conditioning control of a vehicle as an example. In order to speed up the detection, the gas is detected by the differential of the first floor or the second floor. Next, the random fluctuation of the gas sensor signal is extracted to detect that the door or window is open. The extraction of random fluctuations is the extraction of global features. Preferably, if the extraction result of the random fluctuation is corrected with a certain time constant, the influence of the extraction result of the random fluctuation remains in the initial stage of the drift after closing the window or door, and the initial stage of the drift can be detected. This is nothing but correcting the detection conditions by classifying the atmosphere in the passenger compartment into three states: the open state of windows and doors, the subsequent drift, and the stable period after the end of the drift. The classification of the atmosphere in the passenger compartment is to globally classify the atmosphere in which the gas sensor is placed. This is an average of the long-term average of the sensor signal and the signal of the sensor signal over a long period. It can be realized by using one. For example, time constant 30
An average value obtained by convoluting the past gas sensor signals by convolution is used for about 4 minutes to 4 hours. The time constant is, in principle, the time until the signal drops to 1 / e with no input. The difference, ratio, etc. between this average value and the current gas sensor signal is useful in classifying the atmosphere during window and door opening, drift, and stable periods. The smallest gas sensor signal at present is the characteristic of the opening period of windows and doors, and the next smallest is the characteristic of the drift period. The largest current gas sensor signal is characteristic of the stable period. The gas sensor signal is smooth and the random noise is extremely small during the drift and stable periods.
Therefore, if the differential value of the sensor signal is processed by, for example, a bandpass filter, the characteristics of the drift period and the stable period can be extracted. For example, a bandpass filter that passes signals in the drift period is provided, a counter is added when the differential coefficient of the sensor signal enters the band of the bandpass filter, and a bandpass filter that passes differential signals when the window or door is open is used. The counter is subtracted or reset by this signal. The signal of the counter indicates the time after entering the drift period, and the signal of the counter can classify the atmosphere into the opening period of the windows and doors, the drift period, and the stable period.

A characteristic of the opening of the windows and doors is that the sensor signal drops off first. Therefore, this sudden decrease may be detected to detect the opening of the window or door.

[0012]

According to the present invention, attention is paid to the global characteristics of past gas sensor signals, and the S / N ratio of the gas detection device is improved. This means that in the case of an air purifier for a vehicle, malfunctions due to opening of windows and doors and malfunctions due to drift after the windows and doors are closed can be eliminated. In addition, since malfunctions can be reduced, it becomes easy to increase the detection speed by differentiation. In addition, the non-linearity of the gas sensor can be corrected from the global characteristics.

[0013]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 shows a main part of an embodiment of a control device for an air purifier for an automobile. In the figure, 2 is a metal oxide semiconductor gas sensor, 4 is its heater, and 6 is its metal oxide semiconductor. The metal oxide semiconductor 6 may be made of SnO2, WO3, In2O3, or the like. In addition to the metal oxide semiconductor gas sensor 2, for example, a catalytic combustion type gas sensor utilizing temperature change due to combustion of combustible gas, or a proton conductor gas sensor utilizing detection of H2 and CO generated by smoking can be used. Reference numeral 8 is a power source such as a battery, and RL is a load resistance of the metal oxide semiconductor 6.

A signal processing microcomputer 10 is a 4-bit 1-chip microcomputer. Reference numeral 12 denotes an air purifier, for example, an electric dust collector that removes smoke associated with smoking, and an ozone generator that removes odor associated with smoking. Instead of the air purifier 12, a ventilation device, an outside air introduction device, and an air purification catalyst may be controlled.

The structure of the microcomputer 10 is shown. Reference numeral 14 is an AD converter, and 16 is a differentiating circuit, which detects the gas accompanying smoking by the first-order differentiation or second-order differentiation of the gas sensor signal. The time constant used for the first-order differential or the second-order differential is, for example, about 2 seconds to 10 seconds, the time width is shortened within a range where the distinctiveness from noise is obtained, and the gas due to smoking is promptly detected. A noise filter 18 detects random and non-directional noise having a short time constant of the gas sensor signal V due to noise such as wind from a window or a door. Reference numeral 20 denotes a counter, for example, 1 is added when the differential signal ΔV obtained by the differentiating circuit 16 is the detection threshold value D or more, and when ΔV <D, for example, 2
Is subtracted. The target value J of the counter 20 is determined by the signal of the noise filter 18, and J is increased as the noise becomes more intense. When the value of the counter 20 reaches J, the air purifier 12
To operate. Reference numeral 22 is a stop control circuit for the air purifier 12.

FIG. 2 shows the internal structure of the microcomputer 10. 16 is the above-mentioned differentiating circuit for performing first-order differentiation,
The noise filter 18 convolves the negative differential signal with the time constant k according to the equations (1) and (2). The time constant k is ΔV
When ≧ 0 continues, the time until the value of F decreases to 1 / e is about 30 to 120 seconds. Expression (1) is the convolution, and Expression (2) is the normalization of the noise F. F = F · k + | ΔV | ΔV <0 (1) F ′ = F / V (2) The operator of the digital first derivative is determined by, for example, the formula (3). ΔV = V1−V2 (3) (V1 is the current value of the sensor signal V, V2 is the value one second before) The stop control circuit 22 has a memory 23 and a timer 24 for storing the sensor signal Von when smoking is detected. , Multiplication circuit 25,
It is composed of a comparison circuit 26, and formula (4) is a condition for stopping the air purifier 12. Von (1 + T) ≧ V (4)

[0017]

Second Embodiment FIG. 3 shows the detection of noise similar to the 0-crossing method, and in FIG. 3 and subsequent figures, the same reference numerals as those in FIGS. Reference numeral 31 is an average value calculation circuit, which calculates the time average of the sensor signal V by simple averaging or convolution.
Reference numeral 32 is a comparison circuit, which is an average value and Δ1 and Δ2 from the average value
The three detection lines at points (Δ2> Δ1) that are only small are drawn, and when the sensor signal V passes through the detection lines, the counter 3
Add 3 For example, the addition value is set to 1 when the average value is crossed, 2 when the average value-Δ1 is crossed, and 3 when the average value-Δ2 is crossed. Reference numeral 34 is a timer, which subtracts the data of the counter 33 by 1 every 30 seconds, for example, and the counter 33 is provided with a circuit for preventing generation of a negative number and overflow.

[0018]

Third Embodiment FIG. 4 shows a third embodiment. An edge detection circuit 41 further digitally differentiates the first-order differential signal ΔV of the differentiating circuit 16 to obtain a digital second-order differential signal Δ ² V. For example, an operator of (V1-2V2 + V3) is used for the digital second-order differentiation, and detection is performed using two seconds from V1 to V3 (V1 is the current value of the sensor signal V, V3 is the sensor signal two seconds before).

[0019]

Fourth Embodiment FIG. 5 shows a fourth embodiment. According to experiments, the sensor signal V decreases first when the window or door opens. This would be a response to the gas sensor 2 being cooled by the air flow. Since the sensor signal V increases when smoking, a sharp decrease in the sensor signal is an indication that the window or door has opened. In FIG. 5, reference numeral 51 denotes a noise detection circuit that causes noise when the digital first-order differential signal ΔV of the sensor signal V is −B or less, and smoking is performed by the comparison circuit 52 when ΔV is the threshold value A or more. In the logic circuit 53, when the noise is detected, the timer 34 is activated to set a flag, the AND circuit 54 prohibits the operation of the air purifier 12 while the flag is set, and when there is no flag and there is a smoking detection signal, the air purifier is detected. Drive twelve.
Preferably, -B is used as a threshold value, and the flag setting time is extended as the magnitude (absolute value) of the first-order differential signal increases from -B.

[0020]

Fifth Embodiment FIG. 6 shows a fifth embodiment. The differentiating circuit 16 obtains the digital first-order differential signal ΔV of the sensor signal V, and the edge detection circuit 41 further digitally differentiates this to obtain the digital second-order differential signal Δ ² V. Then, if the digital second-order differential signal is greater than or equal to the threshold value D2, smoking is assumed. On the other hand, digital 1
Assuming that the floor differential signal ΔV is −B or less and noise due to windows, doors, etc. has occurred, the timer 3 is used as in the embodiment of FIG.
4 is activated and a flag is set. The AND circuit 54 operates the air purifier 12 when an edge caused by smoking occurs and there is no flag for noise.

[0021]

Sixth Embodiment FIG. 7 shows a sixth embodiment. 71 in the figure,
Reference numeral 72 denotes an addition circuit, which adds the absolute values of the changes in the gas sensor signal V, and the addition circuit 72 adds the digital first-order differential signal ΔV of the gas sensor signal V with a sign. When noise is generated, the output of the adding circuit 71 for adding the absolute value of ΔV is large, and the output of the adding circuit 72 for adding ΔV with a sign is smaller than this. Therefore, if these ratios are calculated by the division circuit 73, it is possible to discriminate between noise and smoking. This is input to the AND circuit 54, and when there is a smoking detection signal from the digital first-order differentiation circuit 16 and there is no noise detection signal from the division circuit 73, the air purifier 12 is driven.

In these embodiments, a specific example is shown, but the present invention is not limited to this. For example, the first digital differential signal ΔV and the second digital differential signal Δ ² V may be obtained by analog differentiation instead of digital differentiation. 5 to 7
In the embodiment, the signal from the differentiating circuit 16 is invalidated when the noise is generated, and the air purifier 12 is not operated. On the other hand, in the embodiment shown in FIGS.
0 is used to delay detection when noise occurs,
The detection is limited to the case where the sensor signal V continues to increase monotonously (up to 50 seconds).

[0023]

[Operation of Embodiment] FIG. 8 shows an operation of the embodiment when the microcomputer 10 of FIG. 2 is used. 3 to 7
The operation in this embodiment is obvious from the description of FIG. The microcomputer 10 AD-converts the gas sensor signal V every one second, for example, and differentiates it by the first order by the differentiating circuit 16. The ratio between the digital first-order differential signal ΔV and the sensor signal V is obtained, and smoking is carried out when this value is D or more. The noise filter 18 calculates the noise amount F by convoluting the differential value of the sensor signal every second by convolution with a time constant of about 30 to 120 seconds only when the differential value is negative, and calculates the noise amount F and the gas sensor. The noise F'is detected from the ratio with the signal V. The target value J of the counter 20 is
J0 (for example, 2) is minimized and increased in accordance with F'using a constant p1 so that only the increase in the sensor signal V that continues monotonically for a long time is detected as the noise is severe, and the noise is small or no noise. Sometimes smoking is detected promptly.

FIG. 9 shows operation waveforms in each embodiment. A characteristic of the gas sensor signal V due to smoking is that the sensor signal V increases simply and monotonously as shown on the right side of FIG. 9. Therefore, for example, if the sensor signal V is subjected to digital first-order differentiation, smoking can be detected. Further, if the sensor signal V is subjected to digital second-order differentiation, smoking can be detected from the edge of the increase in the sensor signal V accompanying the start of smoking. As shown in the left side of FIG. 9, a characteristic of noise due to windows, doors, etc. is a random fluctuation of the sensor signal V, which is detected to suppress the smoking detection signal. The small 0 mark on the left side of FIG. 9 is an edge in noise. When smoking is detected, the stop threshold value Von (1 + T) of the air purifier 12 is gradually increased by the timer 24, and the air purifier 12 is stopped at the point where the sensor signal V is crossed. As a result, the operating time of the air purifier 12 is proportional to the increase in the sensor signal V due to smoking, and the operating time becomes longer as the amount of generated gas increases.

10 and 11 show a smoking signal, water vapor and odor generated by an occupant, drift of a sensor signal due to operation of an air conditioner, and noise from a door. In FIG. 12, 80
2 shows noise when the vehicle is run at km / hr and two windows are opened in the front and back to blow air into the passenger compartment for 1 to 6 minutes. In FIG. 12, the sensor resistance decreases at about 4 minutes,
This is because the truck passed around and inhaled the bad odor.
In these figures, the sensor resistance RS which is the reciprocal of the sensor signal V
Is shown on the vertical axis, and particularly in FIGS. 10 and 11, the maximum value of the sensor resistance is normalized to 1. As shown in FIGS. 10 and 11, drift caused by water vapor, odor, etc. generated from the air conditioner and passengers is slow, and can be distinguished from the smoking signal of FIG. 10 by a differential coefficient. Most of the noise when the door is opened can be eliminated by setting the time constant of the above convolution integration to 30 seconds (constant k is 0.97 when performing convolution every 1 second). Figure 1
In the case of 2, the sensor resistance continues to fluctuate randomly while the window is open, and malfunction can be prevented. If you close the door or window,
Since the time constant of the noise filter 18 is as long as 30 to 120 seconds, the air purifier 12 does not malfunction even if the noise filter 18 temporarily drifts with a gradient of the detection threshold D or more in the initial stage of the drift. Since the counter 20 is used in the embodiments of FIGS. 1 to 4, when the window or door is opened, for example, when the sensor signal V exceeds the threshold value D for 50 seconds or more,
The air purifier 12 is operated and, for example, 20 during the drift period.
The air purifier 12 is operated when it exceeds D for a second or more continuously, and when the drift ends, for example, the air purifier 12 is operated in 2 seconds for D or more. That is, as the noise becomes more severe, the detection speed is sacrificed, and only a more reliable signal is taken out. When the noise is small, the signal is detected promptly.

[0026]

Seventh Embodiment In FIG. 13, 2 is the gas sensor,
Reference numeral 8 is the power source. Sensor output V to load resistance RL
Is taken out using a resistor R2, and a capacitor C1 and a resistor R4
The sensor output V is differentiated by a differentiating circuit having a time constant of about 1 to 2 seconds. Resistor R so that the differential output does not become negative
A positive bias is applied to the differential output using 3 and R4, and this is amplified by, for example, about 10 times in the amplifier circuit A1. In this way, amplification is performed in the preceding stage of the AD converter to improve the resolution.

100 is a new microcomputer,
Reference numeral 102 is, for example, a 2-port AD converter, 104 is a smoking detection circuit using logarithmic differentiation of the sensor signal V, and 106 is a random noise detection circuit for detecting random noise of the sensor output, which uses | dlnV / dt | integration. 1
Reference numeral 08 denotes a window opening detection circuit for detecting the opening of a window or a door from a sudden decrease in the sensor output V, 110
Is a long-term average detection circuit for obtaining a long-term average of the sensor output V, 112 is a stop level generation circuit for generating a stop level of the air purifier, and 114 is a temporary memory of the sensor output V. Further, 116 is a NOR gate, 118 is an AND gate, 120 and 122 are inverters, and 124 is a comparator. The window open detection circuit 108 may not be provided.

FIG. 14 shows the signal flow in the seventh embodiment.
The sensor output V is analog-differentiated by the capacitor C1 and amplified by the amplifier circuit A1, and the sensor output differential signal dV / dt
Take out. This is port P1 of AD converter 102.
And supplies it as a logarithmic differential dlnV / dt to the smoking detection circuit 104, the random noise detection circuit 106, and the window opening detection circuit 108. The sensor output V is applied to the port P2 of the AD converter 102, and this is supplied to the sensor output long-term average detection circuit 110, the air purifier stop level generation circuit 112, and the temporary memory 114.

As shown in FIG. 12, when the window opens, the sensor resistance RS increases and the sensor output V decreases. Figure 10,
As shown in FIG. 11, the sensor output V decreases even when the door is opened. However, the decrease in the sensor output V when the door is opened is smaller than the decrease when the window is opened. The window opening detection circuit 108 detects that dlnV / dt is more negative than the threshold value -X, and outputs the window or door opening signal. The window opening detection circuit 108 has a built-in timer, and when it is detected that a window or door is opened, for example, 1-2
Start the minute timer. Random noise detection circuit 10
In 6, the gate G is provided to pass the logarithmic differential signal only when the logarithmic differential of the sensor output V is negative. The random noise amount C stored in the memory is multiplied by a correction constant k in the multiplication circuit M to obtain k · C, and the logarithmic differentiation of the sensor output V is added to this by an addition circuit (absolute value subtraction since it is a negative signal). K · C-dlnV / dt is fed back to the memory, and this is compared with the threshold value by the comparator C. When the threshold value is exceeded, a random signal detection circuit F2 is generated. On the other hand, the smoking detection circuit 104 uses the logarithmic derivative d (l
Smoking is detected because nV) / dt is larger than the threshold value k1.

The long-term average detection circuit 110 of the sensor output utilizes the NOR gate 116. When there is no smoking detection signal, no random noise is detected, and there is no window opening detection signal, the long-term average L of the sensor output V is detected. To fix. For example, convolution is used to correct the long-term average L. The long-term average time constant is, for example, 10 minutes to 1 hour, and the correction constant is k3. This corresponds to finding the average value of the sensor output when the window and door are closed and there is no smoking. Alternatively, instead of obtaining the long-term average, the sensor output immediately before the window opens may be stored. In the long-term average detection circuit 110 of the sensor output, the ratio between the current value V of the sensor output and the long-term average L is obtained, and the detection threshold value k1 in the smoking detection circuit 104 and the air cleanliness are obtained from the reference table by this ratio V / L. The correction constant k2 of the stop level in the stop level generation circuit 112 of the reactor is determined. The smaller the value of V / L, the larger the smoking detection threshold value k1 and the larger the correction constant k2 of the stop level L2 of the air purifier. This means that, for example, in FIG. 12, the smoking detection threshold value k1 is increased in a state where the window is opened and the sensor output V is decreased to reduce the sensitivity to smoking. Further, when the window is closed after 6 minutes, the sensor output V monotonically increases. During this period, the sensor output V is smaller than the long-term average L, and the detection sensitivity is lowered by increasing k1 to prevent malfunction. If smoking overlaps here, the increase in the sensor output due to smoking and the increase in the sensor output due to the closing of the window overlap, and smoking can be detected even if the threshold value k1 is increased. Similar effects are obtained in FIG. 10 and FIG.
It is also obtained for the increased drift of the sensor output after the door is closed at. Sensor output V when the door is open
Is small, and with respect to the drift of the sensor output after the door is closed, the malfunction threshold due to drift can be prevented by increasing the detection threshold value k1 for smoking as V / L becomes smaller. If smoking overlaps here, smoking can be detected because dlnV / dt is larger than in the normal case.

In the embodiment of FIG. 13, smoking detection signal F1
Occurs, and the air purifier 12 is driven only when there is no random noise detection signal F2 or window opening signal F3. When the air purifier 12 is operated, the sensor output V before smoking detection is stored as the initial value of the stop level L2 of the air purifier. The air purifier 12 only removes smoke accompanying smoking and does not remove gas, and the sensor output V does not decrease even when the air purifier 12 is operated. According to the increase in the sensor output V due to smoking (peak height due to smoking), the air purifier 1
2. Determine the operating time. Thereby, the operating time of the air purifier 12 is made to correspond to the generated smoke concentration. The sensor output V is not proportional to the gas concentration but to the 0.6th power of the gas concentration. Further, the higher the background gas concentration and humidity, the smaller the increase in the sensor output with respect to the gas generated by smoking. Therefore, from the ratio V / L of the sensor output V and the long-term average value L of the sensor output, the correction constant k2 is determined in the reference table
As the V / L increases, the background gas concentration and humidity increase, and the sensor has an apparently low sensitivity.
The correction constant k2 of the stop level L2 is reduced.

In the stop level generation circuit 112 of the air purifier, the sensor output V before the detection of smoking is stored as the stop level L2, and the timer output (L) is set to L at predetermined intervals by using a timer (not shown).
2 is multiplied by the modified constant k2 (0 <k2 <1), and this is set as a new stop level L2. This means that, for example, in FIG. 10, an oblique curve is drawn from the sensor resistance RS immediately before the detection of smoking on the side where RS decreases, and the air purifier 12 is operated until this curve intersects the actual sensor resistance RS. To do. Then, this time becomes a function of the depth of the valley of the decrease in the sensor resistance due to smoking in FIG. 10, and the influence of the background humidity and the combustible gas that has entered from the outside air, and the effect of the gas from the previous smoking when continuously smoking. Is compensated for, and the time is proportional to the generated smoke concentration. If the vehicle air is replaced and the sensor output V decreases, the operating time of the air purifier 12 is shortened accordingly. Required air purifier 12
Operating time depends on the internal volume of the vehicle, the capacity of the air purifier 12,
Determined by the smoke density generated. The capacity of the air purifier 12 is known, and if the smoke concentration generated in the above process is obtained from the gas concentration, the air purifier 12 can be controlled appropriately.

Here, it is shown that the detection threshold value k1 of smoking is corrected by the ratio V / L of the sensor output V and the long-term average L. However, simply detecting smoking may be prohibited when the ratio V / L decreases below a predetermined range. In this case, in case the window or door keeps opening for a long period of time and the meaning of the stored long-term average L is lost, when V / L is below a predetermined value, L is gradually decreased, and after a long period of time V / L May be increased and detection of smoking may be restarted.

[0034]

Eighth Embodiment FIG. 15 shows the basic configuration of the eighth embodiment. In the figure, 202 is a gas sensor, for example, a metal oxide semiconductor gas sensor, a proton conductor gas sensor, a solid electrolyte gas sensor, or the like is used. Reference numeral 204 is a differentiating circuit, and 206 is a filter for generating an auxiliary signal for evaluating the reliability of the differential signal. Reference numeral 208 denotes a comparison circuit, which receives a detection threshold value from the filter 206 and generates a gas detection signal when the time derivative of the sensor signal is equal to or greater than this detection threshold value.
In the embodiment, the first-order differentiating circuit is used as the differentiating circuit 204, but a second-order or higher-order differentiating circuit may be used. In the example,
Although the auxiliary signal is generated by the filter 206 from the signal of the gas sensor 202 itself, for example, the filter 20 of FIG.
A compensation sensor such as a humidity sensor is arranged on the input side of 6.
The signal of the sensor for compensation may be processed by the filter 206 to generate the auxiliary signal. Differentiating circuit 204, filter 2
06, the comparison circuit 208 is realized by a microcomputer, for example.

16 to 19 show examples of the filter 206. The noise filter 226 of FIG.
The signal of No. 02 is subjected to the second differentiation by the second differentiation circuit 222, and the integration circuit 224 performs the convolution integration of the absolute value of the second differentiation to detect the random fluctuation of the sensor signal.

FIG. 20 shows a signal waveform of the gas sensor 202 inside the vehicle when it is used to control an air purifier for an automobile. The atmosphere in the passenger compartment can be classified into three states: a state where a door or a window is opened, a drift period thereafter, and a stable period after the drift ends. A characteristic of opening a door or a window is that the sensor signal V sharply decreases and intense random fluctuations occur. The characteristic of the drift is that the sensor signal V continues to increase gradually and monotonically over a long period of time. When the drift ends, the sensor signal V becomes almost constant. However, in reality, there is no stable period in the exact sense, and it seems that the velocity of drift is merely sufficiently small. The purpose of detection in the control of the air purifier is to detect smoking, it is not necessary to detect smoking when the window or door is opened, and it is sufficient to detect only smoking in the drift period or the stable period. The characteristic of smoking is that the sensor signal V rapidly increases.

[0037]

[Table 1] Main part of program Read the noise filter READ VS sensor output Vs in Fig. 16 (every second) V6 = V5: V5 = V4: V4 = V3: V3 = V2: V2 = V1: V1 = VS Latest Output 5 to V1 and input 5 to V6 D1 = (V1 + V2-V3-V4) / V1 1st derivative (time constant 2 seconds) D2 = (V1 + V2-2 * V3-2 * V4 + V5 + V6) / V1 2nd derivative (time constant 2 seconds) IF D2 ≦ 0 THEN F = F * (1-P) + ABS (D2) ELSE F = F * (1-P) Evaluate noise with a noise filter, negative Convolve the edge of J = INT (F) * K1 + J0 Increase the threshold value from J0 according to the output F of the noise filter IF D1 ≧ J THEN X = 1 Drive the air purifier with the differential signal above the threshold value.

Referring to Table 1 and FIG. 20, FIG. 15 and FIG.
The operation of the sixth embodiment will be described. The sensor signal V is read every 1 second, and the first-order differentiation and the second-order differentiation are performed with a time constant of 2 seconds.
When the second differential is negative, that is, when the edge on the side where the sensor signal V decreases is detected, the absolute value thereof is integrated. The integration time constant is determined by a constant P, and the integration value is reduced to 1 / e every 30 seconds to 120 seconds (when no noise is input). As is apparent from FIG. 20, the sensor signal V fluctuates drastically when the door or window is open, so that many edges occur and the second derivative is not zero. On the other hand, during the drift period, smoking period, or stable period, the sensor signal V changes substantially linearly and the second derivative is small. In the embodiment, in order to prevent the noise filter 226 from detecting the edge of the increase in the sensor signal due to smoking in the drift period or the stable period,
Only the negative second derivative was integrated. Therefore, the output of the noise filter 226 increases when the door or window is opened, and decreases with a time constant of about 30 to 120 seconds when the door or window is closed. The detection threshold value J of the sensor signal V with respect to the first-order differential signal D1 is J0
Is defined as the minimum value by the output of the noise filter 226.
As a result, the detection threshold value J increases when the door or window is open, and even if a differential signal having a gradient indistinguishable from smoking is temporarily generated, it is possible to prevent malfunction. Further, in the initial stage of drift, the output of the noise filter 226 remains because the time constant is as long as 30 to 120 seconds, and even if an accidentally large gradient drift occurs during this period, malfunction of the air purifier can be prevented. it can. The output of the noise filter 226 becomes small when the drift ends, and the differential detection threshold becomes the smallest in the stable period. Therefore, smoking is detected even with a small differential signal in the stable period with little noise.

The differential detection threshold value J is set to the noise filter 22.
The correction with the signal of No. 6 also has the meaning of correcting the non-linearity of the output of the gas sensor 202. Gas sensor 2
The signal 02 is proportional to the 0.5th power of the gas concentration, and the higher the background gas concentration, the smaller the rate of change of the sensor signal V per cigarette. When the doors and windows are open, ventilation is progressing and the gas concentration inside the vehicle is decreasing. On the other hand, during the stable period, the background gas concentration is increasing due to body odors and fragrances generated by passengers, odors generated by the vehicle itself, and odors generated by the filter of the air purifier. Therefore, the noise filter 226 suppresses the detection of smoking in the state where the door or the window is open and speeds up the detection of smoking in the stable period. Therefore, it is easy to detect smoking in the stable period when the background gas concentration is high. And has a role of correcting the non-linearity of the gas sensor 202.

FIG. 17 shows a modified noise filter 236. The operation is shown in Table 2. In the noise filter 236, the signal of the gas sensor 202 is once differentiated from the differential circuit 2
Differentiate by 32, and the integrating circuit 234 convolves and integrates only the change on the side where the sensor signal V decreases.

[0041]

[Table 2] Main part of program Read noise filter READ VS sensor output Vs in Fig. 17 (every 1 second) V4 = V3: V3 = V2: V2 = V1: V1 = VS Input to V4 D1 = (V1 + V2-V3-V4) / V1 1st derivative (time constant 2 seconds) IF D1 ≦ 0 THEN F = F * (1-P) + ABS (D1) ELSE F = F * ( 1-P) Evaluate the noise with the noise filter and convolve the negative difference J = INT (F) * K2 + J0 Increase the value from J0 according to the output F of the noise filter IF D1 ≧ J THEN X = 1 Driving the air purifier with a differential signal above the specified value

As is apparent from FIG. 20, opening the door or window is accompanied by a decrease in the sensor signal at first. Further, since the sensor signal V also changes randomly thereafter, a large value can be obtained when the window or door is opened by integrating the difference on the side where the sensor signal decreases. On the other hand, in the drift period, the smoking period, or the stable period, the sensor signal either increases or is almost constant, and the value is small even if the difference on the side where the sensor signal decreases is integrated. Therefore, if the difference is negative, it is integrated and convolved with a time constant of about 30 to 120 seconds. If the door or window is open, the noise filter 2
36 emits a large signal, after which the doors and windows close and 6
The noise signal 236 continues to generate a signal for about 0 to 240 seconds. The time constant for the signal to decrease to 1 / e was set to 30 to 60 seconds, and the duration of the signal of the noise filter 236 was set to 60 to 240 seconds. As a result, when the door or window is open, the differential detection threshold value J is maximized, and the detection threshold value is set to an intermediate value during the drift period.
In the subsequent stable period, smoking can be detected quickly by minimizing the detection threshold.

[0043]

[Table 3] Main parts of program Envelope detection READ VS Read sensor output Vs in Fig. 18 (every 1 second) V4 = V3: V3 = V2: V2 = V1: V1 = VS The latest output is 3 before V1. Input to V4 D1 = (V1 + V2-V3-V4) / V1 1st derivative (time constant 2 seconds) IF T MOD 10 = 0 AND V1 ≧ VH THEN VH = VH * (1 + p2) IF T MOD 10 = 0 AND V1 ≦ VH THEN VH = VH * (1-p2) Correct the envelope value VH with the correction constant P2 every 10 seconds Q = VH / V1 Find the ratio of VH and V1 J = INT (Q * K3) + J0 Increase the threshold value from J0 according to the ratio value IF D1 ≧ J THEN X = 1 Drive the air purifier with a differential signal above the threshold value

FIG. 18 shows a filter using the global filter 246. In the figure, 242 is an envelope detection circuit, 24
Reference numeral 4 is a division circuit. Table 3 shows the operation of the embodiment shown in FIG. When the value of the envelope that follows the sensor signal V with a time constant of about 30 minutes to 4 hours is calculated, this is the sensor signal V
The detailed behavior of is omitted. This value is, so to speak, a long-term average value of the sensor signal V. Comparing the long-term average of the sensor signal V with the current value of the sensor signal V, the current value of the sensor signal V is small when the window or door is open, and the current value of the sensor signal V is stable during the stable period or smoking period. It will be big. Therefore, in the global filter 246 of FIG.
For example, the envelope value VH is corrected every 10 seconds, and the correction constant P2 (a time constant of about 30 minutes to 4 hours) is used.
The envelope value VH is made to follow the sensor signal V gradually. Note that instead of using the envelope, a past simple average of the sensor signal V or the like may be used, as long as the past gas sensor signal is statisticized. The dividing circuit 244 obtains the ratio between the envelope value VH and the current value V1 of the gas sensor signal, and thereby the differential detection threshold value in the comparing circuit 208 is generated.

Returning to FIG. 20, the operation in this case will be described.
When the door or window is open, the sensor signal V decreases,
VH / V1 increases and the differential detection threshold value J increases. On the other hand, if the sensor signal V1 increases due to drift, VH / V
1 gradually decreases and the differential detection threshold value J gradually decreases. The value of VH / V1 becomes smaller at the stable period, and accordingly the differential detection threshold value J becomes minimum. The filter 246 of FIG. 18 also has an effect of increasing the detection threshold value when the gas sensor signal V is small and compensating for the non-linearity of the gas sensor signal.

FIG. 19 shows a filter 256 using a bandpass filter for the differential signal, and Table 4 shows its operation. In FIG. 19, 252 is a differentiating circuit, 254, 25
Reference numeral 5 is a window for confirming whether the differential coefficient is within a predetermined range, and 258 is a counter. Window 254
Is 1 or more and less than or equal to b, the counter 258 is incremented by 1, and the window 255 has a first-order differential D1 of -c.
In the following, the counter 258 is subtracted, for example, to 3/4 of the original value. Here, c>b> a.

[0047]

[Table 4] Main part of program Read differential filter READ VS sensor output Vs in Fig. 19 (every 1 second) V4 = V3: V3 = V2: V2 = V1: V1 = VS Input to V4 D1 = (V1 + V2-V3-V4) / V1 1st derivative (time constant 2 seconds) IF D1 ≦ b AND D1 ≧ a THEN C2 = C2 + 1 IF D1 ≦ -C THEN C2 = INT (3 / 4 * C2) c>b> a If the differential signal is within the drift range, add the counter, and if the sensor output sharply decreases, subtract the counter. J = J0-C2 * K4 The differential value with the counter value. Correct IF D1 ≧ J THEN X = 1 Drive the air purifier with a differential signal above the threshold value

The operation in this case is shown in FIG. 20. When the door or window is opened, the sensor signal V sharply decreases, so that the first-order differential signal D1 becomes -c or less and the value of the counter 258 is rapidly subtracted. . On the other hand, when the sensor signal V increases between the slopes a and b in the drift period, the window 254 is opened and the counter 258 is added. As a result, as the drift progresses, the value of the counter 258 increases, and when the door or window opens, the value of the counter 258 rapidly decreases and approaches 0. The larger the value of the counter 258, the smaller the differential detection threshold value, and the smaller the value of the counter 258, the larger the differential detection threshold value. When the door or window is open, the differential detection threshold value becomes the maximum. In the drift period, the differential detection threshold value becomes slightly large, and in the stable period, the differential detection threshold value for smoking becomes the minimum. Note that this filter 25
6 also has a role of correcting the non-linearity of the gas sensor signal and increasing the sensitivity in the stable period when the apparent sensitivity becomes small.

The operation of any of the filters shown in FIGS. 16 to 19 can be summarized as shown in FIG. Each of the filters in FIGS. 16 to 19 classifies the atmosphere in the vehicle compartment into three states of a state in which windows and doors are open, a drift state, and a stable period in some sense. For example, with the filters of FIGS. 16 and 17, it is detected that the window or door is open, and the initial time of the subsequent drift is detected by setting the time constant of the filter to a long time of about 30 to 120 seconds. In addition, in the filter 246 of FIG. 18, the envelope value VH and the gas sensor 20
From the comparison with the current value V1 of the signal of No. 2, three states, that is, the state where the window or door is opened literally, the drift period, and the stable period are detected. In the filter 256 of FIG. 19, the drift is detected in the window 254, the opening of the window or the door is detected in the window 255, and the counter 258 is subtracted. The detection in this case is centered on detecting the drift in the window 254. In this way, when the atmosphere is classified into the stable period, the drift period, and the window and door opening period, the differential detection threshold value J is corrected accordingly. As a result, the detection threshold value J of the differential increases as the noise becomes more intense as shown in FIG. If the noise is extremely intense, the threshold value J may be set to infinity to substantially prohibit the detection of the differential.

In the embodiment, the differential detection threshold value J is corrected instead of invalidating the differential signal when noise is detected. I will explain another meaning of this. For example, even when the window or door is open, the sensor signal V may continuously increase for about 30 seconds. 2 here
If a filter with a time constant of about 0 seconds is used and the differential detection signal is invalidated by the signal of the filter, a malfunction may occur with a short time constant. However, if the time constant of the filter 206 is extended here, a dead time will occur during the drift period after the window or door is closed. However, when smoking is actually performed in the drift period, the increase in the sensor signal due to the drift is added to the increase in the sensor signal due to the drift, and the differential coefficient of the sensor signal becomes larger. Therefore, if the detection threshold value J is corrected according to the magnitude of noise and the time constant of the filter 206 is lengthened to, for example, about 30 to 120 seconds, noise can be removed without causing dead time in the drift period.

[Brief description of drawings]

FIG. 1 is a block diagram of an air conditioning control device according to an embodiment.

FIG. 2 is a block diagram of a main part of an air conditioning control device according to an embodiment.

FIG. 3 is a block diagram of a main part of an air conditioning control device according to a second embodiment.

FIG. 4 is a block diagram of essential parts of an air conditioning control device according to a third embodiment.

FIG. 5 is a block diagram of essential parts of an air conditioning control device according to a fourth embodiment.

FIG. 6 is a block diagram of a main part of an air conditioning control device according to a fifth embodiment.

FIG. 7 is a block diagram of a main part of an air conditioning control device according to a sixth embodiment.

FIG. 8 is a flowchart showing an operation algorithm of the first embodiment.

FIG. 9 is a characteristic diagram of a gas sensor signal showing the detection principle of the present invention.

FIG. 10 is a characteristic diagram showing gas sensor signals inside the vehicle when the door is open and when smoking.

FIG. 11 is a characteristic diagram showing gas sensor signals in the vehicle when the door is open and when the vehicle drifts.

FIG. 12 is a characteristic diagram showing a gas sensor signal when a window is opened.

FIG. 13 is a block diagram of essential parts of an air conditioning control device according to a seventh embodiment.

FIG. 14 is a diagram showing a signal flow in the air conditioning control device of the seventh embodiment.

FIG. 15 is a block diagram of a gas detection device according to an eighth embodiment.

FIG. 16 is a block diagram of a noise filter used in the eighth embodiment.

FIG. 17 is a block diagram of another noise filter.

FIG. 18 is a block diagram of a global filter used in the eighth embodiment.

FIG. 19 is a block diagram of a differential filter used in the eighth embodiment.

FIG. 20 is a characteristic diagram of a gas sensor signal in the vehicle interior.

FIG. 21 is a characteristic diagram showing a relationship between noise and a detection threshold value for differentiation of a gas sensor signal in the eighth embodiment.

[Explanation of symbols]

 2 Gas Sensor 4 Heater 6 Metal Oxide Semiconductor 8 Power Supply 10 Microcomputer 12 Air Purifier 14 AD Converter 16 Differentiation Circuit 18 Noise Filter 20 Counter 22 Stop Control Circuit 31 Average Value Calculation Circuit 32 Comparison Circuit 33 Counter 34 Timer 41 Edge Detection Circuit 51 Noise detection circuit 52 Comparison circuit 54 AND circuit 71 Addition circuit 72 Addition circuit R2 to R4 Resistance C1 Capacitor A1 Amplification circuit 100 Microcomputer 102 2-port AD converter 104 Smoking detection circuit 106 Random noise detection circuit 108 Window open detection circuit 110 Sensor output Long-term average detection circuit 112 Air purifier stop level generation circuit 114 Temporary memory for sensor output

Claims (7)

[Claims] 1. A means for detecting a gas from a local behavior of a gas sensor signal, a means for extracting a global characteristic of the gas sensor signal from a past global behavior of the gas sensor signal, and a local means. And a means for correcting the detection signal of the gas obtained from various behaviors with the global characteristics of the gas sensor signal,
Gas detector. 2. The gas detecting device according to claim 1, wherein the means for extracting the global characteristic of the gas sensor signal is adapted to statistically quantify the degree of random fluctuation of the gas sensor signal. . 3. A means for extracting global characteristics of a gas sensor signal is characterized in that a past statistical data of a gas sensor signal is globally statistically compared with a gas sensor signal in the vicinity of the present time. The gas detector according to claim 1. 4. The means for extracting the global characteristic of the gas sensor signal is adapted to detect that the drift of the gas sensor signal is within a predetermined range. Gas detector. 5. A device for detecting a gas generated in a vehicle compartment by a gas sensor, for differentiating the gas sensor signal by at least one floor or more to detect the gas, and for detecting a random fluctuation of the gas sensor signal. And a means for controlling the air conditioner by combining the signal from the gas detection means and the signal from the random fluctuation detection means. 6. The air conditioner is an air purifier, and the detection target gas is a smoking gas.
The gas detection device described in 1. 7. An apparatus for detecting gas generated in a vehicle compartment with a gas sensor, for differentiating a gas sensor signal by at least one floor to detect gas, and for detecting a sudden drop in the gas sensor signal. And a means for correcting the detection of the gas by the gas detection means so as to suppress the detection of the gas after the detection of the sudden drop of the gas sensor signal. .